<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Market Therapy on Medium]]></title>
        <description><![CDATA[Stories by Market Therapy on Medium]]></description>
        <link>https://medium.com/@MarketTherapy?source=rss-f341c5cf49ec------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*bNP5srk3PDyXil5Netaxlw.png</url>
            <title>Stories by Market Therapy on Medium</title>
            <link>https://medium.com/@MarketTherapy?source=rss-f341c5cf49ec------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 16:59:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@MarketTherapy/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Bullish on Introspection. Don’t Be the Wile E of Wall Street.]]></title>
            <link>https://medium.com/the-investors-handbook/bullish-on-introspection-dont-be-the-wile-e-of-wall-street-8f19cd76d2d4?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/8f19cd76d2d4</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[trading-psychology]]></category>
            <category><![CDATA[a16z]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Mon, 30 Mar 2026 12:31:00 GMT</pubDate>
            <atom:updated>2026-03-30T17:12:56.462Z</atom:updated>
            <content:encoded><![CDATA[<h4>Marc Andreessen says introspection is a weakness. The AI arms race &amp; my own journey say otherwise.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WoVvFGskkM0T7Vso2eulfw.png" /></figure><p>We’re going to try something new with this piece. I’m not here to unpack a thesis behind a new position. There’s no earnings breakdown or price target being stress-tested. This will be a thought piece sparked by an <a href="https://x.com/davidsenra/status/2033297822691516853">X post</a> featuring one of the most widely known names in the VC arena. Because sometimes what I want to pass on to those that I invest for can’t be found in a discussion around a single ticker. It takes a broader discussion around how we think when we invest.</p><p><strong>Marc Andreessen</strong>, arguably the most influential venture capitalist in Silicon Valley, was asked recently about his levels of introspection. Although he blurs the lines between introspection and retrospection, his answer was clear: <a href="https://www.youtube.com/shorts/b6Zw50f5jJk"><strong>“Zero. As little as possible. Move forward. Go.”</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*07GPnIL4yT7S8prnxK-eaA.png" /><figcaption>Andreessen’s take on introspection on the Founders Podcast.</figcaption></figure><p>That stuck with me. The man’s Twitter is not for the politically faint of heart, but when it comes to building <strong>A16Z</strong> into a unicorn farm that got in early on Airbnb, Coinbase, and Databricks, there is no track record like it. And honestly? As someone who professionally came of age in the trenches of startups, speed and execution weren’t nice to haves. It was expected and many times required to keep the business afloat for the next 18 months.</p><p>However, when looking at all the CAPEX being burned in the AI race right now, it looks like we could use a bit more introspection. <strong>Meta</strong> burned <strong>$80 billion</strong> on a metaverse that only gave everyone the heebie-jeebies. <strong>OpenAI</strong> just axed Sora after a little more than a year. Meanwhile <strong>Apple</strong>, the company that even I knocked for not having a clear AI strategy at the time, is suddenly looking like the only strategic adult in the room.</p><p>So instead of following Andreessen’s advice, let’s go in the opposite direction. Let’s examine how three AI leaders are operating with varying levels of introspection, how my largest realized loss to date set me on the right path, and why the system I built from my list of failures matters to me more than any winning trade.</p><h3>The Cost of Just Moving Forward</h3><p>I want to be fair to Andreessen. I’m not bringing him into this piece just to dunk on his take. I fully recognize that he’s playing a completely different game where hitting on an early stage bet could cover all the losses -and then some. In early-stage startup land, you prototype your way into partnerships and product-market fit. I’ve lived it and get it.</p><p>But here’s why his comments land at such an interesting time and sparked some introspection, for lack of a better term. We’re in the middle of an AI arms race that many believe is starting to look a lot like a prisoner’s dilemma. Hyperscalers are pouring billions into infrastructure, launching products at breakneck speed, and racing to plant flags out of fear of missing out on any potential AI revenue.</p><p>Even for someone like me who’s been bullish on the underlying technology and the infrastructure powering it over the last few years, the recent wave of backtracking and billion-dollar write-offs has me wondering if the “move fast, break things” mentality is finally starting to run its course.</p><h4><strong>Meta</strong></h4><p>Because Meta isn’t a Series A and we’re not talking about burning through a seed round. It took four years and a <strong>$80 billion</strong> hole in Zuck’s wallet before he finally <a href="https://communityforums.atmeta.com/blog/AnnouncementsBlog/updates-to-your-meta-quest-experience-in-2026/1369435">pumped the brakes on Horizon World</a> and the Metaverse. And honestly? No matter how much money the company now prints from AI ads, Zuck’s lack of introspection makes me extremely nervous about their superintelligence lab and the next chapter for the Family of Apps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/898/1*kF9y26EHdxCTPTqEPYlMOA.png" /></figure><p>When push comes to shove, will he be able to learn from the past and cut his losses if his superintelligence lab packed with super expensive talent leads to a dead end? I’m not sure, which is why Meta stock is now falling into the “trade not own” territory for me.</p><p>That said, Zuckerberg recently announced he’s building an AI agent to<a href="https://www.wsj.com/tech/ai/mark-zuckerberg-is-building-an-ai-agent-to-help-him-be-ceo-eddab2d5?gaa_at=eafs&amp;gaa_n=AWEtsqetwJqL_PktQu5rECnpwOTtocsYjwa1pd7NqsG2ZX_6krmTTgGuEkXSj85pB_o%3D&amp;gaa_ts=69c8d797&amp;gaa_sig=MixPU-uzkJtr5taMg38huOXiIdL5E6FDVUuuRXMXrS0Bgh3w1HPz4ebqio5tjhH7gT1x3vG_eVTWD08Iw2GDaA%3D%3D"> help him make decisions as CEO</a>. As ironic and as meme-worthy as the headline is, if tapping an AI sidekick helps Zuck become more reflective, I’m game. Until then, Meta is fundamentally strong and deserves your attention, but I see the lack of introspection leading to more execution risk down the road. Especially if he’s on the verge of creating something new (like Metaverse), as opposed to acquiring an asset and scaling it like Instagram.</p><h4><strong>OpenAI</strong></h4><p>Now let’s look at OpenAI. I’ll give Altman a little bit of credit since some introspection is clearly better than none, but zoom out and you start to see how “move forward, go” has caused more harm than good. Sora is the most recent casualty but far from being the only one. <a href="https://archive.is/20260330112101/https://www.bloomberg.com/news/articles/2026-03-30/openai-s-chatgpt-app-store-took-aim-at-apple-but-results-lag-so-far">The GPT Store</a> never became the consumer-driven marketplace it was hyped up to be. DALL-E pioneered text-to-image and then watched Midjourney take the crown for years. And ChatGPT’s market share continues to slide as users opt for Claude and Gemini.</p><p>Altman killing off distractions for the team is the right move, but had he done so sooner, the door wouldn’t have been cracked open just enough for Claude and Gemini to walk right on through and take a seat at the table. With a two year gap between the release of GPT-3 in 2020 and other comparable models in 2022, it doesn’t look like Altman will be able to convert this lead into a compounding moat like NVIDIA has.</p><p>I don’t think OpenAI will have issues raising capital and having a historic IPO, but what I find more troubling is how introspection isn’t more deeply rooted in Altman’s DNA. The man led <strong>Y Combinator</strong> for years, helping early-stage founders steer past shipping too fast, spreading too thin, and losing focus. He also has <strong>Nadella</strong> and <strong>Microsoft</strong> in his corner, a company that understands distribution and scale better than anyone else. Yet, OpenAI continues to stumble while their competition closes the gap.</p><h4><strong>Apple</strong></h4><p>Then there’s Apple, which I’ve consistently hammered on for sitting on their hands while other hyperscalers had a bias towards action. The jury is still out on whether the move was strategic or serendipitous, but the outcome is undeniable. Apple is about to unleash AI across their <strong>2.5 billion</strong> network of active devices with limited scar tissue and an absolutely pristine balance sheet compared to its peers. And regardless of how much better ChatGPT or Claude is, I reckon most of us would gladly opt for a Gemini-backed Siri if it means native AI functionality across iCloud and all of our Apple devices.</p><p>I wrote last summer that <a href="https://medium.com/@MarketTherapy/not-a-core-holding-of-mine-0d2df54cd960">AAPL wasn’t a core holding</a> of mine because of the AI lag. I still don’t own it. However, I’m starting to recognize Cook’s ability to sit still and introspect as an edge over other tech leaders. I guess the most strategic move at times is to make no move at all, and Apple’s zen approach to AI now has the potential to shake things up for all of Big Tech.</p><p>I bring up these timely examples not to take cheap shots at tech CEOs, but to examine the human nature driving the decisions. Although these individuals operate on a much different scale than us retail investors, most of us should be quite familiar with the forces at play. We’ve all had to manage conviction, FOMO, and the reluctance to reverse course when our pride is on the line. Any seasoned investor knows these forces very well, and anyone who doesn’t simply has yet to pay their tuition in full.</p><p>I understand why Andreessen wants the CEOs in his portfolio to be untethered by introspection, especially at the stage where their paths cross. That said, the implications of AI on our society have raised the stakes substantially, and we’re now talking about CAPEX figures that are comparable to entire GDPs of small nations. Having zero introspection while leading the charge on such an impactful tech would be irresponsible.</p><p>As for those of us who aren’t hardwired to turn off our introspection at a moment’s notice, I refuse to believe that this should be seen as a hindrance or a weakness. Especially in the field of investing, I would argue that the ability to reflect on one’s journey is the only thing keeping us from becoming the Wile E. Coyotes of Wall Street.</p><h3>SMCI: Paying my Tuition</h3><p>To talk about how I learned to harness introspection and retrospection as an edge, we’d have to painfully rehash the details around one of my largest realized losses first. The time I lost <strong>$17,783</strong> investing in <strong>Super Micro Computer (SMCI).</strong></p><h4><strong>The Chase</strong></h4><p>This was back in 2024 and I had just scored some wins with big name data center players, and I was hungry to prove that I had the Midas touch. Prior to their most recent scandal, SMCI was at the forefront of building liquid-cooled rack-scale servers for data centers. They had also recently been tapped by the man in the leather jacket himself to become one of NVIDIA’s manufacturing partners for their next-gen AI accelerators.</p><p>Fueled by a burning desire to beat the rest of the herd to the watering hole, I chased the SMCI roadrunner off the edge of the cliff. Never mind the well-documented battle scars and countless promises I made myself to never chase a stock that had already ripped to all-time highs. Forget about the clearly marked retracement and entry zones that I had been mapping out along the way. Jensen’s partnership became a proxy for my own research.</p><p>I bought SMCI eight times in total, a handful of times even on the way down as a<a href="https://hindenburgresearch.com/smci/"> <strong>Hindenburg Research</strong> short report</a> exposed countless sketchy accounting practices. I had been up on my position around 20% at one point, and I wasn’t going to let a couple of salty bears talk me out of strapping myself to Wile E.’s infamous ACME rocket.</p><p>Spoiler alert: it ended how it always ended in the Sunday morning cartoons. Following the short report, all Big Four accounting firms refused to put their stamp of approval on SuperMicro’s books, and I lost 72% of my invested capital on the ticker. But as painful as that loss was, it needed to happen. I needed a trade to shake me to my core so violently that habitual introspection would become a permanent part of how I invest.</p><p>Which is why I don’t want this piece to be about judging tech CEOs or throwing stones at glass houses. I think most of us have some level of introspection when it comes to making decisions, but for whatever reason, our own systems and rules are so easily cast aside when we’re staring down what always seems like an opportunistic buy of a lifetime.</p><p>Having zero introspection might be freeing for some. For me, introspection is the hedge against my own human nature, and the only thing keeping me from stepping on the same rakes over and over again.</p><h4><strong>The Temptation</strong></h4><p>Shortly after the dust settled, Seeking Alpha blew up with Buy ratings as value investors piled back into SMCI. Forward P/E at 6.9x, a 63% discount to its three-year average. I would be lying if I told you that the human urge to get a revenge trade in and make up all the losses didn’t cross my mind countless times. Locking in the loss was difficult, but watching the stock make higher highs on its way to a 250% rally might have been the most painful part of the experience.</p><p>Fast forward to just earlier this month, after management was caught orchestrating a <a href="https://www.cnn.com/2026/03/19/politics/super-micro-computer-founder-charged-ai-chips-china"><strong>$2.5 billion</strong> scheme to smuggle NVIDIA chips into China</a> and employees were literally taking hairdryers to the serial numbers of NVIDIA GPUs to hide where they were going. SMCI is recirculating amongst investing circles, and even after the indictment, investors are once again calling it a bargain and a buy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XBzkO74hvbmO4R6mSibQSg.png" /><figcaption>SMCI employees were caught by the warehouse’s security survellance system.</figcaption></figure><p>This is important to call out, because even though I was able to chuckle and scroll right by all those bullish articles on SMCI this time around, I recognize that the temptation and human nature never fully go away. Introspection is useful in identifying gaps in your thinking, but it takes application and a system to actually benefit from the reflection.</p><p>The reps matter. And every time you journal a trade regardless of outcome, or recall a previous experience to fight off your impulses, you’re force multiplying your edge as an introspective investor. As your confidence grows in your system, that’s when all the noise gets quieter.</p><h4>The System</h4><p>I get it. Journaling and building a system sounds like a lot of work, especially when we’re now able to get in and out of multiple trades faster than we can sip our morning coffee. The friction between impulse and acting on it has never been lower. But I’d argue that the same can be said about AI and how it can just as easily remove some of the friction behind building your own guardrails.</p><p>Every psychological trap I’ve described in this piece is something I still grapple with every market open. None of that fully goes away with time and experience. What changed is that I built a system that holds a mirror up to my own decision making, recognizes traps and patterns that I’m still painfully oblivious to, and forces me to come face to face with my own convictions and experiences before locking in a trade.</p><p>When both CNBC and “finfluencers” can’t stop recommending a ticker, the system pulls up my own thesis from two weeks ago and puts it right in front of me. It helps me cut through the noise and trust in the time and research that I’ve committed to the stock. Same goes for when I’m on the losing end of a trade. My system surfaces all the receipts and reminds me of why I bought in the first place, why I ignored the red flags and developed diamond hands, and forces me to sit with a thesis that might not be working out. You’d be surprised how easily you can start seeing through thought traps when the full story is chronologically laid out in front of you.</p><p>And that’s why when Reddit went straight to meming Zuckerberg’s announcement of an AI twin, I actually thought it was a pretty good idea. As someone who’s bootstrapped a system together and has developed a better mindset because of it, I think most folks are dismissing what AI has to offer too quickly.</p><p>The popular take is that AI is great for trading because it strips away all emotion from the process. While that might be true, <strong>I think AI’s greatest contribution is that it gives you a memory that your ego can’t tamper with.</strong> Journaling in theory is a no-brainer, but logging entries takes time. Actually taking the time to review it is a whole other challenge.</p><p>AI now turns my transaction logs and notes into a living ledger. One that pattern matches across years of decisions, warns me when a thesis is stale, and effortlessly keeps track of all my analysis, regardless of whether it’s fundamental, technical, or macro narratives. For once, journaling can actually keep up with the pace at which we trade, and has a chance to beat our impulses to the punch when we’re about to pull the trigger.</p><h3>Closing Thoughts</h3><p>I started this piece because a very successful VC guy told his audience that introspection is a liability. After looking at how even the top companies stumble with the same advice and paying my own five-figure tuition on SMCI, I’d argue it’s the most undervalued edge any investor can hold.</p><p>The funny thing is, the more I refine the system, the more it’s changing how I think about my investing goals. P&amp;L and beating the market will always be a focus.<strong> </strong>Nowadays<strong>, every bad trade I don’t end up taking feels just as good as some of my largest gains. The process is becoming the point, and I’m becoming a stronger investor because of it.</strong></p><p>If any of you want to nerd out on trading psychology or are looking for someone to help you get started with a system, you know where to find me. Thanks for reading!</p><p>📩 <strong>Join </strong><a href="https://investorshandbook-newsletter-c72f94.beehiiv.com/"><strong>Investor’s Handbook Digest</strong></a><strong> — get the best investing, markets, and wealth-building insights each week.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8f19cd76d2d4" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/bullish-on-introspection-dont-be-the-wile-e-of-wall-street-8f19cd76d2d4">Bullish on Introspection. Don’t Be the Wile E of Wall Street.</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NVIDIA GTC 2026: From Keynote to Portfolio]]></title>
            <link>https://medium.com/the-investors-handbook/nvidia-gtc-2026-from-keynote-to-portfolio-29bcb0f21eba?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/29bcb0f21eba</guid>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[nvidia]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Thu, 19 Mar 2026 10:31:00 GMT</pubDate>
            <atom:updated>2026-03-19T10:31:00.640Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>Takeaways from Jensen’s keynote &amp; how it impacts my AI investment thesis.</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wInZ5NmCOKD7UR9GZyhgSQ.png" /></figure><h3>TLDR:</h3><p><strong>Tickers mentioned:</strong> NVDA, TSM, ASML, AVGO, NOW, CRM, GOOGL, AMZN, UBER</p><ul><li><strong>AI Infra is Alive &amp; Well:</strong> Jensen revealed a <strong>$1 trillion</strong> demand pipeline through 2027. 40% now comes from non-hyperscalers, and the demand is pulling through every layer of the supply chain. When demand is broadening and the whole ecosystem is posting records, the AI infrastructure trade is healthy.</li><li><strong>SaaS to Agents-as-a-Service: </strong>SaaSpocalypse<strong> </strong>was built on the fear that AI kills seat pricing. But SaaS to Agents-as-a-Service doesn’t weaken pricing power, it adds new levers. OpenClaw finally gives enterprises the agentic blueprint to do more with enterprise AI, and NemoClaw solves the critical security gap. IGV is still down 15% since the selloff but our thesis behind software hasn’t changed.</li><li>Jensen reframed performance per watt to <strong>revenue per watt</strong>, and SemiAnalysis benchmarks showed that ASICs still lag NVIDIA’s last-gen chips. With GTC wrapping up, it’s up to ASICs to throw the next punch, but for now the Groq acquisition seems to have cemented NVIDIA’s lead once again for the next three generations of chips.</li></ul><p>Jensen opened his <strong>GTC 2026 keynote</strong> by calling out all the traders and investors sitting in the auditorium. Finance was the largest attendee segment this year, and while he was happy with the engagement, he jokingly hoped that he was speaking to developers within financial services, not just a sea of traders hanging on to his every word. I couldn’t help but laugh as the man of the hour was basically telling me that I was not his target demographic.</p><p>But when <strong>NVIDIA</strong> is by far your largest holding, you don’t just skim the press release. You sit through the full two-hour keynote, bob your head to the dramatic Hans Zimmer-like background music during each product trailer, and try to decipher the product roadmap as if you were a developer who just happens to have a brokerage account full of NVDA stock.</p><p>Jensen has never been shy about where he thinks NVIDIA and the broader AI industry are headed. In fact, many of the biggest announcements at GTC 2026 were teased by him at the same event exactly one year ago. This time, however, he left absolutely nothing to the imagination.</p><p>He laid out the economics behind the AI factories that hyperscalers and governments are racing to build, revealed how CEOs are evaluating productivity and performance through the lens of compute, and showed exactly how the next three generations of NVIDIA hardware will power the shift. Here are my notes on the keynote and how I’m connecting the dots back to the AI investment thesis I’ve been building for the past two years.</p><h3>AI Infrastructure is Alive &amp; Well</h3><p>My first takeaway is the broadest but it also supports the entire thesis behind the portfolio. We’re now in year three of this AI buildout, and based on the demand signals shown at GTC 2026, AI infrastructure and chips continue to be one of the strongest sectors poised for growth.</p><p>Let’s start with the headlining number, then get to the much more interesting “why.” Jensen revealed visibility into a demand pipeline worth <strong>$1 trillion</strong> through 2027, doubling from $500B at last year’s GTC.</p><p>Again, I get all the skepticism surrounding AI, but look at the adoption and the shift in user behavior that got us here. Despite the sentiment around Sam Altman, ChatGPT is now approaching<strong> 1 billion</strong> weekly active users. Regardless of the negative connotation behind vibe-coding and the slop it creates, Anthropic’s Claude Code managed to hit a <strong>$2.5B run-rate</strong>. And then there’s OpenClaw, which we’ll cover in depth later, that accrued <strong>250K+ stars</strong> in two months because of the overwhelming demand for chatbots to do more than just hold a conversation.</p><p>Regardless of who provides the infrastructure, these are the demand signals from the enterprises and end users who’ve now had a taste of what AI can offer, and not only <em>want</em> but <em>expect</em> AI to do more. This requires inference, and inference requires tokens. It’s exactly why inference demand has grown roughly<strong> a million times</strong> according to Jensen, and how NVIDIA has managed to double their demand pipeline in just 12 months.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WnWXXA6SvSVBaROO8W5TOg.png" /><figcaption>Growth in the 40% pie is where I’ll be watching closely.</figcaption></figure><p>A year ago, this slide would have looked completely different as nearly all demand was driven by the hyperscalers. The only ones with the talent and balance sheets to justify their big bets on compute. Jensen has always claimed that commoditization of tokens would broaden demand, and this 2026 snapshot of the pipeline is proof.</p><p>60% is still covered by hyperscalers and the frontier labs, but it’s the other 40% that gives me confidence to continue investing within the ecosystem. Sovereign AI programs, industrial leaders, and enterprises across every industry that have successfully argued for bigger AI budgets are now lining up alongside <strong>Microsoft</strong> and <strong>Amazon</strong> for their share of AI-enabling hardware and a chance to turn their data centers into token factories.</p><p>What’s crazy is that Jensen later confirmed that the $1 trillion in demand only accounts for <strong>Blackwell</strong> and <strong>early Rubin</strong>. It doesn’t include <strong>Rubin Ultra, Feynman</strong>, or any of the other segments we’ve covered in the latest earnings report. With hyperscaler CAPEX already at <strong>$650B</strong> in 2026 (nearly 3x since 2023), and non-hyperscaler demand broadening quickly, the $1 trillion figure actually looks to be achievable and somewhat conservative.</p><p>And it’s not just NVIDIA reaping the benefits. The demand is flowing through every layer of the infrastructure stack. Upstream, <strong>TSMC</strong> is building additional fabs which is causing the backlog for EUV machines to grow over at <strong>ASML</strong>. On the power side, utilities and energy providers are experiencing record levels of customer commitments and partnerships. Even competitors like <strong>Broadcom</strong> and <strong>AMD</strong> continue to benefit as customers diversify their supply chain or try their hand at creating their own AI chips.</p><p>When a cutting-edge technology becomes more accessible and the entire supply chain supporting it is busy posting record numbers simultaneously, it’s hard to argue that the infrastructure buildout is peaking. This is exactly why AI infrastructure continues to be the foundation of my portfolio, and why Jensen’s keynote only strengthens this conviction.</p><h3>From SaaS to AaaS</h3><p>Six months ago, I published <a href="https://medium.datadriveninvestor.com/crm-the-unsung-hero-to-agentic-ai-5c31f6bb2952">“The Unsung Hero to Agentic AI,”</a> arguing that <strong>Salesforce</strong> wasn’t getting disrupted by AI, and in fact was going to grow as the context and distribution layer for agents. Then one week before SaaSpocalypse took out the entire software sector, I expanded my thesis and portfolio <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">to include other AI data players such as <strong>ServiceNow</strong> and <strong>Uber</strong>.</a> The timing could have been better, but I remain confident that the thesis will play out, especially now thanks to <strong>OpenClaw</strong>.</p><p>While I definitely could have come up with a better acronym or at least sound it out before putting it on the teleprompter, Jensen proudly proclaimed that every SaaS company will eventually become an <strong>AaaS (Agent-as-a-Service)</strong> company. Nice.</p><p>Now when I wrote those pieces, I knew agentic AI was going to be the next stop, but the path to get there was murky. Chatbots were awesome six months ago, but even the most successful tech companies couldn’t get them to do more on their own at scale across an entire business. Then <strong>Peter Steinberger</strong> built OpenClaw right when it seemed like enterprise AI was losing steam.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EcD3lnxfPqJsEMB33SN4rA.png" /><figcaption>OpenClaw hit 250K GitHub stars faster than React or Linux ever did.</figcaption></figure><p>While the open-source project went viral for entrepreneurs and personal use, it provided enterprises with a blueprint forward. AI can now call on LLMs, spawn sub-agents, and interact with other software directly. The industry went from building AI-wrapped custom tools to adopting their own OpenClaw strategy that includes their entire tech stack and database as context and rails for AI agents.</p><h4>An Upgrade to Pricing Power</h4><p><a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0">SaaSpocalypse</a> was built on the fear that AI would make the per-seat pricing model obsolete. However, with how the industry is already moving, this couldn’t be further from the truth:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Any87X1NntIdKJmawOIgmA.png" /><figcaption>Token pricing tiers show how AaaS scales revenue with work complexity, not headcount.</figcaption></figure><p>AaaS just gave software leaders another lever on pricing power. Look at the pricing tiers along the bottom of this chart. Basic inference, your vanilla chatbot, essentially becomes freemium. But with the help of NVIDIA hardware, CRM and NOW can charge a premium for the workloads that get more done. Agents that run 24/7, handle complex reasoning, and manage multi-step workflows autonomously can now be part of a premium tier.</p><p>Seat pricing was always easy to sell, but it’s capped by employee count. Token pricing scales with the complexity and volume of the work being cranked out. If a customer support agent costs X tokens, a financial modeling one now costs 10X. A multi-agent workflow that spans across the entire deal pipeline can now cost 100X. As each generation of NVIDIA hardware prints tokens faster and more efficiently, the margins for AaaS only get better, which ties back directly to my thesis behind software.</p><h4>The Security Roadblock</h4><p>But there’s been one major barrier holding enterprise adoption back, and it’s not technological. It’s security. Nobody wants proprietary data leaking over the web, and in our <a href="https://medium.com/the-investors-handbook/nvda-the-safest-seat-is-at-the-top-61610ceaec7e">coverage of NVIDIA’s latest earnings</a>, we flagged the <strong>RTX on-prem</strong> opportunity as a critical enabler for enterprises looking to adopt agentic workflows.</p><p>During the keynote, Jensen delivered exactly that. <a href="https://www.nvidia.com/en-us/ai/nemoclaw/?ncid=pa-srch-goog-788853&amp;_bt=800486735026&amp;_bk=nemoclaw&amp;_bm=p&amp;_bn=g&amp;_bg=196797973640&amp;gad_source=1&amp;gad_campaignid=23663275624&amp;gbraid=0AAAAAD4XAoEaby0-_kbGSOwuTa6mCiPH2&amp;gclid=Cj0KCQjwmunNBhDbARIsAOndKpm9OHFTUpONtBO52Avayvvrt21j4YIxiRn_hw_GFGFz6ohnQO_4YkAaAh6QEALw_wcB"><strong>NemoClaw</strong></a> now gives enterprises the policy engines, guardrails, and audit trails to deploy agents safely across their business. Pair the new security software layer with the RTX hardware, and the security risk behind enterprise AI just lost its teeth.</p><h4>Investing in the Data Layer</h4><p>And this is where it circles back to the data layer companies we’ve been writing about while SaaS was thrown to the wayside. Jensen called structured data “the foundation of trustworthy AI,” and as an ex-GTM systems guy, I couldn’t agree more. Agents are only as good as the data they can access and the systems they can plug into.</p><p>When customers no longer have to pick between ripping out their existing stack and adopting AI, the decision becomes much easier. The metrics behind early AaaS platforms speak for themselves. Salesforce’s <strong>Agentforce</strong> is up <strong>330% YoY</strong>. And with the help of their <strong>AI Control Tower platform</strong>, ServiceNow grants customers an orchestration layer across their IT stack, which is pushing their AI-related ACV to go beyond $600M.</p><p>And then there’s <strong>Uber</strong>, which most people still don’t recognize as an AI data play at all, but I find the partnership momentum to be extremely compelling. Instead of trying to pick winners out of Waymo, Tesla, and Zoox, why not just bet on the demand and data layer that every AV player needs? That’s exactly what Jensen did at GTC, <a href="https://nvidianews.nvidia.com/news/nvidia-announces-open-physical-ai-data-factory-blueprint-to-accelerate-robotics-vision-ai-agents-and-autonomous-vehicle-development">announcing a multi-city robotaxi partnership with Uber</a> and validating the platform as a critical layer of NVIDIA’s physical AI stack.</p><p>I’m not saying that agents are going to run the world tomorrow. But the path from chatbot to enterprise-grade agents just got clearer at GTC. <strong>IGV</strong> is still down over 15% since Anthropic dropped their legal plugin earlier this year, and the SaaS names in my portfolio felt every bit of that selloff. However, enterprise AI adoption isn’t slowing down, and the AV partnerships just keep stacking up. I remain bullish on AaaS, and I think the market will eventually catch up to what Jensen presented at GTC.</p><h3>Trading Per-Watt Punches</h3><p>Up until now, I’ve kept the focus broad for readers wanting investing takeaways that aren’t exclusive to NVIDIA. However, this is an NVIDIA conference, and in <a href="https://medium.com/the-investors-handbook/avgo-im-buying-the-gigawatt-gap-63178dc2c7a1">my recent Broadcom piece</a> I specifically called out that Jensen would need to respond to the ASIC efficiency threat. <strong>Performance per watt</strong> has become increasingly important with the power grid bottleneck and CAPEX concerns, and Broadcom’s <strong>XPUs</strong> were winning it by design. Well, in true Jensen fashion, the CEO responded by reframing the whole conversation and immediately turning up the heat.</p><p>According to Jensen, as SaaS transitions to AaaS, and agents become more common, performance per watt really becomes revenue per watt. As soon as a data center is built it is power-constrained since a 1 gigawatt facility will never become 2. With that constraint in mind, the only metric that matters becomes how many tokens you produce per watt, because tokens turn into outcomes that turn into revenue.</p><p><strong>SemiAnalysis</strong> ran a third-party <a href="https://newsletter.semianalysis.com/p/inferencex-v2-nvidia-blackwell-vs">AI inference benchmark</a> and the results were astonishing. NVIDIA’s latest <strong>Blackwell</strong> system delivered <strong>50x</strong> performance per watt versus <strong>H200</strong>, NVIDIA’s own last-gen chip. And here’s the knockout punch for the competing ASICs. Based on the benchmark, the chips that were supposed to challenge NVIDIA on efficiency haven’t caught up to what the company is already racing to make obsolete. This has major implications during a time when CAPEX is under strict scrutiny.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q5pO-g6k6zyvQyD5yG4H_A.png" /><figcaption>Blackwell delivers 50x higher perf/watt and 35x lower token cost. The competition can’t match last-gen H200 on either metric.</figcaption></figure><p>Jensen showed his back of the napkin math. A one gigawatt data center costs roughly $40 billion over 15 years regardless of what chips you put inside it. The cost of building and powering the data center doesn’t change. The only variable NVIDIA customers have control over becomes token production, which in turn is driven by the hardware in the racks. If your data center produces fewer tokens per watt, you’re undertaking a $40 billion infrastructure project just to get outperformed by a facility running NVIDIA’s last-gen hardware. As Jensen puts it, “If you have the wrong architecture, even if it’s free, it’s not cheap enough.” This $40 billion headache will be exactly what all the Big Tech CEOs will be grappling with when it comes time to filling up the racks and choosing their vendors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CjuU0mR5-aUR0Io1enexnA.png" /><figcaption>Revenue per gigawatt jumps from $30B on Blackwell to $300B with Vera Rubin and Groq LPX (10x leap).</figcaption></figure><p>And unfortunately for NVIDIA competitors, Jensen is already <a href="https://medium.com/datadriveninvestor/nvda-groqin-around-the-christmas-tree-ce0b19b6d6a4">putting the Groq acquisition to good use</a>. Just three months after the deal went through, <strong>LP30</strong> is in production and set to ship with Vera Rubin later this year. The heavy-hitting GPU will now have a more than capable counterpart to handle the rapid-fire workloads that agents demand. Instead of fighting the efficiency advantage that ASICs have always had over GPUs, NVIDIA simply absorbed it and you can see it everywhere on the roadmap with Ultra and Feynman. Again, neither of which are included in the $1 trillion pipeline.</p><p>For years, NVIDIA’s moat never felt seriously threatened. Then Alphabet’s TPUs gained traction, Broadcom stacked up six AI whale customers, and even the flip from training to inference worked against the GPUs strengths. GTC was the perfect platform for Jensen to throw down the gauntlet, and the team did exactly that.</p><p>Instead of letting ASICs dictate the rules of the game, NVIDIA showed up competing on revenue per watt, a metric NVIDIA defined, benchmarked, and is already two generations deep into with Groq’s help. Broadcom’s next-gen XPUs are in development and Alphabet’s TPU v6 is ramping, so the punches will keep coming. But for now, per usual, NVIDIA is the one setting the pace and the terms.</p><p>I hope my coverage of GTC has been worth a read. If you’d like to hear the dramatic background music or see the man in the leather jacket in action for yourself, check it out <a href="https://www.youtube.com/watch?v=jw_o0xr8MWU&amp;t=1921s">here</a>. He also did a <a href="https://video.ibm.com/recorded/134740555">Q&amp;A</a> where he broke down many of the concepts further for analysts.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=29bcb0f21eba" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/nvidia-gtc-2026-from-keynote-to-portfolio-29bcb0f21eba">NVIDIA GTC 2026: From Keynote to Portfolio</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AVGO: I’m Buying the Gigawatt Gap]]></title>
            <link>https://medium.com/the-investors-handbook/avgo-im-buying-the-gigawatt-gap-63178dc2c7a1?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/63178dc2c7a1</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[semiconductors]]></category>
            <category><![CDATA[broadcom]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[investing]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Wed, 11 Mar 2026 11:01:00 GMT</pubDate>
            <atom:updated>2026-03-11T17:48:47.840Z</atom:updated>
            <content:encoded><![CDATA[<h4>Broadcom’s CEO thinks Wall Street is hallucinating. After Q1 earnings &amp; checking his math, I think he’s right.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MQo9OWte5iJ3_RKstvrD3w.png" /></figure><h3>TLDR:</h3><p><strong>Companies Mentioned:</strong> AVGO, NVDA, GOOGL, META, AAPL, OpenAI, Anthropic, ByteDance</p><ul><li>Broadcom’s fifth consecutive double beat featured <strong>$19.3B</strong> in revenue (↑29% YoY) and a Q2 guidance of <strong>$22B</strong> that crushed the $20.56B estimate. CEO Hock Tan confirmed 6 AI whale customers and targeted 2027 AI chip revenue to be $100B+.</li><li>Back-of-the-napkin math points to the CEO potentially sandbagging once again. With gigawatt commits implying <strong>$150–200B</strong> in AI chip revenue alone, and current FY2027 estimates of <strong>$153B </strong>covering<strong> </strong>the <em>entire </em>business, I believe Wall Street’s models are still catching up.</li><li>The industry is shifting from raw buildout to efficiency at scale, and <strong>performance per watt</strong> is the new metric to watch. Jensen’s <strong>$20B</strong> Groq acquisition validates the specialization trend that Broadcom’s XPUs have been designed for.</li></ul><p><strong>Broadcom (AVGO)</strong> reported Q1 FY2026 earnings last week, and class was in session as <strong>CEO Hock Tan</strong> treated the analysts to an arithmetic lesson. When UBS analyst Tim Arcuri asked about margin compression, Hock fired back firmly: <em>“Hate to tell you that you must be a bit hallucinating.”</em></p><p>That’s right. He responded to the question on a live earnings call as if the analyst was Claude on a bad day. And honestly? After spending the weekend digging through the transcript and the power commitments behind their backlog (or as management calls their <strong>“line of sight”</strong>), the typically tight-lipped CEO might have a point.</p><p>AVGO is <strong>up nearly 200%</strong> in the last two years, and it’s been incredibly rewarding being a shareholder since 2021 as analysts rerated the stock over and over again to catch up with ASIC demand. While it is never easy to add to a position that has already taken off, Hock Tan’s latest insights into Broadcom’s future, paired with the <strong>~20% pullback</strong> from <strong>$415</strong> highs, presents a window of opportunity that I won’t be sleeping on.</p><p>Let’s cover the earnings highlights, the bottoms-up math behind the gigawatt buildouts, and break down why Broadcom has hyperscalers buying full racks and ordering their next generation of chips way before the current ones are deployed.</p><h3>Q1 FY2026 Earnings Highlights</h3><ul><li><strong>Top &amp; Bottom Beat:</strong> Revenue of <strong>$19.3B</strong> beat the $18.6B estimate (↑29% YoY, ↑7% QoQ). EPS of <strong>$2.05</strong> beat the $1.88 consensus (↑28% YoY). Q2 revenue guided to <strong>$22.0B</strong> (↑47% YoY), crushing the $20.56B consensus.</li><li><strong>Semiconductor Solutions:</strong> <strong>$12.5B</strong> (↑52% YoY), representing 65% of total revenue. AI semiconductor revenue of <strong>$8.4B</strong> (↑106% YoY), with Q2 AI revenue guiding to <strong>$10.7B</strong> (↑140% YoY).</li><li><strong>Networking:</strong> Now 33% of AI revenue in Q1, growing to 40% in Q2. Broadcom’s Tomahawk 6 and 1.6T DSP have cornered the market and are used whether or not the customer uses XPUs or GPUs.</li><li><strong>Capital Efficiency:</strong> <strong>$8.3B</strong> in operating cash flow on just <strong>$250M</strong> in CAPEX. Revenue grew <strong>$1.3B</strong> QoQ on declining OPEX.</li></ul><h4>Commentary</h4><p>Naturally, many investors throw the NVIDIA measuring stick at Broadcom when evaluating the stock. The ~33x forward P/E is indeed higher than NVIDIA’s ~25x, but keep in mind that this is a company that just funneled <strong>$1 billion</strong> in revenue to operating income while also lowering operating expenses. This is the fifth consecutive quarter of double beats, and while the financials are phenomenal, it was management’s commentary that made me add to my position.</p><h3>Adding Up the Gigawatt Gap</h3><p>Hock Tan isn’t one to volunteer the underlying details of his business, but he uncharacteristically laid it all out on the call. As if he was Thanos collecting the Infinity Stones, he revealed that <strong>Google, Meta, Anthropic, OpenAI</strong>, and two other undisclosed AI whales (speculated to be <strong>Apple</strong> and <strong>ByteDance</strong>) are all fighting over Broadcom’s capacity. According to the CEO:</p><blockquote>“We have line of sight that our revenue in ’27 will be <strong>significantly in excess of $100 billion</strong>. I’m focusing on the fact that these are pretty much all based on chips, whether they are XPUs, whether they are switch chips, DSPs, <strong>these are silicon content.</strong>”</blockquote><p>Anyone who’s been following Hock Tan knows that he consistently sandbags and overdelivers. With the weight of all six AI whales behind him, he’s signaling that the $100B+ isn’t padded with software or services. It’s all silicon. Everything beyond that is icing on the cake.</p><p>When Bernstein’s Stacy Rasgon asked Hock Tan to show his work, he obliged, confirming that Broadcom expects to ship around <strong>10 gigawatts</strong> of AI compute in 2027 across the 6 customers. When Rasgon estimated that the company would be receiving <strong>$15–20 billion per gigawatt</strong>, the CEO didn’t bat an eye. This implies that the company now has line of sight into <strong>$150–200 billion</strong> in chip revenue alone. Old habits die hard, and even with all the infrastructure commits and their capacity tied up through 2028, Hock Tan is once again busy sandbagging and tempering expectations even though they have been tapped as the premier co-designer for custom chips.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yf7TNJ-iGXCZY9p4FhiTIQ.png" /><figcaption>Generated by Author with Gemini.</figcaption></figure><p>Here’s the gap and the opportunity. Wall Street’s FY2027 consensus for total Broadcom revenue currently sits around <strong>$153 billion</strong> across 45 analysts. That estimate is for the entire company. The gigawatt math we just walked through implies <strong>$150–200 billion in AI chips alone</strong>, before adding any software or non-AI semiconductor revenue into the mix. Simply put, the analysts are slowly catching up to the AI demand, but their current models are effectively zeroing out all of Broadcom’s other segments.</p><p>It’s a setup that reminds me of NVIDIA back in May 2023, when Jensen guided Q2 to <strong>$11 billion</strong> versus the analysts’ <strong>$7.2 billion</strong>. I see Broadcom entering that same phase now, and as we’ll cover later, I expect the core business to funnel growth into all other segments as they progress deeper into vertical integration and co-design.</p><h3><strong>Efficiency Over “One Size Fits All”</strong></h3><p>With CAPEX budgets under intense scrutiny and power becoming the leading bottleneck across every data center project, <strong>performance per watt</strong> has quietly become the headlining metric hyperscalers are focused on when evaluating their AI infrastructure. Unlike GPUs that have largely remained “one size fits all” across multiple tech trends, the architecture of XPUs wins this metric by design.</p><p>If you recall from our <a href="https://medium.com/@EmulsifiedInvestor/nvda-groqin-around-the-christmas-tree-ce0b19b6d6a4">article on NVIDIA’s Groq acquisition</a> back in January, I covered how Meta’s <strong>MTIA chips</strong> (co-designed with Broadcom) cut total cost of ownership by <strong>44%</strong> compared to GPUs and performed up to <strong>90% more efficiently</strong> in live production. XPUs strip out all the extra silicon and circuitry that GPUs carry, and the result is more inference per watt, lower cooling costs, and a smaller power footprint per workload.</p><p>Google needs chips tuned for Search across billions of daily queries. If Apple is indeed one of the mysterious customers, they’ll need Broadcom to officiate the marriage between Gemini and Siri. ByteDance, the other rumored spender, is undoubtedly looking to replicate Meta’s efficiency gains across their own ad and revenue platforms. Last but not least, Anthropic is busy printing cash and crashing sectors, but margins appear to be buckling under the weight of running Claude at scale:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/665/1*fv1uSVFrQk-4DbrKRW5ORA.png" /></figure><p>Multiply the performance per watt across all of their networks and the demand behind Broadcom becomes clear. Here’s Hock Tan’s take:</p><blockquote>“The one size fits all of a general purpose GPU gets you only that far… the design starts to depart from what is the traditional standard GPU design, which is why XPUs will eventually be more the choice.”</blockquote><p>Jensen’s acquisition of Groq’s LPUs was an attempt to get ahead of this exact specialization and efficiency trend. While I’m guessing that NVIDIA will have a direct response to the ASICs threat during GTC next week, the <strong>$20 billion</strong> vote of confidence confirms how the industry is moving, and more importantly, how Broadcom potentially has an edge for the next leg of the AI race.</p><h3>Unraveling the Receipts</h3><p>We’ve covered the size of the opportunity and why XPUs win on efficiency. Now let’s look at how these customers are spending with Broadcom, as their behavior tells you as much as the numbers do.</p><p>Anthropic is currently consuming <strong>1 GW</strong> of Broadcom-made silicon and ramping to <strong>3+ GW in 2027</strong>. Instead of shipping chips or switches to the AI lab’s data centers, Broadcom is shipping end-to-end <strong>Ironwood Racks</strong>, fully optimized out of the box. If that sounds familiar, it should. It’s a page right out of Jensen’s playbook, and it’s how NVIDIA continues to squeeze revenue out of their Data Center segment.</p><p>As NVIDIA broadened their reach to own more of the rack, <a href="https://medium.com/p/b0d638e9c742">networking and software revenue followed</a>. Broadcom is now running the same play. While shipping entire racks may compress margins in the short term, owning more of the stack means increased stickiness and opportunities to funnel revenue beyond just silicon and into networking and software.</p><p>In addition to shipping racks, Hock Tan revealed on the call that his customers are now developing <strong>two chips simultaneously</strong>:</p><blockquote>“We are seeing very rapidly more for those customers who are much more matured… that they will start to develop 2 chips each year simultaneously, one for training, one for inference.”</blockquote><p>Thanks to the flip from training to inference, and the demand for incremental improvements from AI users, Broadcom’s line of sight into revenue per customer is essentially doubling. While Wall Street is starting to get a handle on the high-level numbers, I think the depth of customer engagement and the trickle down effect to the rest of the business has yet to be factored into their estimates.</p><p>Zooming out to the big picture, for anyone still trying to call the top of AI infrastructure spending, this is your latest signal. The big spenders that are running the models today and most likely footing the power bill tomorrow aren’t being conservative. Instead, they are doubling down on custom silicon and stacking up on multi-generation commitments in search of efficiency and better unit economics. And they are building all of this with Broadcom as their preferred partner.</p><h3>Closing Thoughts</h3><p>I confidently own both NVDA and AVGO as two of the best AI infrastructure names. I expect NVIDIA to continue being steps ahead of everyone else, but I also respect the hyperscalers’ desire to forge their own path with the more than capable Broadcom. It’s two sides of the same coin, and a trade I’ll remain in as long as the infrastructure build out shows no sign of slowing down.</p><p>While Broadcom’s customer base has some of the deepest pockets we’ve ever witnessed, any pullbacks may break the math. However, with supply locked through 2028, a long track record of execution, and a CEO who has made a career out of sandbagging and overdelivering, I like where the risk-reward sits from here.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*5tF2KtuHj68hnmRX6TejZQ.png" /><figcaption>My AVGO position going back 1 year.</figcaption></figure><p>With my cost basis sitting around <strong>$195</strong>, and my latest add being at <strong>$345, </strong>the<strong> </strong>2x increase in sticker price doesn’t feel great. However, for reasons we covered above, I’m long Broadcom and remain a big fan of Hock Tan.</p><p>Thanks for reading!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=63178dc2c7a1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/avgo-im-buying-the-gigawatt-gap-63178dc2c7a1">AVGO: I’m Buying the Gigawatt Gap</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NVDA: Letters From the Infrastructure Graveyard]]></title>
            <link>https://medium.com/the-investors-handbook/nvda-letters-from-the-infrastructure-graveyard-b0d638e9c742?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/b0d638e9c742</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[trading]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Thu, 05 Mar 2026 21:53:23 GMT</pubDate>
            <atom:updated>2026-03-05T21:53:23.722Z</atom:updated>
            <content:encoded><![CDATA[<h4>History hasn’t been kind to Tech’s heavy lifters. Why I’m still long NVIDIA &amp; the AI megatrend.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L5PNwKyul_CeGobd2Q5x5Q.png" /></figure><h3>TLDR:</h3><p><strong>Tickers mentioned:</strong> NVDA, INTC, AMD, GOOGL, TSM, UBER</p><ul><li>Nadella’s “winner’s curse” warns that the companies doing the heavy lifting often watch the platforms built on top of them collect all the rent. Railroads, fiber optics, and Intel all prove the pattern is real.</li><li><strong>Overbuilding:</strong> Jensen is creating additional layers of demand with every new era of the buildout, but if demand ever teeters, history hasn’t been kind to those laying the tracks.</li><li><strong>Commoditization:</strong> What happens to NVIDIA’s pricing power if AMD, Alphabet, and all the ASICs close the performance gap? If vertical integration and extreme co-design stop being a differentiator, compute could become the next buried glass.</li><li><strong>Execution risk:</strong> Intel had the same moat, margins, and dominance for 30 years. Betting that any leader stays right for the next 30 years is a foolish bet against history.</li><li><strong>My bull case:</strong> Only 9.7% of US firms have adopted AI, sovereign AI is just ramping, and NVIDIA’s capacity has been sold out a year in advance going on three years. NVIDIA is a “platform company, not a chip company,” and is already future-proofing growth with sticky software lock-in, vertical integration, and a release cycle that gives them visibility into where the next workload goes.</li></ul><p>I don’t know about you all, but I get kind of paranoid when it comes to investing. Especially when things are going well. <strong>Jensen Huang</strong> just put on a Bad Bunny Super Bowl quarter, and how did I choose to celebrate? By publishing <a href="https://medium.com/the-investors-handbook/nvda-the-safest-seat-is-at-the-top-61610ceaec7e">my earnings analysis</a> and immediately booking a weekend getaway to the nearest NVIDIA bear camp.</p><p>You’ve seen all the memes on <a href="https://www.bloomberg.com/graphics/2026-ai-circular-deals/?embedded-checkout=true">circular financing</a> and we’ve all done our fair share of speculation on the China export situation. I’ll leave it to the Reddit crowd to cover those risks. Instead, I’d like to revisit something we covered in the beginning of the year called the <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">“winner’s curse,”</a> and why it doesn’t always pay to be the first mover or the premier picks and shovels enabler.</p><p>As <strong>Satya Nadella</strong> originally framed it, AI labs are plagued with this curse because they are potentially doing all the heavy lifting of building an incredible LLM, only to be <strong>“one copy away from that being commoditized”</strong> and watching the platforms built on top of their work reap all the benefits and collect all the rent.</p><p>While I remain bullish on NVIDIA and will share why I don’t think the company will suffer from the winner’s curse, I think it’s worth exploring the long and bloody timeline of companies that called the future correctly, put in the blood, sweat, and cash to build out the infrastructure, and still ended up as a sacrificial lamb for those that came after them.</p><h3>The Builders That Lost it All</h3><p>Railroads made America and still went bankrupt. <strong>Eighty million miles</strong> of underground fiber optic cable was laid to set the stage for the internet boom, but does anyone remember <strong>WorldCom</strong> or <strong>Global Crossing</strong>, the companies that did most of the work? And from this era, Intel won the chip race for three decades but is now relying on lifelines from the US government and NVIDIA to stay relevant. Three different eras, three critical technologies, yet all shared similar outcomes.</p><p>This is why the winner’s curse haunts me. All of these companies were right on the money and had the same conviction that a lot of the semiconductor and AI infrastructure leaders are touting right now. Hyperscaler CAPEX has already been revised <strong>up beyond 60% YoY</strong>, with JPMorgan projecting over <strong>$5 trillion</strong> in data center spend over the next five years. The order of magnitude of the capital being deployed makes me squeamish, especially when considering that history has not been kind to the ones shouldering the infrastructure build out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/540/0*3Y2xZNdAlfJhe1qu" /><figcaption>CAPEX continues to ramp to fund the AI infra build out.</figcaption></figure><p>So what went wrong? Let’s take a deeper look at each technology mentioned above as a case study for the patterns that kept reappearing throughout history.</p><h4>1. Overbuilding (Railroads)</h4><p>Every infrastructure boom sprouts from a demand thesis that has it all. A bright and shiny new technology that solves real pain points and promises exponential growth. However, builders consistently overestimate how fast demand materializes, and how much infrastructure is actually needed. It’s a common thread that is present throughout all three examples, but let’s focus on the American railroad expansion first.</p><p>I’m not disputing that America’s railroads literally paved the way for industrialization down the line. The thesis was correct. However, builders eventually stopped building for demand as they began racing one another to claim territory and chase political influence. When the <strong>Panic of 1873</strong> hit, 121 railroad companies, or a third of all operators, went bankrupt within five years. By 1893, 30% of all railroad mileage in the country had been seized by creditors, representing <strong>$1.8 billion</strong> in losses at a time before AI CAPEX figures had desensitized us all.</p><p>Track mileage <strong>tripled</strong> between 1860 and 1880. Over 70,000 miles of track were added, and with all the excess capacity flooding the market, <strong>freight rates fell by over 80%</strong> from their peak. While the ones that laid the tracks became footnotes in bankruptcy filings, industrialists like Andrew Carnegie took the checkpoint and built generational enterprises on the back of cheap transportation he didn’t pay a dime for.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1SWulUp-d1NoLHXdspyTtw.png" /><figcaption><a href="https://fred.stlouisfed.org/series/A02F2AUSA374NNBR">https://fred.stlouisfed.org/series/A02F2AUSA374NNBR</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/829/1*gfAaWMccVngfB6KACPmktw.png" /><figcaption>Builders went belly-up but the infra they built generated 40x the value in GDP. Source: Exponential View</figcaption></figure><p>Jensen is hard at work, creating additional layers of demand with every new era of the AI buildout: training, inference, physical, and sovereign. But if demand ever teeters under the weight of the buildout, history says that the builders pay he ultimate price.</p><h4>2. Commoditization (Fiber)</h4><p>Time and time again, infrastructure is built, supply catches up, and the product of the buildout becomes a commodity. When the only thing separating providers is price, it quickly becomes a race to the bottom, no matter how important of a building block the technology is. When this happens, <strong>this arms race</strong> <strong>rarely translates into value for the company creating it</strong>. Just ask the companies that buried all that fiber.</p><p>So what exactly kicked off one of the largest infrastructure buildouts in US history? A single spreadsheet. A WorldCom capacity planner’s model claimed that internet traffic was going to double every 100 days. The <strong>Telecommunications Act of 1996</strong> had just deregulated the industry, and suddenly every carrier and newcomer wanted a piece of the fiber optic pie. The result was a gold rush of debt-financed projects that buried over <strong>$500 billion</strong> worth of fiber optic cable across the country. Unfortunately for all the carriers, traffic only doubled once a year, meaning that the math was <a href="https://www-users.cse.umn.edu/~odlyzko/doc/itcom.internet.growth.pdf">off by multiple factors.</a></p><p>Overbuilding infrastructure rears its head here too, but here’s where commoditization became the nail in the coffin. B<strong>andwidth is bandwidth</strong>. There was no proprietary advantage within the strand of glass, and once supply caught up to demand, prices fell by 90%. Once fiber became a commodity, WorldCom’s logo was a non-factor and it didn’t matter who laid the cable first or who laid the most.</p><p>Ironically, <strong>over 85%</strong> of those fiber lines laid dormant for years, only to become the backbone of our digital economy decades later. Once again, the application builders showed up fashionably late and inherited the infrastructure for free. This freebie went on to help build all of Big Tech.</p><p>In terms of NVIDIA, what happens to their pricing power if <strong>AMD</strong>, <strong>Alphabet</strong>, and all the ASICs end up closing the performance gap in a couple years? Today, NVIDIA’s hardware is clearly a cut above the rest. But <strong>Moore’s Law</strong> once felt permanent too. If vertical integration and extreme co-design stop driving gains, NVIDIA’s margin reacceleration stops in its tracks, and compute could easily become buried glass.</p><h4>3. Execution Risk (Intel)</h4><p>Let’s diverge from the winner’s curse in the traditional sense and offer up a antithesis to my Q4 takeaway: <a href="https://medium.com/the-investors-handbook/nvda-the-safest-seat-is-at-the-top-61610ceaec7e">that NVIDIA’s position of power makes it one of the safer AI investments</a>. <strong>Intel</strong> had the same moat, the same ecosystem, and the same data center margins. Much like NVIDIA today, Intel was winning in every computing category and kept the winning streak alive for 30 years. But even the deepest moats can’t always insulate a company from execution risk, especially when you stretch the timeline out by decades.</p><p>Intel designed <strong>x86 CPUs</strong> to dominate the PC era in the late 1970s, then leveraged that installed base into servers when enterprise computing took off. The strategy paid off, as every laptop, server, and cloud rack ran on Intel chips. For all intents and purposes, much like NVIDIA is for AI, Intel was the premier picks and shovels player that was in a league of their own during the PC and computing era.</p><p><strong>Then smartphones exploded and the workload shifted.</strong> Although Intel had the <strong>XScale</strong> chip ready to serve that market, <strong>CEO Paul Otellini</strong> sold it off in 2006 because it had razor-thin margins compared to their server business. Unfortunately for Otellini, the mobile market would become several times bigger than PCs, and Intel had zero skin in the game.</p><p>When AI took center stage years later, GPU specialization in parallel computing became a hot commodity, and Intel once again couldn’t serve the demand with their CPUs. While competitors like AMD was able to claw back CPU market share while simultaneously bringing their<strong> MI series GPUs</strong> to market to compete directly against NVIDIA for AI workloads, Intel struggled to keep up with the pace of innovation. Their foundry business has been plagued with the same struggles, as a botched transition to 10nm put them light-years behind <strong>TSMC</strong>, an overseas competitor that now prints cash making chips for the entire world.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/775/1*WIxWnpQu8VPkPk7OIvRabQ.png" /></figure><p>As an NVIDIA investor, it’s hard to imagine that the Jensen-led company could one day fall behind as well. I felt compelled to include Intel as a case study in this article because the company also looked impenetrable during an era where they were on top of their game. Could AI suddenly bring about another workload shift where NVIDIA is a half-step behind AMD or Broadcom? While it certainly feels like Jensen has a crystal ball hidden underneath the black leather jacket, betting that any leader stays right for the next 30 years is a foolish bet against history.</p><h3>Why I’m Still Long NVDA</h3><p>After all of that, you might be wondering why NVIDIA is still my largest holding. I covered the full bull case in my Q4 earnings analysis, so I won’t rehash it here. Instead, I want to give my honest take on the risks we just discussed earlier.</p><h4>On Overbuilding</h4><p>Out of all the risks covered in this piece, I think the overbuilding comparison is the weakest bear case against NVIDIA. Unlike the railroad and fiber companies, NVIDIA isn’t the one at risk of holding the bag. But more importantly, NVIDIA isn’t passively waiting for demand to materialize. Since the AI buildout began three years ago, the company has been actively shaping where the entire sector moves next, and therefore has a direct hand in creating demand.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/474/1*xblVYxZYht7u4updxHYdrQ.png" /><figcaption>NVIDIA Investor Presentation 2024</figcaption></figure><p>Although the markets have already priced in AI’s impact on enterprise software, <strong>only 9.7% of US firms have adopted AI</strong> according to <a href="https://www.anthropic.com/research/anthropic-economic-index-september-2025-report"><strong>Anthropic’s Economic Index</strong></a>. Autonomous vehicles have made massive progress, but NVIDIA’s partnership with Uber and Stellantis to put Level 4 vehicles on the road doesn’t even start until 2027. Then there’s sovereign AI, where most nations are still in the early stages of turning ambitions into infrastructure. Just last year, the EU unveiled a<a href="https://commission.europa.eu/topics/competitiveness/ai-continent_en"> <strong>€200 billion</strong> <strong>AI Continent Action Plan</strong></a>, and now over 50 countries are building sovereign AI on NVIDIA systems.</p><p>We’re now going on year three where NVIDIA’s next-gen capacity has been sold out a year in advance. And in just <a href="https://www.nvidia.com/gtc/?ncid=pa-srch-goog-225-ret-rsa-en-us-3-l1&amp;_bt=794864499359&amp;_bk=gtc%202026&amp;_bm=b&amp;_bn=g&amp;_bg=193067419355&amp;gad_source=1&amp;gad_campaignid=23502881208&amp;gbraid=0AAAAAD4XAoE3gclztQwiu0vX9ZH8rqvZu&amp;gclid=Cj0KCQiA8KTNBhD_ARIsAOvp6DJDlpnjOm6bMnUxauJzyX9BuZuC-ScH2mO3AXuybgiIdY_MJg2oxFUaAmJZEALw_wcB">two weeks at GTC</a>, Jensen is set to unveil what he’s calling “chips the world has never seen before.” Overbuilding means supply gets ahead of demand. In my eyes, NVIDIA has the opposite problem, and until that reverses, the demand risk is moot.</p><h4>On Commoditization</h4><p>Any AI investor knows that NVIDIA hardware is the furthest thing from being interchangeable. At least not while Jensen keeps finding ways to squeeze performance gains out of the company’s full stack solutions. But for the sake of the argument, let’s entertain a scenario where the performance gap narrows and compute starts looking more like a utility.</p><p>Now hidden within Q4’s massive <strong>$62.3 billion</strong> Data Center segment sits a budding <strong>AI Enterprise</strong> segment that crossed the <strong>$1 billion</strong> run rate two years ago. It’s a managed software subscription layer that is fully integrated with <strong>CUDA</strong> and helps enterprise customers deploy, manage, and scale AI applications on their shiny new NVIDIA hardware. At <strong>$4,500 per GPU per year</strong> in subscription fees, I’ll let the math and opportunity speak for itself.</p><p>Think of their flagship product, <a href="https://developer.nvidia.com/nim"><strong>NVIDIA NIM</strong></a>, as a ready-made meal kit that comes with all the premium ingredients, except what’s for dinner is a full serving of fully optimized inference out of the box that comes tailored to your stack. In a sector where <strong>time-to-deployment is a competitive edge</strong>, getting up and running in days instead of quarters makes a compounding difference, and I fully expect this segment to expand in coming quarters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/650/0*gapZlQIaHfMhed1h.png" /><figcaption><a href="https://developer.nvidia.com/blog/">https://developer.nvidia.com/blog/</a></figcaption></figure><p>One can argue that AI accelerators are already on their way to becoming commodities, but can you really make the argument that NVIDIA’s installed base as a whole can be replaced? CUDA is 100% free to use, but <strong>4 million developers </strong>and every major machine learning framework have been building on top of it for over a decade. Even if AMD closes the hardware gap tomorrow, you’re asking NVIDIA customers to rebuild all of the optimization from scratch, and essentially throw out years of performance tuning that can’t just be forked like a Github repo.</p><p>I’ve been following Jensen pretty closely these last couple of years, and I noticed that the CEO now explicitly calls NVIDIA a “<strong>platform company, not a chip company.</strong>” It demonstrates that Jensen is aware of the winner’s curse and is already positioning accordingly. Just as Nadella has a <strong>49% stake in OpenAI</strong> and <strong>Microsoft 365</strong> to cover all their bases, Jensen is looking to build the same kind of coverage. Hardware grants NVIDIA pricing power today, CUDA ensures customer lock-in, and AI Enterprise is a bet on platform capturing more value tomorrow when dynamics shift.</p><h4>On Execution Risk</h4><p>If there’s one thing that stands out about Jensen, it’s that <strong>he competes with himself harder</strong> than anyone else within the sector competes with NVIDIA. They were the first to adopt an <strong>annual release cycle</strong>, setting a new standard for the pace of innovation, but also <strong>rendering their own hardware obsolete</strong> long before even their own customers have had the time to get set up. From <strong>Hopper to Blackwell to Rubin</strong>, each generation is designed to make the last one irrelevant, and I believe Jensen will shake things up once again at GTC.</p><p>Then there’s the <strong>vertical integration</strong>. NVIDIA didn’t just stay in its lane like Intel did with x86. When Moore’s Law no longer applied, they took it upon themselves to redesign the entire stack. Jensen calls it <strong>“extreme co-design”</strong>, but he’s selfishly just bringing more components of the data center in-house, and trusting his engineers over anyone else in the field.</p><p>While the annual release cycle and vertical integration create undeniable performance gains, it also <strong>compounds the sheer amount of proprietary knowledge</strong> the company gains with each generation of hardware. Execution risk stems from not knowing what you don’t know. What actually works, what breaks, where all the hidden landmines are. This expertise across hardware and software separates NVIDIA from those that get left behind, and Jensen never needs to chase the next workload since he’s the one steering the industry towards the next big thing.</p><h3>Closing Thoughts</h3><p>The winner’s curse is real. Every NVIDIA investor should spend some time in the graveyard of infrastructure companies that predicted the future but still ended up losing it all. I wrote this piece because I wanted to stress test my own conviction, and I hope it provided you with something to chew on as we stay invested through 2026.</p><p>NVIDIA has real exposure to the patterns that killed previous infrastructure players, but I remain confident that Jensen is mitigating every single one of them. Even so, those of us holding NVDA should be keeping tabs on the gross margin trajectory from Blackwell to Rubin, how Google’s TPUs and AMD’s MI Series stack up against Jensen’s upcoming lineup, and whether the full stack solution remains as sticky as it’s been.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/851/1*zmZcJ5hNck6-CARns9pq_g.png" /><figcaption>A snapshot of my NVDA holding going back 1 year.</figcaption></figure><p>My current cost basis for NVDA is around $92, and I continue to pick up shares when it trades down to around the $170s. The stock has more or less remained flat between $170-$200 for months, but I’m convinced that the fundamentals and upcoming announcements at GTC should eventually overtake the overhead resistance and bearish narrative on AI.</p><p>Like Nadella, the winner’s curse haunts me, but Jensen hasn’t given me a reason to flinch. Yet.</p><p>Thanks for reading!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b0d638e9c742" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/nvda-letters-from-the-infrastructure-graveyard-b0d638e9c742">NVDA: Letters From the Infrastructure Graveyard</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[NVDA: The Safest Seat is at the Top]]></title>
            <link>https://medium.com/the-investors-handbook/nvda-the-safest-seat-is-at-the-top-61610ceaec7e?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/61610ceaec7e</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Fri, 27 Feb 2026 13:01:02 GMT</pubDate>
            <atom:updated>2026-02-27T16:23:38.943Z</atom:updated>
            <content:encoded><![CDATA[<h4>Q4 earnings deep dive &amp; why NVIDIA is still the safest bet in AI.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vUllvBINGObssf7oK0l2CQ.png" /></figure><h3>TLDR:</h3><p><strong>Tickers Mentioned: </strong>NVDA, MSFT, AMZN, GOOGL, META, TSLA</p><ul><li>NVIDIA posted an incredible top and bottom beat with margins reaccelerating, but the real showstopper was the <strong>$78B</strong> in Q1 guidance.</li><li>While hyperscalers are burning cash on data centers and power plants, NVIDIA collects the checks. Asset-light, <strong>60% operating margins</strong>, and a way lighter CAPEX burden.</li><li><strong>$120B</strong> in annual net income is being recycled into extreme co-design and vertical integration. Every acquisition makes the stack stickier. Mellanox turned into a <strong>$31B</strong> segment. Groq’s team brought inference talent. Rinse and repeat.</li><li>Waymo, Uber, Tesla, and Amazon are all building with NVIDIA. No matter who wins, NVIDIA benefits the most from AVs and physical AI.</li><li>The stock is currently cheaper than Walmart. The market is pricing 73% growth with 75% gross margins like day-old cupcakes at the grocery store bakery. I’m buying every last one.</li></ul><p>With Satya, Sundar, Zuck, and Jassy all setting up their direct deposits straight into Jensen’s bank account during their earnings calls, NVIDIA’s recent earnings beat was simply a formality. Yes, somewhere along the line, <strong>$68.1 billion</strong> in Q4 revenue and a <strong>$3 billion</strong> beat over consensus has become business as usual for the world’s most valuable company.</p><p>What’s not business as usual is the Q1 guidance of <strong>$78 billion</strong>, and the visibility the man in the black leather jacket gave into how he plans to keep growing a business that so many believe has already peaked.</p><p>We’ll of course dig into the stellar numbers. GPUs powering the entire AI mega trend. Gross margins accelerating back to <strong>75%</strong>. We’ll even cover a sleeper segment that grew by <strong>159% YoY</strong> and might have Fortune 500 companies rethinking their cloud strategy. Most importantly, I want to lay out a thesis around why I think NVIDIA is paradoxically the best AI stock to own, but also one of the safest.</p><h3>Q4 FY 26 Earnings Highlights</h3><p>Let’s kick things off with the numbers, followed by the big picture implications:</p><ul><li><strong>Top &amp; Bottom Line Beat:</strong> Revenue of <strong>$68.1B</strong> beat the $65.56B estimate (↑73% YoY, ↑20% QoQ). EPS of <strong>$1.62</strong> beat the $1.52 consensus. Full-year revenue hit <strong>$215.9B</strong> (↑65% YoY) with net income of <strong>$120.1B</strong>.</li><li><strong>Data Center:</strong> <strong>$62.3B</strong> (↑75% YoY, ↑22% QoQ). Hyperscalers represented over 50% of the revenue. More on Blackwell and Vera Rubin later.</li><li><strong>Networking:</strong> <strong>$11B</strong> (↑263% YoY, ↑34% QoQ). <strong>Spectrum-X</strong> alone is now a <strong>$10B+</strong> annualized business. The segment brought in <strong>$31B </strong>annually, up 10x from the year they acquired Mellanox to get skin in the game.</li><li><strong>Professional Visualization:</strong> <strong>$1.32B</strong> (↑159% YoY). This is the sleeper segment. More on this later.</li><li><strong>Automotive:</strong> <strong>$604M</strong> (↑6% YoY). Below the $654.8M consensus.</li></ul><h3>Reading Between the Numbers</h3><p>Despite what everyone tries to tell you, NVIDIA’s growth story is far from over. But don’t just take my word for it. Look at the numbers for yourself:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FOrYfAhck5iI7xstDEXmGA.png" /><figcaption>Created by author based on NVIDIA’s reported numbers.</figcaption></figure><p>Most analysts point to the <strong>$4.7 trillion</strong> market cap and <strong>$215.9 billion</strong> revenue base as a sign that growth has to slow. At this scale, they’re typically right. But somehow the company keeps finding yet another gear and resisting the pull of large numbers.</p><p><strong>So how does a $200+ billion revenue company keep accelerating?</strong></p><h4>Data Center</h4><p>I <a href="https://medium.datadriveninvestor.com/nvda-take-the-moat-over-the-memes-ff7897a686e0">wrote back in October</a> about NVIDIA’s pace of innovation as a compounding moat, and Q4 showcases exactly that. The product roadmap embodies the “next man up” mentality we see in sports, and it reminds me a lot of the amazing rosters that went on to build dynasties.</p><p>If the Dodgers have the best pitching rotation in baseball, NVIDIA has the AI hardware equivalent. Opponents get the same scouting report, but they still can’t do much against the talent already on the mound in addition to what’s warming up in the bullpen.</p><p><strong>Blackwell</strong> is your <strong>Ohtani</strong>. The two-way star handles both training and inference, and is the main driver behind the $3 billion Q4 beat. The kind of franchise player you build a dynasty around.</p><p><strong>Blackwell Ultra</strong> is your <strong>Snell</strong>. The mid-season pickup that immediately prepares you for another playoff run. Blackwell shows no sign of slowing down, but the improvements promised by Blackwell Ultra already have hyperscalers scouting whether it’s time for a mid-cycle upgrade.</p><p><strong>Vera Rubin</strong> is your <strong>Sasaki</strong>. An up-and-coming talent that hasn’t taken the mound yet, but is already making the rest of the league extremely nervous. Samples have shipped, production is on track for H2 2026, and everyone knows about the <strong>10x lower inference costs</strong> that could reshape the landscape.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uj_YzGcFM4aIW51XymnrlA.png" /><figcaption><em>Three generations of NVIDIA platforms shipping simultaneously across every layer of the stack.</em></figcaption></figure><p>CFO <strong>Colette Kress</strong> confirmed that the combined <strong>Blackwell</strong> and <strong>Vera Rubin</strong> pipeline now exceeds <strong>$500 billion</strong>, supported by token production and the hyperscalers’ shift in focus from building to running and monetizing their models. <strong>ChatGPT</strong> is at roughly <strong>700 million weekly active users, and Alphabet</strong> pushed <strong>980 trillion tokens</strong> in the month of June alone.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*idfXGVkdoMSdu_tRdN1K8g.png" /><figcaption>Token generation is doubling every two months.</figcaption></figure><p>The demand is there, the innovation is keeping pace, and naturally the growth follows. NVIDIA’s margin recovery is indicative of that, as they are now growing and also <strong>holding gross margins at 75% </strong>at the same time. While most analysts just stop at Data Centers, I believe there are a few more segments worth highlighting this quarter.</p><h4>The Sleeper: Professional Visualization (ProViz)</h4><p>Remember that sleeper segment I mentioned earlier? The one that grew by <strong>159% YoY </strong>and crossed the <strong>$1 billion mark</strong>? It gets lost in the Data Center numbers, but I think it has the potential to go viral. Hear me out.</p><p>Over the last few months, the internet has been losing its mind over AI agents that can run entirely on local hardware like a Mac Mini. <a href="https://openclaw.ai/"><strong>OpenClaw</strong></a> can manage your email, make dinner reservations, and even become the co-founder for the side business that you’ve always wanted to start. It even spawned <a href="https://www.moltbook.com/"><strong>Moltbook</strong></a>, an entire social network where AI agents interact with each other and humans can only follow along.</p><p>The tech is amazing, but that’s not why people are obsessed with it. All of it runs locally, which means <strong>your data never leaves your machine.</strong> That business plan you brainstormed with your AI co-founder never touches a third-party server. Your GTM strategy and competitive edge never leaves your device through API calls or the cloud. <strong>Privacy and security</strong> is the the actual value prop.</p><p>Now scale that concept to the enterprise level. CEOs can’t wait to inject AI into their database and tech stack, but what’s the biggest hurdle keeping them from pulling the trigger? <strong>It’s IT’s biggest nightmare, data security, and </strong>it’s essentially the Achilles heel of the entire enterprise AI thesis.</p><p>That’s where <a href="https://www.nvidia.com/en-us/data-center/rtx-pro-6000-blackwell-server-edition/">NVIDIA’s <strong>RTX PRO Server</strong></a> comes in. To keep company secrets from leaking over the internet the moment agents join the workplace, companies are essentially reversing course on cloud and prioritizing on-prem solutions to keep their data in-house and secure.</p><p>Today, it’s only 2% of total revenue, but I’m choosing to view the 159% growth as a preview for what’s to come. There was a time when accessibility and ease of use were the primary concerns of tech companies, but I think we’re starting to see a shift from the cloud back into those musty IT closets down on the basement floor.</p><h4>Networking</h4><p>I wrote about Jensen’s relentless pursuit of vertical integration last year, and it’s showing up in the Q4 numbers for Networking. Why go with any other vendor when NVIDIA can offer a fully optimized solution which includes networking out of the box?</p><p>We all know about the <a href="https://medium.com/datadriveninvestor/nvda-groqin-around-the-christmas-tree-ce0b19b6d6a4">“de-NVIDIAcation” trend</a> and the push for ASICs, but when push comes to shove, customers are still sticking with NVIDIA as a one-stop shop. After all, their decision immediately impacts performance per dollar today, and the gains only compound further as each generation of networking silicon gets co-designed with the rest of the stack.</p><p>Networking is a testament to the effectiveness of Jensen’s vertical integration playbook which we’ll cover in more detail in the next section.</p><h3>The Safest Seat at the Top</h3><p>The higher they fly, the harder they fall. That’s the mentality behind every bearish take on NVIDIA right now, and I get it. Staying #1 while carrying a massive target on your back is difficult enough. Doing it without stumbling a single time is even harder.</p><p>Just look at <strong>OpenAI</strong>. Untouchable a year ago, and now <strong>Google</strong>, <strong>Anthropic</strong>, and a giant wave of open-source models have crashed against the early mover’s moat, making <strong>ChatGPT</strong> look more like a commodity with each passing month.</p><p><strong>So why do I think NVDA is both the strongest AND the safest AI play in my portfolio today?</strong></p><h4>1. Low Asset &amp; CAPEX Risk</h4><p>This earnings season told two very different stories. Investors in Big Tech only heard the sound of cash being vacuumed out of the balance sheets, while NVIDIA investors couldn’t hear over the sound of the cash register.</p><p>As we covered a few weeks ago, <strong>Microsoft</strong>, <strong>Amazon</strong>, and <strong>Alphabet</strong> were all <a href="https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-9bea22b9cfbc">raked over the coals</a> for their out of control spending, and it doesn’t look like they’re about to get any breaks any time soon. As of this month, the Trump administration announced a <a href="https://www.reuters.com/business/energy/trump-says-he-has-told-big-tech-companies-build-their-own-power-plants-2026-02-25/">“protection pledge”</a> that essentially orders Big Tech to “BYOP”. Bring your own power.</p><p>All the big players and AI labs are expected to sign on, meaning that they’re now footing their own power bill on top of all the delayed data center projects, all while burning cash just to keep the LLMs running.</p><p>The opposite is happening over at NVIDIA, as they don’t have data centers to build, no power plants to fund, not even a single autonomous vehicle to manufacture. The company stands to benefit from this enormous AI tailwind while somehow remaining asset light and flushed with cash.</p><h4>2. Cash for Co-Design</h4><p>So where does <strong>$120 billion</strong> in annual net income go when you’re not saddled with power plants and data centers? <strong>Talent and acquisitions</strong>. And it’s the hidden cost of maintaining the yearly release cadence and racing against yourself to outdo last quarter.</p><p>We’ve already covered how the <strong>Mellanox</strong> acquisition turned into a <strong>$31 billion</strong> Networking segment. And when the market flipped from training to inference, Jensen dropped <strong>$20 billion</strong> on Christmas Eve to bring in <strong>Jonathan Ross</strong>, the brains behind <strong>Google’s TPU</strong>, along with 80% of <strong>Groq’s</strong> engineering team to Santa Clara, CA. This is extreme co-design and vertical integration in motion, and at this point, it’s rinse and repeat.</p><p>Every acquisition unlocks yet another layer of the stack for NVIDIA to own. From GPUs, CPUs, networking, liquid cooling, to CUDA, Jensen trades part of the balance sheet for strategic partnerships and the freedom to eventually own every component within the data center. Every dollar spent on co-design and full-stack optimization will make NVIDIA’s stack that much stickier, and ultimately pave the way to the Q1 guidance number.</p><h4>3. Physical AI</h4><p>I know, I know. Tech bros have been saying autonomous vehicles and robots are right around the corner for over a decade. We aren’t exactly there yet, but look at how much progress we’ve made within this last year.</p><p><strong>Waymo</strong> is completing over <strong>400,000 rides per week</strong> and targeting 1<strong> million by the end of 2026</strong>. Uber just launched an entire autonomous division and is deploying <strong>thousands of NVIDIA DRIVE-powered L4 vehicles</strong>. Even Tesla has had some wins, launching its robotaxi service in Austin and converting its flagship EV factory in Fremont, CA to ramp up <strong>Optimus</strong> production.</p><p>I credit my success in 2025 largely on <strong>betting on the racetrack</strong> instead of trying to pick the winning horses, and I believe the same applies here for Physical AI. Yes, Uber has the data moat and the distribution network. Alphabet has Waymo, and Amazon is probably employing more robots than humans in the warehouses by now. I’m invested in all of those names, but NVIDIA is clearly poised to benefit the most from the physical AI era.</p><p>Just like training and inference, every single one of these companies are building on NVIDIA’s platform. NVIDIA’s <strong>DRIVE</strong> powers the autonomous vehicles and <strong>Jetson</strong> powers edge robotics. No matter who wins, NVIDIA benefits from all of it.</p><h3>Closing Thoughts</h3><p><strong>$78 billion</strong> in Q1 guidance sounds crazy until you peel back all the layers. The Data Center segment is firing on all cylinders, with generations of products already waiting to get called into the game. <strong>ProViz</strong> has the potential to become the on-prem game-changer that could actually move the needle on enterprise AI adoption.</p><p><strong>Networking</strong> is living proof that vertical integration isn’t just a buzzword and can continue to grow the core business. And <strong>Physical AI </strong>is lining up as the next massive TAM expansion with NVIDIA already cornering the market. All that’s left is to execute, and I would never bet against Jensen and his team when it comes to that.</p><p>I know how paradoxical it sounds to call the world’s most valuable company a “safe” investment, especially when we’re dealing with something as nebulous as AI. However, your rotation destinations are all completely overbought and Big Tech’s free cash flow is in the gutter for the next few quarters. For my growth investors out there, where else can we park our money where the business is leading the sector, collecting 60% operating margins, offloading the CAPEX and overhead risk to everyone else, and just comfortably reinvesting $120 billion back into their own tech?</p><p>Lastly, I know there’s been a lot of doom and gloom circulating around AI lately. I even felt compelled to <a href="https://medium.com/activated-thinker/anthropics-roadmap-is-now-the-market-signal-eeb582b26d1a">respond to the viral doomsday scenario</a> put out by <strong>Cintrini</strong> over on Substack. But for investors truly interested in investing in AI and tech, I want to remind you that a lot of these folks are writing from a position where they literally can’t lose.</p><p>If NVIDIA succeeds, they’ll claim that it’s one step closer to the AI bubble pop. If NVIDIA fails, they’ll get up on their soapbox and talk about how they predicted the outcome.</p><p><strong>I simply refuse to believe that we’ve come all this way just to innovate ourselves into a grave.</strong></p><p>That’s why I remain bullish on NVIDIA and continue to pick up shares when it dips into my buy zone ~<strong>$170.</strong> I’ve built up most of my position thanks to post-earnings drawdowns, and this quarter is no different.</p><p>NVIDIA is now sitting at a forward PE of roughly <strong>25x</strong>. For the record, that’s now cheaper than Walmart. 73% YoY growth with 75% gross margins, and the market is pricing it like day-old cupcakes at the grocery store bakery.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*86DDRFM4_L47W15besy43Q.png" /><figcaption>My NVDA position and buys in the last 6m.</figcaption></figure><p>I don’t know about you, but I’ll take every last one of them.</p><p>Thanks for reading!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=61610ceaec7e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/nvda-the-safest-seat-is-at-the-top-61610ceaec7e">NVDA: The Safest Seat is at the Top</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Anthropic’s Roadmap Is Now the Market Signal]]></title>
            <link>https://medium.com/activated-thinker/anthropics-roadmap-is-now-the-market-signal-eeb582b26d1a?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/eeb582b26d1a</guid>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[stocks]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Wed, 25 Feb 2026 19:24:03 GMT</pubDate>
            <atom:updated>2026-02-25T19:24:03.407Z</atom:updated>
            <content:encoded><![CDATA[<h4>An honest response to the viral AI bear case from an investor, GTM systems guy, &amp; Claude Coder.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*smEkV6FEGiFmIwdURzXY9A.png" /><figcaption>Image by author</figcaption></figure><p><strong>Citrini Research</strong> recently dropped their latest analysis on Substack called <a href="https://www.citriniresearch.com/p/2028gic"><strong>“Consequences of Abundant Intelligence.”</strong></a> It’s a fictional thought exercise from the future, looking back at how AI broke society. In their own words:</p><blockquote>“What follows is a scenario, not a prediction. This isn’t bear porn or AI doomer fan-fiction. The sole intent of this piece is modeling a scenario that’s been relatively underexplored.”</blockquote><p>Then they proceeded to unravel a dark dystopian timeline that reads like a Sweeney Todd adaptation of The Terminator. The dominoes fall like this: AI replaces SaaS, companies slash headcount, displaced workers consume less, credit and banks collapse, game over. Yikes.</p><p>As an investor that is optimistic on AI, I still see A<a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0">nthropic’s plugins as a collection of beautifully crafted markdown files</a> that showcased what a leading model can do in a neatly crafted sandbox. The scaffolding is there, but there’s still a massive schism between what is being demoed and what it actually takes to execute and drive deeper value in the real world.</p><p>As AI companies bump up against the old order of software and tech, I wanted to slow things down with my readers. AI moves fast but the whiplashing headlines and opinions have moved even faster. Let’s pick apart what I think is happening today versus what the market is pricing in.</p><p>To be clear, I’m not placing all the blame on Anthropic. For the most part, they’re letting the demos speak for themselves, and staying silent while the rest of us rabble about the implications. Turn on CNBC nowadays and it sounds a lot like a <a href="https://www.youtube.com/watch?v=wphbN1h4BZ4">South Park town-hall meeting</a>.</p><p>Because nobody fully understands where AI is going and every week brings a new groundbreaking development, both institutional and retail investors are being forced to shoot from the hip and ask questions later. Positions that took years to compound and build are now being liquidated within days, and <strong>Anthropic’s product roadmap has very much become the signal on where billions of dollars should be vacuumed out of next.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FOghsWQOLANwGEdmjG1kfg.png" /><figcaption>S&amp;P 500 sector performance YTD: the rotation out of growth is clear as day.</figcaption></figure><p>This is precisely how a markdown file turns into SaaSpocalypse. How a viral Substack post can rattle <strong>Visa and Mastercard</strong> even though they’ve <a href="https://medium.com/@MarketTherapy/stablecoins-destabilizing-a-duopoly-visa-and-mastercard-03c3e764cbb2">stayed ahead of the curve with stablecoins</a> and other technology. And how <strong>Microsoft and Amazon </strong>are being punished for CAPEX while bankrolling the very “startups” that are producing the technology that is displacing, well, technology.</p><h3>Separating the Foam from the Beer</h3><p>I recently heard<strong> </strong><a href="https://www.youtube.com/shorts/SKb5vWFtt2Y"><strong>JL Collins</strong> use an analogy of beer and foam</a> when analyzing stocks and it stuck with me. Instead of using it to separate the hype from the fundamentals of a stock, I want us to apply it to what the market thinks AI can do versus what the technology is actually doing.</p><p>Picture a Friday night at a packed bar. The music’s bumping, there’s no room at the counter, and the bartender is cranking out drinks like he’s going for a new high score. He grabs your glass, yanks the tap, and you end up with the world’s worst pour. A glass that’s all foam. Right now, the foam is everything we just talked about. The headlines. The product releases. All the viral tweets and Substack posts that continue to rise up and take up even more space in the glass.</p><p><strong>That’s all foam</strong>. And that’s the stuff moving the markets.</p><p>I’m not going to counter the foam by beer-tapping with my own arsenal of “what-ifs” and<a href="https://www.a16z.news/p/death-of-software-nah"> a list of technological shifts</a> that were perceived as the beginning of the end. Instead, I’m just going to fall back on lived experience.</p><p>Before I started writing about investing and AI, I managed GTM systems for tech startups. My days were spent figuring out what to buy, build, or rip out to better serve my coworkers and our customers, and whenever I’d talk to operations folks at other companies, the pain points were always similar.</p><p>A decade of “move fast and break things” has left behind doom boxes of technical debt and spaghetti code that even Claude wouldn’t be able to stomach. Disparate databases, bad data hygiene, and integrations duct-taped together by contractors or SMEs who left the company years ago. Even if AI was ready to integrate with years of institutional knowledge and unlock true ROI without leaking data to competitors, that’s the enterprise landscape AI is walking into, and that’s what’s standing between us and finally tasting the beer itself.</p><p>Now let’s talk about my experience with the actual tech. I’ve been a Claude Max user for about a year, and have been using Claude Code to create my own investing intelligence platform that bridges the gap between my portfolio tracker and my brokerages. Prototyping and seeing your own ideas come to life is a real rush, but I’ve also been on the losing end of endless debugging loops, folders packed with useless markdowns, and Opus going off the rails because I was vibing and compacted the conversation one too many times.</p><p><strong>That’s the beer.</strong> Tangible and getting better, but nowhere near ready to replace the mission-critical systems the market is already replacing.</p><p>It’s one thing to bootstrap a platform that I use myself, where the worst case scenario is wasted tokens and hitting usage limits. It’s another thing entirely to hand an AI agent CRUD credentials and the keys to your database, point it to years of proprietary institutional knowledge, and trust it to interact with your customers without the security, compliance, and guardrails that traditional SaaS providers have spent decades building.</p><p>The Citrini post was a fun read and well-constructed, but it’s crazy to see the markets so willingly price in the worst case scenario for SaaS and AI.</p><h3>When Fundamentals &amp; Sentiment Diverge</h3><p>So what does all this mean for investors?</p><p>Buying the dip never gets easier. Especially when the heatmap is showing 50 shades of red and every chart looks like it’s about to fall off a cliff. But I’ve seen this pattern before, and every time the fundamentals diverge from market sentiment, it’s been a pretty good time to dig in and look into whether or not I should <a href="https://medium.com/the-investors-handbook/when-buying-the-dip-takes-a-decade-7e091862b9ab">buy the dip</a>.</p><p><strong>Alphabet</strong> was an AI laggard and Search was dead back in June of last year. Then it became one of the best performers in my portfolio. <a href="https://medium.com/@MarketTherapy/part-of-my-pie-issue-1-1a3e8ff71b9e">My original thesis</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/976/1*for9KFaKzw12cd4Y5Hi8tQ.png" /><figcaption>My GOOGL position and buys within the last year.</figcaption></figure><p>Likewise, I bet that the demand for chips and AI would outlast tariff concerns and it played out perfectly. <a href="https://medium.com/datadriveninvestor/part-of-my-pie-issue-3-asml-1ff281d8bbb6">Thesis here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/984/1*BBrGfuEarafNFZ2IqlIsNA.png" /><figcaption>My ASML position and buys within the last year.</figcaption></figure><p>I’m not saying SaaS is the same turnaround waiting to happen. But I’m willing to bet that <strong>the gap between what the market thinks AI can do for enterprises today and the years of change management and hybrid work before it actually gets there is where the opportunity lives.</strong></p><p>Software multiples have compressed by over 33%. For a sector that traded at nosebleed valuations for most of the last decade, it’s worth a closer look. A cheaper price tag should never be the reason you pull the trigger, but I strongly believe that <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">SaaS will become the data layer</a> and the best distributor of AI. This is something that will become clearer once enterprises take the leap to terminate their contracts with their vendors.</p><p>Even if SaaS isn’t your cup of tea, consider where the money is rotating to. Consumer staples are trading at all-time high multiples, the highest the sector has been since the dot-com bubble. At these levels, I’d rather <a href="https://medium.com/the-investors-handbook/buy-tech-while-the-street-rents-safety-b08e8e0be0bc">keep my exposure to AI</a> while also benefiting from the diversified advertising, cloud, and e-commerce segments that are all growing much faster.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BEh-d3S4TasSUje_pxnNjQ.png" /><figcaption><strong>S&amp;P 500 Consumer Staples Sector: </strong>well above the 20-year average of 18.26x</figcaption></figure><p>With <strong>Salesforce and NVIDIA</strong> earnings dropping later today, I’ll be getting back into the rhythm of analyzing results and doing deep dives into companies that interest me. It’s been therapeutic to invest out loud, and I hope it has been the same for those reading along!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eeb582b26d1a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/activated-thinker/anthropics-roadmap-is-now-the-market-signal-eeb582b26d1a">Anthropic’s Roadmap Is Now the Market Signal</a> was originally published in <a href="https://medium.com/activated-thinker">Activated Thinker</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When Buying the Dip Takes a Decade]]></title>
            <link>https://medium.com/the-investors-handbook/when-buying-the-dip-takes-a-decade-7e091862b9ab?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/7e091862b9ab</guid>
            <category><![CDATA[personal-finance]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[trading]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[stocks]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Fri, 20 Feb 2026 20:26:10 GMT</pubDate>
            <atom:updated>2026-03-03T18:31:50.740Z</atom:updated>
            <content:encoded><![CDATA[<h4>Market crashes &amp; the simple math behind bouncing back quickly.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xQiBgtcw6LlUVaZa6sZ8xQ.png" /></figure><h3><strong>TLDR:</strong></h3><ul><li>The <strong>Shiller CAPE ratio</strong> is at <strong>39.8</strong>. The last time it was higher, the dot-com bubble burst. Nobel laureate Robert Shiller is projecting just <strong>1.5% annualized returns</strong> for the next decade.</li><li>I dug into three <strong>lost decades</strong> and the <strong>dollar-cost averaging</strong> data through each of them is surprisingly reassuring. The recovery math tells a completely different story than the crash headlines.</li><li>I share my February buys, portfolio performance, and break down how I’m applying the method to my own tech-concentrated holdings.</li></ul><p>It’s no secret that I’ve been bullish on the entire AI ecosystem and even the SaaS sector as it continues to get hammered.</p><p>The tech powering AI is real. So is the hybridized agent-human workforce.</p><p>But instead of locking myself in a complete echo chamber, I try to spend some time in the opposing camp. Given the panic that’s seized the market lately, wading through lazy bear takes is unavoidable. However, every once in a while, you come across a few ideas that nag you like the itchy tags on the back of your shirt collar until you finally stop what you’re doing to scratch the itch.</p><p>My latest itch? <strong>The lost decade</strong>. We all know the crashes. We doom-scroll through analysis of the dot-com bust and the financial crisis regularly. But how often do we zoom out and look at the full picture? A lost decade is the longer timeframe crashes live inside. A stretch over a decade or more where crashes and rebounds ultimately cancel each other out and the market ends up going absolutely nowhere.</p><p>So I hit the books. What causes a lost decade? How often do they happen? Are we headed for one right now? Most importantly, <strong>what happens to investors who stay the course?</strong></p><p>Turns out, the path forward, however painful, is actually stupidly simple.</p><h3>CAPE Ratio &amp; The Three Lost Decades</h3><p>If the lost decades got my attention, it was the <strong>Shiller CAPE ratio</strong> that put everything into perspective. Named after Nobel Prize-winning economist <strong>Robert Shiller</strong>, it smooths out earnings over a decade to measure how expensive the market really is relative to what it earns. Historically, the average sits around <strong>17.3,</strong> and the last time the CAPE was higher than it is today was right before the dot-com bubble burst.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5AgtO60TwkbRyAM9I3P2YQ.png" /><figcaption><a href="https://ycharts.com/indicators/cyclically_adjusted_pe_ratio">https://ycharts.com/indicators/cyclically_adjusted_pe_ratio</a></figcaption></figure><p><strong>Today it’s at 39.8.</strong></p><p>At these levels, Shiller projects <a href="https://finance.yahoo.com/news/nobel-prize-winning-economist-robert-130000225.html">just 1.5% annualized returns</a> for the S&amp;P 500 over the next decade. Let’s take a quick look at where the ratio was each time things actually went sideways.</p><p>During the <strong>dot-com bubble</strong>, the CAPE came in at a scorching <strong>44.2</strong>. Companies with no revenue and rich valuations were going public with a domain and a dream. The bubble burst and tech stocks lost <strong>77%</strong> of their value. When the market spent seven years crawling back to its 2000 highs, the financial crisis curb-stomped it right back down. Two crashes in one decade.</p><p><strong>The Great Depression</strong> started with a CAPE of <strong>32.6</strong>. The Dow had gone up six-fold in eight years, and retail investors were draining their savings accounts, mortgaging houses, and trading on margin to get in on radio, automotive, and electricity stocks. When the bubble popped, the banking system went with it, and markets that peaked in 1929 didn’t recover until 1954. Researchers still call it a lost decade, but this one took twenty-five years. And the CAPE was lower than it is today.</p><p>Then there’s the 1970s, the lost decade that actually came after the pop. Following the Nifty Fifty bubble, with the CAPE back down to <strong>24.1</strong>, it was Vietnam, Watergate, and a Fed chair whose best options were to crush the economy or surrender to inflation that kept the S&amp;P flat for seventeen years. While returns were flat on paper, inflation tripled the price of goods, and every dollar lost two-thirds of its purchasing power.</p><p>It’s not hard to see why so many bears are sounding the alarm. Today’s CAPE of <strong>39.8</strong> is higher than the starting point of two out of three lost decades. We have the valuations and concentration risk that come with bubble territory. Deglobalization, trade tensions, and a newly nominated Fed chair whose rate policy is still a giant question mark echo the geopolitical backdrop that ended up plaguing the 1970s.</p><p>It’s one thing to have a bearish take on AI. It’s another to have 150 years of market data to back it up.</p><p>However, history cuts both ways. The same data that supports the bear case also tells a surprisingly bullish story about investors who stuck to the plan and remained in the market through it all.</p><h3>How DCA Defeats a Lost Decade</h3><p>When I dug into the data, I realized that every crash analysis used a fatal assumption. The textbook studies assumed that you put all of your money in at once, right at the peak, and rode the elevator down for years. Or worse, sold at maximum pain.</p><p>Don’t get me wrong. The Great Depression was truly devastating. The wipeout totaled around <strong>90%</strong> and took 25 years to recover. But the other two? When looking at the entire lost decade, not just the crash, the dot-com era ended down about <strong>9%</strong> and the stagflation era ended flat. The real killer wasn’t the wipeout. It was the slow bleed with no light at the end of the tunnel that convinced many investors to sell. It was the inflation that slowly ate away at every dollar.</p><p>Now there’s a real time and place for lump-sum investing. Especially for those sitting on a pile of cash. But that’s not most of us. Most investors are still building as we go. A contribution from every paycheck. A little more when that performance bonus hits, and a little less when you’re not feeling as brave. I’m talking about dollar-cost averaging, and the math behind how it performs during a lost decade blew my mind.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r8E4_1uKiQlAQoBW7fiBxA.png" /><figcaption>Lump-sum recovery: 25 years. DCA recovery: 7 years</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7WdJ9pZSPulzh9SChEh6JA.png" /><figcaption>Lump-sum recovery: 17 years. DCA recovery: 5 years.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xbMjijNiJ0_R1qicjIhBpw.png" /><figcaption>Lump-sum recovery: 15 years. DCA recovery: 4.5 years.</figcaption></figure><p>In terms of recovery timelines, the data speaks for itself. The lump-sum investor waits for a whole generation, while the DCA portfolio recovers faster than LeBron James in the cryo-chamber. With the DCA strategy, every share purchased at lower prices ends up fast-tracking the recovery once the market is on the mend.</p><p>And here’s the kicker. DCA investors didn’t just reach their break-even points light-years ahead of lump-sum investors. They brought in real returns during the worst decades where the texbooks and headlines really only mention the <strong>50% drawdowns</strong>.</p><p>A DCA investor with balls of steel who invested $100 in the S&amp;P back in 1929 through the Depression and World War II earned a <strong>7.19% annualized return</strong> by 1954. Now add in reinvested dividends and that number jumps to <strong>12.1% annualized</strong>. A positive return during one of the worst economies on top of a global conflict.</p><p>The 2000s tell a similar story. For someone contributing $100 a month, the total return came in at <strong>5.9%</strong> with dividends reinvested, as opposed to a lump-sum return of <strong>-9.1%</strong>. By 2012, that would have grown to <strong>32.5%</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*g3BF4HmbKkYJAMwItEqUAg.png" /><figcaption>Lump sum: -9.1%. DCA with dividends: 5.9% by 2009 &amp; 32.5% by 2012.</figcaption></figure><p>I completely get why the dot-com era has become the poster child for market risk. However, I also think it’s just as important to provide a path forward, especially when the recovery mechanics are stupidly clear. You might not like my takes on AI or my methodology behind picking stocks, but the math I’m highlighting in this article is hard to argue with. It’s reassuring to know that if we continue what most of us are already doing, even through the roughest stretches in US stock market history, we’ll most likely be fine.</p><h3>Where the Math Meets My Portfolio</h3><p>I’m not saying that buying into a bloodbath is easy, especially when there’s so much uncertainty on so many different fronts. To be honest, I’m not even sure I can confidently apply what we’ve learned here while staring down the barrel of a real lost decade. All I can say is, the start of 2026 has been rocky, and though it’s early, I’ve been sticking to the game plan so far.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J1roDeTU6YPDeyvfimuAWw.png" /><figcaption>My recent buys in February</figcaption></figure><p><strong>The DCA math works best when you’re contributing to a diversified index like the S&amp;P 500 that continually rebalances itself. </strong>When you apply it to a more concentrated index like the tech-heavy NASDAQ, the math still maths, but the recovery gets delayed. The NASDAQ took 15 years to recover from the dot-com bubble compared to the S&amp;P’s 7. For most, I’d recommend dollar-cost averaging and diversification.</p><p>But as an investor with a long horizon ahead who actually enjoys reading up on the market, digging into moats, and understanding the tech driving the most recent trends, <strong>I’m willing to bet on myself</strong> over the wealth managers at my local Chase Bank.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iuThHBCYc7raF8jOL3xCZw.png" /><figcaption>My performance from Feb 2023-Feb 2026. Roughly when I pivoted hard into AI.</figcaption></figure><p>Rotating away from a more diversified portfolio and <a href="https://medium.com/datadriveninvestor/silicon-valleys-sprint-for-gold-9c4b7edb2b06">concentrating my thesis</a> took me from matching the market to outperforming. But as you can see, I&#39;m approaching my first gut-punch moment where I need to apply what I&#39;ve been writing.</p><p>We’re all staring at the same charts and reading the same earnings reports, but I believe that picking true AI enablers and <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">investing in sticky applications over AI wrappers</a> will continue to lead to outperformance over a much longer stretch.</p><p>The <a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0">current CAPEX cycle and the overall sentiment toward AI</a> have created buying opportunities, and I plan to buy the dip. <strong>Even if it takes a decade.</strong></p><p>Whether you want to put our heads together and chase outperformance, or just want to bookmark this article to come back and laugh in my face like <a href="https://www.youtube.com/shorts/ArXx5lqCZS8">Marshawn Lynch laughed at Pete Carroll</a>, welcome. I’d love to have you along for the ride.</p><p>For now, keep doing what you’re doing. It’s not exciting. But it works. Thanks for reading!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7e091862b9ab" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/when-buying-the-dip-takes-a-decade-7e091862b9ab">When Buying the Dip Takes a Decade</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Buy Tech While The Street Rents Safety]]></title>
            <link>https://medium.com/the-investors-handbook/buy-tech-while-the-street-rents-safety-b08e8e0be0bc?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/b08e8e0be0bc</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[stocks]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[stock-market]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Mon, 16 Feb 2026 17:02:19 GMT</pubDate>
            <atom:updated>2026-02-17T14:29:32.869Z</atom:updated>
            <content:encoded><![CDATA[<h4>Rotating now means overpaying for safety &amp; underselling growth.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VWxXHDaOdBv0vDLSaIMqSw.png" /></figure><h3><strong>TLDR:</strong></h3><p>Tickers mentioned: GOOGL, META, AMZN, MSFT, XLP, PG, COST</p><ul><li>The “defensive” rotation into consumer staples is the expensive trade with less margin of safety. Staples are at the <strong>97th percentile</strong> of their 5-year valuation range while hyperscalers are at the <strong>39th</strong>. Same sticker price with glaringly different fundamentals.</li><li>CAPEX only tells half the story. Hyperscaler OCF grew <strong>127%</strong> over 5 years while staples grew <strong>4%</strong>. Free cash flow will take a hit, but OCF growth points to strong fundamentals and effective capital deployment.</li><li>Growth has outperformed value by <strong>4x</strong> over the last decade. Rotating now means buying high, selling low, and potentially getting left behind when market sentiment on tech and AI improves.</li><li>It comes down to personal preference and risk tolerance, but if the goal is to outperform, I’d rather dollar cost average into quality growth at a discount than chase overpriced staples to limit the downside.</li></ul><h3>Introduction</h3><p>All of last year, investors had front row seats to one of the greatest love stories to ever unfold on the market. Tech and AI, our very own Jack and Rose, standing at the bow of the ship as companies crushed estimates and revenues began to soar. The quartet was playing a banger, bubbly was flowing, and the vibes were strong as we optimistically sailed into 2026.</p><p>Then came the glaciers.</p><p>First it was the <strong>$750 billion</strong> in projected CAPEX that rocked the boat. Then we hit the <a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0"><strong>SaaSpocalypse</strong></a>, which filled the cold market air with the sound of shrieking alarms. Then fears from Anthropic’s legal plugin spread like a contagion, cratering the finance, real estate, and now logistics sectors.</p><p>Where are all the lifeboats? Apparently, any sector that has absolutely nothing to do with AI. Think utilities, tobacco, burgers, and Cheerios. Piper Sandler is calling this a “passing of the torch,” as investors frantically jump ship from growth to value. Leaving tech and going defensive.</p><p>Although I’m long tech, I completely get the concerns. When tech bros and CEOs have already blazed through <strong>$420 billion</strong> in 2025, anchoring your portfolio with soaps and toothpaste feels like the smart move.</p><p>Building positions in tech during large investment cycles has paid off for me so far. But when everyone’s streaming for the exits and the lifeboats are filling up quickly, I figured I’d at least take a deeper look.</p><p>After a long weekend of research, I decided I’m staying put. Here’s why.</p><h3>Defensive Stocks at Offensive Prices</h3><p>If you moved into staples early, before the rest of the herd arrived, I tip my hat to you. So far, the consumer staples ETF, <strong>XLP</strong>, is up <strong>13% year to date</strong> while tech continues to bleed out week after week. After a full year of net outflows, XLP recently pulled in over <strong>$750 million</strong> in a single month.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SOVZ-Pj0dhT0LxupjGmfzA.png" /><figcaption>XLP inflow shows the “Great Rotation” accelerating. Source: ETF Database</figcaption></figure><p>Zooming out, staples have for the most part softened the blow of pretty much every major selloff. During the Great Recession, staples fell by <strong>28%</strong> while the S&amp;P plummeted by <strong>56%</strong>. Even in shorter recent drawdowns, rotating to staples can effectively put your portfolio on ice while other sectors experience complete meltdowns. The track record is there. Those that execute deserve a spot in the bunker.</p><p>However, if you’re like me, exposed to tech and <em>still</em> thinking about rotating, think again. Big tech has already shed <a href="https://seekingalpha.com/news/4548077-magnificent-seven-erases-nearly-3t-in-market-cap-as-wall-street-slides">nearly <strong>$3 trillion</strong></a> in market cap since last year. And as we learned in the case of Jack’s passing, the life raft can only support so many people. At this point, I’d argue that you’re not getting a ton of safety and you’re really only padding the gains of others.</p><p>I don’t like to lead with tables. That’s sort of why I write on Medium as opposed to Seeking Alpha. However, when it comes to consumer staples vs tech stocks, this table tells a darn good story:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JFeIbcipUrYRj9oluKrC3w.png" /><figcaption>Similar PE. Very different businesses. Generated by Author.</figcaption></figure><p>After all the chilling headlines lately, hyperscalers like Google, Meta, Amazon, and Microsoft are actually trading at an average forward PE that is identical to consumer staples. Exact same sticker prices for two very different products.</p><p>Now look at where price for both groups sits relative to history. Staples are now trading at the <strong>97th percentile</strong>, meaning they’ve only been more expensive 3% of the time within the last 5 years. The mass exodus away from AI has ballooned the sector to near all-time high valuations.</p><p>What about hyperscalers? Microsoft is now trading well below its own 5-year average. Amazon’s in the bottom third, and even the recent winners Google and Meta are hovering just above their midpoints. Combined, they’re averaging around the <strong>39th percentile</strong>, signaling value relative to their own historical valuations.</p><p>Staples are priced like the Cheerios are made of 24 karat gold, and the best of tech is on sale like it’s a ratchety Xbox 360 found at the local thrift store.</p><p>Apart from dividend yield, hyperscalers are running <strong>2x the operating margins</strong>, <strong>3x the net margins</strong>, and almost <strong>4x the revenue growth</strong>. On a $100K position of XLP you’d be making around $3K a year in dividends. But are you really banking on a $3K bandaid to stop the bleeding when you’re hemorrhaging $30K? Unless you’re Buffett sitting on $25B in Coca-Cola stock and had access to Robinhood in the womb, it just doesn’t make sense.</p><p>In past articles I’ve said that <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">not all SaaS is created equal</a>, so let’s give staples the same respect and look at some specific companies that are synonymous with being black swan event insurance.</p><p><strong>Procter &amp; Gamble</strong> is a unanimous blue chip, but it’s trading at a forward PE of ~25x with revenue growth less than 3%. In comparison, <strong>Meta</strong> with a similar PE is growing revenue over <strong>20%</strong>. <strong>Costco</strong>, which I frequent weekly to power my toddler, is a phenomenally run business but sits at <strong>50x+ PE</strong>. The company keeps the dad jeans and rotisserie chicken cheap, but that’s about it. There’s an undeniable premium when it comes to the stock itself.</p><p>And before you call me a tech fanboy or a staple hater, I’ve owned most of these names over the years. However, at some point the opportunity cost became too real to ignore, especially once the AI trend emerged.</p><p>In my eyes, the data is telling you the same thing. If you’re still rotating into staples at these prices, you’re unequivocally committing the cardinal sin of investing. Buying high and selling low. Overpaying for safety while underselling growth.</p><p>Receipts of my PG and COST trades:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PWue0zPnrQN0hzTaO3QTxg.png" /><figcaption>Bought PG at $144. Sold at $160.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pnEp1m59JGHJhABWCa22aA.png" /><figcaption>Bought COST at $566. Sold at $873.</figcaption></figure><h3>CAPEX is Only Half the Story</h3><p>So if the staples are expensive for what they are, why is everyone still running? It’s the <strong>$750 billion </strong>CAPEX figure, and it only tells half the story.</p><p>For whatever reason, investors are recently infatuated with only looking at how much cash is outflowing to fund data centers. As I covered in <a href="https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-9bea22b9cfbc">Part 1</a> and <a href="https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-part-2-3a155711e46b">Part 2</a> of my recent Mag 7 Megaposts, patience is wearing thin, and the last thing Wall Street wanted to hear was the sound of cash being incinerated. Yes, CAPEX will decimate free cash flow during the largest infrastructure buildout in tech history, but I feel like that’s a given unless you still believe that AI is all smoke and mirrors.</p><p>Let’s take a more balanced lens and look at how much cash these businesses are generating before they reinvest it. Also known as<strong> operating cash flow (OCF)</strong>, it’s a metric that’s being entirely omitted from these doom and gloom headlines. To me, if OCF manages to grow even while the company reinvests heavily, it’s capital well spent, and the underlying business is actually growing rather than in danger.</p><p>To put this into perspective, I pulled OCF data from <a href="https://www.macrotrends.net/stocks/research">MacroTrends</a>, and applied the same base-year indexing method used to track the S&amp;P 500 and CPI. The question is simple: for every dollar generated starting in 2021, how much more are they generating now?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*twJn_RfI0c51X9Z6SshO1w.png" /><figcaption>Hyperscaler OCF grew 127% vs staples’ 4%. Generated by Author using Gemini.</figcaption></figure><p>For hyperscalers, that single dollar turned into <strong>$2.27</strong>. Collectively, OCF went from <strong>$256 billion to $581 billion</strong>, a <strong>127%</strong> increase. Staples grew that same dollar by <strong>4%</strong>, from $89 billion to $93 billion. Considering how much of the defensives’ OCF will be stripped out to pay dividends to shareholders, I’d argue that the full picture and outcome gets even uglier.</p><p><strong>127% vs 4%</strong> within the same timeframe. Even with all the perceived turbulence, one group’s fundamentals are flying high while the other group is glaringly grounded on the tarmac.</p><p>I’m also not oblivious to the dot-com comparisons that investors are being bombarded with. Once again, OCF cuts through the noise. It simply didn’t exist for many of the pre-revenue internet startups of the late 90s.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JIwdGQQ81QY0Np1gCQlXkg.png" /><figcaption>Prices in 2026 more attached to OCF. Generated by Author in Gemini.</figcaption></figure><p>In 2026, stock prices and cash generation are moving in lockstep. Microsoft’s OCF grew <strong>282%</strong> over the last six years while its stock grew by <strong>293%</strong>. Amazon’s OCF actually outpaced its stock. When cash flow and price move this closely together, you’re looking at valuations validated by performance, not remotely close to the dot-com era.</p><p>Alphabet and Meta are the two where stock prices have outpaced OCF, but that’s largely a result of their recent successes as the market has finally caught up to their margin expansion narratives. Alphabet went from AI laggard to AI leader, and Meta nearly doubled their operating margins. Again, hardly a fair comparison to Pets.com from the 90s.</p><p>All this is to say that investors need to look at both metrics to get a more complete picture. OCF illustrates strong underlying fundamentals and the optionality to reinvest back into the business, while CAPEX telegraphs where these companies are placing their bets.</p><p>The results speak for themselves, as Amazon has more than doubled its operating margin from <strong>5.3% to 11%</strong>, and Meta went from a <strong>25% margin</strong> in 2022 to over <strong>42%</strong> today. When margins are expanding like this, the CAPEX cycle is clearly working. The recent pullback only creates cheaper entries to greater upside and future growth.</p><h3>Why I’m Not Renting Safety</h3><p>Here’s the thing about this rotation that most analysts are leaving out. It works until it doesn’t, and even when it does, it requires next-level luck and timing that I definitely do not possess.</p><p>Just as it can protect you from a crash, it can just as easily cap the recovery on the ride back up. The market always claws back staples’ gains once the storm clouds pass. The haven is temporary and the window for execution is smaller than you think.</p><p>When the Great Recession bottomed out in 2009, the Nasdaq surged 78% from its March low while defensives completely flatlined. The S&amp;P climbed 30% within just two months of the bottom. If you were still nestled in the XLP bunker, you would have missed it.</p><p>More recently in 2023, after the markets digested the rate hike news, XLP returned -1% while the S&amp;P surged by 24% that same year. I get the need for short-term shelter, but based on historical data, the market makes for a shitty landlord, and always eventually repossesses the gains.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CXBdKMNPd5kAkJFDfH8W9Q.png" /><figcaption>Growth returned ~393% vs Value’s ~102%. Source: Bloomberg. Generated by author.</figcaption></figure><p>Growth has outperformed value by nearly <strong>4x</strong> over the last decade. Instead of getting in and out during every drawdown, my goal is to let the trend run its course, and let my picks’ fundamentals translate into gains over time.</p><p>Again, congrats to those who rode the AI wave and are now locking in their gains. You’ve completely nailed the entry and made it onto the lifeboats. However, I wrote this post for the folks that, like myself, have either missed the memo or believe in long-term fundamentals as the true life preserver that will eventually stabilize our growth-focused portfolios.</p><p>From here, I’m keeping my playbook simple. I’ve been honest about <a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0">whiffing my timing into SaaS</a>, and I’m not rolling the dice again on sectors that are more foreign to me. I’m opting to buy low on companies that are well within my wheelhouse and have far superior fundamentals, instead of chasing safety in names sitting near all-time highs.</p><p>There’s absolutely nothing wrong with seeking haven from market volatility, but for those of us trying to build portfolios that beat the market’s average return, these are the decisions that make all the difference.</p><p>For me, I’d rather stomach the red days and continue dollar cost averaging into quality growth. Watching from the sidelines while tech takes off like it has time and time again would be far more soul crushing than a few bad stretches on the heat maps.</p><p>That said, I’d love to hear from seasoned investors who have been in the market much longer than I have to stress test this thesis. I’ve made my position clear. Curious to hear about yours.</p><p>Thanks for reading!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b08e8e0be0bc" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/buy-tech-while-the-street-rents-safety-b08e8e0be0bc">Buy Tech While The Street Rents Safety</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SaaSpocalypse: Prompts, Panic, & The Path Forward]]></title>
            <link>https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad80f4cb73e0</guid>
            <category><![CDATA[stocks]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Wed, 11 Feb 2026 23:37:56 GMT</pubDate>
            <atom:updated>2026-02-12T17:35:52.986Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why AI needs SaaS more than the brutal selloff suggests.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*I8uP5CwYrzOtMgBbOSBxDA.png" /></figure><h3>TLDR:</h3><p><strong>Companies mentioned:</strong> OpenAI, Anthropic, NOW, CRM, UBER</p><ul><li>After expanding my thesis for 2026 into software, SaaSpocalypse wiped $300 billion from SaaS and AI in the same week. Wall Street is finally waking up to AI’s potential, but we’re arriving at very different conclusions about what comes next.</li><li>OpenAI’s schema for Frontier shows investors exactly where enterprise AI is headed: platforms built entirely on top of your systems of record and existing software stack. The data layer that AI desperately needs to improve is the very sector the market keeps selling off.</li><li>ServiceNow locked in a multi-year partnership with Anthropic. Salesforce built proprietary models while integrating Agentforce with ChatGPT, Claude, and Gemini. Uber’s AV partners are all building on top of its platform rather than replacing it. Wall Street is screaming that AI kills SaaS, but the ongoing strategies speak otherwise.</li><li>The market hates uncertainty, but that’s exactly where we’re headed. A long stretch of hybrid human-AI work that doesn’t perfectly fit the narrative. NOW, CRM, and UBER are sitting at 52-week lows with decent fundamentals, but I’d suggest waiting for the dust to settle.</li></ul><h3>Introduction</h3><p>Two weeks ago, as I took a sip of celebratory LaCroix to commemorate how the AI infrastructure trade has continued to pay off, I was eager to expand my thesis to cover more emerging plays within AI. This led me to dip my toes into the treacherous waters of SaaS as I <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">identified three out-of-fashion software players</a> I felt were poised to capture more margin as token economics shifted in their favor. Boring apps sitting on decades of data, ready to become the high-speed rails for hybridized agentic workflows.</p><p>Then CEO Dario Amodei of Anthropic grabbed the mic like he was 2009 Kanye West, <a href="https://claude.com/blog/cowork-plugins">dropped a new Legal AI plugin</a>, and watched as the SaaS sector completely went up in flames.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bk_l9tapI0otTQlN6Rk_vw.png" /><figcaption><a href="https://x.com/Speculator_io">https://x.com/Speculator_io</a></figcaption></figure><p>I’d be lying if I said it didn’t sting. Watching double-digit drawdowns on positions you had just confidently put your name next to. CRM down another 8%, adding to a <strong>40%</strong> decline over the past twelve months. NOW down <strong>11% after-hours</strong> despite beating every metric on their Q4 earnings with a 98% renewal rate. Even Uber, my wildcard experience layer play is <strong>down by 16%</strong> just this last month. The pain is real!</p><p>However, after I calmed the nerves by crushing another case of bubbly water and racked my brain on why the market would erase <strong>$300 billion</strong> from both SaaS and AI in the same week, I realized that this widespread panic was actually affirming. With the help of Anthropic, Wall Street was finally waking up to what we’ve been writing about for months. We were digesting the same headlines but arriving at very different conclusions.</p><p>The market wants black and white. <strong>“AI kills SaaS</strong>” offers a clean narrative that lets you sell with conviction and sleep better at night. But if you’re like me and looking for opportunities while the jury’s still out on whether AI is a frivolous intern or a cold-blooded assassin ready to take down an entire sector, I want to share the thought process that kept me grounded amidst the SaaSpocalypse, and why I still think of SaaS as an undervalued AI play.</p><h3>The Plugin That Prompted Panic</h3><p>I don’t want this to turn into the same article that you’ve already been bombarded with over the weekend from CNBC, WSJ, or even here on Medium, but I think it’s important to separate the panic from the tech.</p><p>Anthropic dropped <a href="https://github.com/anthropics/knowledge-work-plugins">a legal plugin amongst others</a> that can triage NDAs, flag clauses, and redline contracts. Tasks that once ran up the billable hours or forced law firms to pay software subscriptions could now be done for twenty bucks a month. Anthropic’s release, paired with the market’s appetite for one-sided narratives, led many to believe that this was the end for all of SaaS.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LaMQkdNonmvwEZO2MEEK8g.png" /></figure><p>Here’s where I think the panic got way ahead of the tech. For those of us who’ve been tinkering in ChatGPT and Claude Code, this is actually pretty baseline stuff. What Anthropic released was a beautifully crafted markdown file or “skill”. A series of detailed instructions and guardrails on how Claude should act when performing these legal workflows. The biggest difference between this plugin and what any vibe-coder has been tossing onto GitHub is that instead of relying on run-on sentences telling it to “act as a lawyer with 20 years of experience,” Claude could lean on a set of handcrafted instructions optimized by an Anthropic engineer.</p><p>The plugin connects the dots between what a majority of folks still see as nebulous technology and a tangible use case. The market reaction might even go down as one of the best PR campaigns for AI ever. But to call this the MVP of the SaaS killer is beyond a stretch. If anything, this plugin just proved that AI still has a long way to go in order to break out of the handcrafted sandbox without the help of knowledge workers and the various systems that the market just took behind the barn.</p><h3>A Look At Frontier’s Foundation</h3><p>We’ve spent a lot of time looking under the hood of Anthropic’s plugin. As a frame of reference, let’s also check out what their greatest rival, OpenAI, is doing to bring AI at scale into the workplace.</p><p>Last week, Sam Altman and <a href="https://openai.com/index/introducing-openai-frontier/">the team announced Frontier</a>, an agentic platform that empowers AI with “shared context, onboarding, hands-on learning with feedback, and clear permissions and boundaries.” To me, this is the release that should’ve actually sent shivers down Wall Street’s spine, but ironically it just flew under the radar. Major corporations like HP, Intuit, and Uber are already integrating it into their tech stack and allowing Frontier to manage multi-step workflows and run autonomously long after the last employee heads home.</p><p>So why am I still convinced that SaaS is an opportunistic AI play? Check out OpenAI’s schema and architecture for this flagship product:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KXK_HT_QYZ1JtueD5wHGdQ.png" /></figure><p>Below all the AI-infused interfaces, shiny new agentic tools, and OpenAI’s proprietary logic architecture lies a boring, almost unassuming layer that will now make up the sole foundation propping everything else up.</p><p>This is a critical reveal in system design. It proves that leading AI labs aren’t looking to replace your CRM, data warehouse, ticketing software, or custom apps. The model makers burning tens of billions in this neck-and-neck race have their eyes set on something far more ambitious than a plugin or a chatbot. Neither OpenAI nor Anthropic has the resources, time, or desire to go sector by sector and recreate the messy, entrenched systems of record that key SaaS players have already laid brick by brick.</p><p>That said, not every SaaS company has what AI labs are looking for, and that’s the major risk behind this thesis. However, like I laid out in my other article, I’m willing to bet that the ones with the data, distribution, security, and malleable architecture will be the software companies that are no longer being offered on the lunch menu.</p><h3>From Projection to Production</h3><p>Enough of my projections. What I find so exciting about this space is how fast entire landscapes can shift. Eight days after I published my SaaS piece, <a href="https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-and-Anthropic-partner-to-help-customers-build-AI-powered-applications-accelerate-time-to-value-and-apply-trusted-AI-to-critical-industries/default.aspx">ServiceNow and Anthropic announced a multi-year partnership</a> to help AI go from just Claude to coworker. ServiceNow gets to deliver the joys of Opus to CEOs that have been pounding their fists for AI, and Anthropic gets deeper hooks into highly industry-specific data that can help Claude become a subject matter expert instead of a generalized assistant.</p><p>Salesforce isn’t waiting around either. Instead of going deep with just one partner, the company built their own proprietary models: xGen-Sales for pipeline management, xLAM for task execution, and xGen-Code for developers. Since all the models were trained specifically on Salesforce data and architecture, the home-court advantage naturally leads to performance boosts when it comes to Salesforce-specific tasks. Although they’ve been bootstrapping their own AI platform, they’ve rapidly integrated with OpenAI, Gemini, <a href="https://www.salesforce.com/news/stories/salesforce-anthropic-trusted-context-ai-actions-on-claude/">and even Anthropic</a> as of earlier this month. With over 90% of the Fortune 500 already on Salesforce, Benioff is betting that allowing customers to “<a href="https://help.salesforce.com/s/articleView?id=data.c360_a_bring_your_own_model.htm&amp;type=5">BYOM</a>” (bring your own model) will compound the company’s dominant network effects even further. These effects have done wonders for Google and Microsoft so I expect it to do the same for Salesforce.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DfFFlVAd2N51E0KCAkI-dQ.png" /><figcaption>Screenshot from 360 Blog</figcaption></figure><p>These dynamics aren’t just happening within enterprise SaaS either. Uber completely backed out of building autonomous vehicles, but Waymo, Waabi, and NVIDIA are all heavily partnering with the company for data and distribution. In fact, on the same day that ServiceNow announced their partnership with Anthropic, Canadian AI lab Waabi announced that they <a href="https://waabi.ai/insights/waabi-secures-1-billion-in-new-funding-to-lead-physical-ai-revolution">planned on committing 25,000 robotaxis exclusively to Uber’s platform</a>. Considering that autonomous vehicles on Uber’s network completed 30% more rides per day compared to other platforms in Q4, the partnership makes complete sense. Waabi gets access to a goldmine of data, and Uber’s utilization rate shows just how valuable they are without ever needing to manufacture another vehicle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vXFz1UftxDsrKbEpB6KQxw.png" /><figcaption>Uber Q4 Earnings</figcaption></figure><p>Different sectors, different strategies, and yet all the same theme. To successfully take that leap of faith out of their carefully constructed sandbox, AI undoubtedly needs and will rely on these systems of record and strong data foundations. This is why I remain bullish.</p><h3><strong>Closing Thoughts</strong></h3><p>I’ve thrown a lot at you as your friendly neighborhood investor. Let me turn on the white noise machine and put on my therapist coat for a moment.</p><p>Market sentiment, whether you agree with it or not, is a powerful force that needs to be acknowledged. While the names I’ve mentioned just hit 52-week lows, I’m waiting for the dust to settle before formulating next steps. When looking at the underlying fundamentals for NOW, CRM, and UBER, none of the most recent metrics are raising any major red flags.</p><p>I’d also encourage you to look past the black and white headlines, however comforting they might be. Investors hate uncertainty and the gray area, but I think that’s exactly where we’re headed. AI is real, but it’s still too early for us to go on PTO forever. We’re going to see a long stretch of hybrid work and trust building before the next huge leap in compute and modeling, and the companies that can accommodate this have the advantage. Here are two studies I found extremely compelling even though they got drowned out by the recent doom and gloom: <a href="https://mitsloan.mit.edu/ideas-made-to-matter/when-humans-and-ai-work-best-together-and-when-each-better-alone">MIT Sloan on human-AI workflows</a> and <a href="https://news.mit.edu/2016/human-robot-teams-more-effective-0602">BMW/MIT on human-robot collaboration</a>.</p><p>All this to say, despite SaaSpocalypse and the ripples it created across the AI ecosystem, I still see opportunities for investors willing to look in the gray area. I’d love to provide a thesis that fits neatly in a box, but that’s just not how investing works. Or life on this planet, for that matter.</p><p>Thanks for reading.</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad80f4cb73e0" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/saaspocalypse-prompts-panic-the-path-forward-ad80f4cb73e0">SaaSpocalypse: Prompts, Panic, &amp; The Path Forward</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mag 7 MegaPost: Winners, Waiters, & Watch Outs (Part 2)]]></title>
            <link>https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-part-2-3a155711e46b?source=rss-f341c5cf49ec------2</link>
            <guid isPermaLink="false">https://medium.com/p/3a155711e46b</guid>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[stock-market]]></category>
            <category><![CDATA[big-tech]]></category>
            <category><![CDATA[stocks]]></category>
            <dc:creator><![CDATA[Market Therapy]]></dc:creator>
            <pubDate>Fri, 06 Feb 2026 21:55:35 GMT</pubDate>
            <atom:updated>2026-02-08T15:28:00.747Z</atom:updated>
            <content:encoded><![CDATA[<h4>How I’m playing GOOGL, AMZN, &amp; NVDA after Big Tech earnings.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-I970a1p84PCXcQv2vecWQ.png" /></figure><h3>TLDR:</h3><ul><li>Despite strong execution from Big Tech, investors are punishing every mention of CAPEX despite cheering for it all throughout 2024–25.</li><li><strong>Alphabet (GOOGL) — A+.</strong> Google Cloud grew 48%, Gemini hit 750 million users, and Pichai revealed a 78% reduction in serving costs. The stock still sold off due to 2026 CAPEX guidance. Alphabet’s full-stack advantage in TPUs and frontier models will create pricing power down the road, but remains underappreciated for now.<strong> Hold for now and will accumulate more on dips.</strong></li><li><strong>Amazon (AMZN) — A-.</strong> AWS captures 24% growth, but the $200 billion CAPEX commitment, an EPS miss, and soft Q1 guidance sent shares down 10%. The execution is strong, but the balance sheet risk and short- to medium-term margin compression warrant a downgrade from last quarter’s A. <strong>Strong Buy in my eyes for long-term investors who believe in Amazon’s aggressive reinvestment strategy and track record.</strong></li><li><strong>NVIDIA (NVDA) — Reporting Feb 25.</strong> The semiconductor canaries (ASML, TSMC) are screaming that demand is real and accelerating. This earnings season shows how infrastructure is still a bottleneck and top of mind for every CEO. With Rubin promising 5x inference performance and 10x lower cost per token, Jensen may have the answer to Big Tech’s CAPEX problem. <strong>Always a Hold ahead of earnings. Post-earnings has been a great time to accumulate shares.</strong></li></ul><h3>Introduction</h3><p>We are back! It’s only been a week since we covered Microsoft, Tesla, Meta, and Apple, and somehow it’s already looking like an entirely different market. <a href="https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-9bea22b9cfbc">Here’s Part 1 if you need a recap.</a></p><p>With Alphabet and Amazon now in the books, a clear pattern is emerging this earnings season. Investors that were once giddy to board the AI rollercoaster have gone through one too many corkscrews and are now demanding to get off the ride. Execution is secondary to CAPEX fatigue, and investors are experiencing intense buyer’s remorse despite many of the Mag 7 showing tangible results. Tech CEOs that were once punished for not doing enough in AI are now being punished for doing too much. The CAPEX numbers are sobering, but so is the irony. Investors spent the last two years demanding Big Tech invest in AI, and now that it’s gut check time, Wall Street is getting cold feet.</p><p>To make matters worse, hundreds of billions were just wiped off of the software sector in a matter of days. This time, there were no Truth Social bombs. No sudden land grab for America’s 51st state. Following my <a href="https://medium.com/the-investors-handbook/the-chip-on-softwares-shoulder-0504994e637f">2026 thesis around investing in software</a> and the experience layer of AI, I’m working on a piece around the<strong> “SaaSpocalypse”</strong> sell-off. However, that’s an article for another day.</p><p>Today, we’re here to wrap up Big Tech earnings. There’s also a ton to cover with the two cloud heavyweights, <strong>Alphabet</strong> and <strong>Amazon</strong>. Although we won’t hear from <strong>NVIDIA</strong> until the end of the month, I’ll share my thoughts on what still is my largest holding even as the markets expect perfection.</p><p>So let’s get into it!</p><h3>Alphabet (GOOGL) Grade: A+</h3><p>This update on Alphabet is annoyingly hard to write for a couple of reasons, despite being one of my largest positions and effortlessly holding onto their <strong>A+ grade</strong> from a quarter ago. I cautioned against expecting immediate revenue growth as a result of all the ramping CAPEX in Part 1 of the Mag 7 Megapost, but Alphabet breezed past all expectations and effectively dismantled my call for delayed gratification as they showed off the various ways that AI was transforming their business at this very moment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_RA0InyAjtwXsDiw-Xlsxw.png" /></figure><p>There’s so much momentum across Cloud, Search, YouTube, and Waymo that it’s hard to distill everything down. Much to Microsoft and Amazon’s disdain, Google Cloud accelerated to <strong>48% growth</strong> while facing the same supply constraints plaguing the entire sector. Google Search continues to dance on its own grave, logging <strong>17% growth</strong> to help fund the company’s AI endeavors. Similar to Microsoft’s adoption numbers around Copilot, Alphabet also reported massive adoption of AI in the workplace, selling <strong>8 million paid Gemini Enterprise seats</strong> in just four months. We won’t get into the accomplishments of Waymo or YouTube here, but having the extra insurance provided by all the other segments if Search or Cloud ever sneezes helps me sleep better at night.</p><p>Still, despite all that growth, Wall Street refused to give Alphabet a free pass on CAPEX. After management announced that the company was planning to nearly double their 2025 spend to <strong>$175–185 billion for 2026</strong>, investors quickly latched onto the CAPEX narrative, driving the stock down <strong>3%</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JQ2_8QFsw_pYJyrP_A5Iyg.png" /></figure><p>Personally, I think grouping Alphabet in with the rest of the hyperscalers and letting the CAPEX headline overshadow future potential is extremely shortsighted. Unlike the others that are dependent on NVIDIA or still figuring out how to optimize their own ASICs, Google has a significant lead both in terms of hardware and in frontier models. Google’s <strong>Ironwood TPUs</strong> are the only custom chips that have come close to competing against NVIDIA GPUs, and Gemini has closed the gap significantly against ChatGPT and Claude. This type of bootstrapped full-stack advantage has led to NVIDIA’s incredible lead in AI, and I think we’re about to see it do the same for Google. If you focused solely on the CAPEX headlines, you would have missed what CEO Sundar Pichai had to say on evolving unit economics:</p><blockquote>“As we scale, we’re getting dramatically more efficient: we were able to lower Gemini serving unit costs by <strong>78%</strong> over 2025 through model optimizations, efficiency and utilization improvements.”</blockquote><p>As Alphabet continues to optimize their stack, this advantage will undoubtedly contribute to the company’s pricing power as they continue to drive down costs and make it extremely difficult for competitors to operate within the same space. Since Wall Street is now allergic to the majority of Big Tech or any mention of CAPEX, I’m not rushing out to pick up shares of GOOGL just yet, as I’m confident that market sentiment will create better buying opportunities soon. As asinine as this might sound, I’m confident that some of Alphabet’s structural advantages have yet to be fully priced into the stock, despite starting off 2026 with a blowout quarter.</p><h4>Earnings Highlights</h4><ul><li><strong>Financials:</strong> Total revenue grew <strong>18% YoY to $113.8 billion</strong>, beating the $111.4 billion consensus. Net income surged <strong>30% to $34.5 billion</strong> with EPS of <strong>$2.82</strong> (↑31% YoY). Google Services revenue increased <strong>14% to $95.9 billion</strong>, led by Search and YouTube, and Alphabet crossed <strong>$400 billion in annual revenue</strong> for the first time in company history.</li><li><strong>CAPEX:</strong> Alphabet is nearly doubling their 2025 spend to <strong>$175–185 billion for 2026</strong>. The investment will go toward AI compute capacity for DeepMind and GCP as they work through the backlog that has grown <strong>55% QoQ to $240 billion</strong>.</li><li><strong>AI Monetization:</strong> Google Cloud grew <strong>48% YoY to $17.7 billion</strong>, crushing the $16.2 billion estimate and significantly outpacing Azure (39%) and AWS (24%). Gemini now has <strong>750 million monthly active users</strong> and Gemini Enterprise has sold <strong>8 million paid seats</strong> in four months. Apple is now reportedly paying Alphabet an estimated <strong>$1 billion annually</strong> to power the new Siri. Waymo is delivering over <strong>400,000 weekly rides</strong> and expanding internationally.</li></ul><h3>Amazon (AMZN) Grade: A-</h3><p>Andy Jassy clearly did not get the memo. His team heard the bloodcurdling screams from investors when Microsoft announced they’d be spending north of <strong>$100 billion</strong> on AI infrastructure in fiscal 2026. They were in the room when Alphabet smashed expectations across the board but got a lukewarm applause after raising CAPEX guidance to <strong>$175–185 billion</strong>. Heads were rolling, investors were heading for the exits, and Amazon decided to lean into the chaos. With <strong>$200 billion</strong> in planned CAPEX for 2026, Amazon is officially the biggest spender in Big Tech heading into NVIDIA earnings. If Jassy mic-dropped, nobody heard it over the sound of the AMZN stock plummeting 10% through the stage.</p><p>When you get past the CAPEX sticker shock and actually evaluate the quarter, you find a pretty solid performance underneath. Last quarter, AWS growth had stagnated while Azure and GCP posted huge numbers. Jassy blamed capacity constraints instead of a lack of demand, but Wall Street wasn’t buying it. Fast forward to Q4 and AWS is absolutely back on track. Growth has accelerated to <strong>24%</strong>, its fastest pace in 13 quarters, and the backlog has ballooned to <strong>$244 billion</strong> (↑40% YoY). The EPS miss was by a penny or two, and a soft Q1 operating income guidance (<strong>$16.5–21.5 billion</strong> vs. $22.2 billion expected) contributed to the stock’s descent.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wyIQudCU-NJYqNS1UJENbg.png" /></figure><p>Beyond the core business, Amazon’s other segments are putting in the work. Back in June of last year, I highlighted Amazon’s retail metamorphosis as the segment to watch, and the momentum has only continued. North America retail now generates <strong>$11.5 billion</strong> in operating income and is within striking distance of AWS’s <strong>$12.5 billion</strong>. My thesis around Amazon as a future leader in physical AI and robotics remains unchanged.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yhF55NBNXvYe_mM3w78ZQA.png" /></figure><p>Outside of Amazon’s usual bread-and-butter money printers, the company is also making serious progress in becoming an AI enabler for its customer base. Amazon Bedrock, their managed platform that lets businesses plug into any foundation models to build without taking on the underlying infrastructure, has quietly become a multi-billion dollar run rate business. Demand is through the roof, and customer spend is growing <strong>60% QoQ</strong>. The custom silicon business also crossed a <strong>$10 billion</strong> annual run rate and has grown by triple-digit percentages. While they haven’t gotten as much press as Google’s TPUs, Amazon has been partnering heavily with Anthropic to optimize Trainium2 and Trainium3, which promise <strong>30–40%</strong> better price-to-performance than comparable GPUs.</p><p>However, when all is said and done, I don’t want to just dismiss that <strong>$200 billion</strong> in CAPEX with a mere hand wave. $200 billion against <strong>$80 billion</strong> in 2025 operating income means Amazon is deploying roughly 2.5 years of future earnings in a single year. It’s absolutely mind-blowing that a company that prints money like Amazon will almost certainly need to take on debt or watch free cash flow go negative to fund future AWS and AI growth. Also, while AWS revenue grew <strong>24%</strong>, operating income only grew <strong>17%</strong>, which means margins are actually compressing. All that new infrastructure will hit the P&amp;L before utilization fully catches up, making Amazon a lot riskier, less predictable, and less attractive compared to its Mag 7 peers.</p><p>In acknowledgment of the heightened risk and growing opportunity cost of holding Amazon over a Google or Meta, I am downgrading the company from an <strong>A to an A-</strong>. To me, this is actually the most <em>Amazon</em> thing Amazon has ever done. It’s literally business as usual. We’re talking about a company that ran razor-thin margins for a decade in order to fund their e-commerce business while investors were pulling their hair out. Wall Street wanted to break off the retail segment and turn Amazon into a pure cloud play, but the aggressive reinvestment into fulfillment turned NA retail into a cash cow that is one step closer to rivaling AWS. The CAPEX numbers have gotten bigger, but the playbook is absolutely the same.</p><p>As Jassy and team undergo what might be the largest investment cycle in Mag 7 history, investors need to reassess their risk tolerance and time horizon on holding AMZN stock. Past performance doesn’t promise the same future returns, but I absolutely will not be betting against AWS, Prime, and the track record of this management team. Even with the recent drop, my cost basis is around <strong>$202</strong>, and I will be accumulating shares after the dust settles from this earnings season.</p><h4>Earnings Highlights</h4><ul><li><strong>Financials:</strong> Total revenue grew 14% YoY to <strong>$213.4 billion</strong>, beating the ~$211.5 billion consensus, but EPS of <strong>$1.95</strong> missed the $1.97 estimate despite net income rising 5.9% to $21.2 billion. Operating income overall jumped 18% to <strong>$25.0 billion</strong>, with North America retail margins expanding to 9.0% and almost bringing in the same amount of operating income as AWS ($11.5B vs $12.5B). Q1 operating income guidance of <strong>$16.5–21.5 billion</strong> came in below the $22.2 billion estimate.</li><li><strong>CAPEX:</strong> Amazon is ramping 2026 spend to approximately <strong>$200 billion</strong>, up from $131 billion in 2025 and well above the $146.6 billion analyst estimate. Amazon now holds the crown for highest CAPEX commitment out of the Mag 7 as they look to continue AWS momentum and establish themselves as an AI leader. Jassy framed the spend as a no-brainer: <em>“As fast as we install this AI capacity, we are monetizing it… this is a very unusual opportunity.”</em></li><li><strong>AI Monetization:</strong> AWS grew 24% YoY to <strong>$35.6 billion</strong>, generating <strong>$12.5 billion</strong> in operating income with a 35% margin — its fastest growth rate in 13 quarters. Amazon Bedrock, their managed AI services segment, is now a multi-billion dollar run rate business with customer spend growing <strong>60% QoQ</strong>. Similar to Meta, AI is also supercharging Amazon’s ad business. Advertising revenue grew 22% YoY to <strong>$21.3 billion</strong>, with their AI shopping assistant Rufus making Prime shoppers 60% more likely to complete a purchase, while Prime Video’s ad-supported audience has grown to 315 million viewers globally.</li></ul><h3>Nvidia (NVDA) Grade: Reporting Feb 25</h3><p>I’ve been around the block enough to know that NVIDIA earnings is more of a minefield than a catalyst no matter how impressive the results actually turn out to be. Analysts are expecting <strong>$65 billion in Q4 revenue</strong> and <strong>$1.52 EPS, </strong>and<strong> </strong>anything short of that will be seen as a stumble. Keep in mind that most of my position has been accumulated from post-earnings dips, so I’ll be ready to pull the trigger if my investment thesis remains unchanged.</p><p>We already have the rest of Mag 7’s CAPEX as a decent proxy for AI infrastructure. I know I sound like a broken record, but pretty much every hyperscaler is shoveling cash into the fire to keep up with customer demand. However, let’s take a look at a few other canaries in the coal mine that are positioned further up in the supply chain.</p><p>ASML reported Q4 bookings of <strong>€13.2 billion</strong>, more than double analyst expectations and a new record for demand. CEO Christophe Fouquet, who typically does his best to temper investor expectations, had this to say:</p><blockquote>“In the last months, many of our customers have shared a notably more positive assessment of the medium-term market situation, primarily based on more robust expectations of the sustainability of AI-related demand.”</blockquote><p>Sustainability and AI have never gone hand in hand, but according to ASML, customers are confidently pulling plans forward and buying capacity for the medium and short term. ASML raised 2026 guidance to <strong>€34–39 billion</strong> and expects EUV revenue to rise significantly. If the premier arms dealer for all the global foundries manufacturing NVIDIA products or ASICs is seeing record orders, that’s a pretty strong argument for semiconductor demand.</p><p>TSMC, one of those foundries and NVIDIA’s main partner in crime for supplying chips, reported Q4 revenue of <strong>$33.7 billion (up 25.5% YoY)</strong> with net income surging <strong>35%</strong>. CEO C.C. Wei described AI demand as <strong>“voracious”</strong> and shared that he spent months personally speaking with global cloud giants before issuing the following take:</p><blockquote>“AI is real. It is starting to grow into our daily life. This is driving need for more and more computation, which supports the robust demand for leading-edge silicon.”</blockquote><p>TSMC guided for <strong>~30% revenue growth in 2026</strong> and raised CAPEX to <strong>$52–56 billion</strong> in an attempt to meet demand. High-Performance Computing (HPC), the segment that tracks GPUs and AI accelerator sales, now accounts for <strong>58% of TSMC’s revenue</strong> (up 7% YoY) and is also the fastest growing segment. Capacity could be a temporary roadblock to TSMC’s success, but it’s clear that the demand for AI is not the issue at hand.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZwBrtuFmVz5M_nOs3PIlaQ.png" /></figure><p>Even amidst the political whiplash on Chinese export controls for NVIDIA’s H200, demand signals are just as strong overseas. Despite President Xi’s call for <a href="https://www.reuters.com/world/us/us-plans-mandate-11-ratio-domestically-manufactured-imported-chips-wsj-reports-2025-09-26/">domestic self-reliance on chips</a>, Chinese companies have already placed orders for over <a href="https://www.reuters.com/world/china/china-gives-green-light-importing-first-batch-nvidias-h200-ai-chips-sources-say-2026-01-28/"><strong>2 million H200 chips</strong> for 2026</a>, with ByteDance alone willing to spend <strong>$14.3 billion</strong> (100 billion yuan). No one can predict how the China situation will shake out, but keep in mind that the market represented <strong>$17.1 billion</strong> in revenue back in 2024 despite existing restrictions. If China reopens, analysts believe H200 access could nearly triple that figure.</p><p>Then there’s Rubin. In a market that’s on its knees for better economics on AI infrastructure, CEO Jensen Huang may have already found the cure. Unveiled at CES earlier this year, the Vera Rubin platform promises up to <strong>5x the inference performance</strong> and a <strong>10x reduction in cost per token</strong> compared to Blackwell. If hyperscalers can scale their AI platforms with a fraction of the hardware, the CAPEX conversation changes entirely.</p><p>Especially after digesting the most recent earnings, I doubt this means Big Tech will spend less. However, Rubin represents a solution where NVIDIA customers could force multiply every dollar spent on AI infrastructure. We’ll see if the economics are enough for hyperscalers to temporarily put down their custom silicon passion projects, or for OpenAI or Anthropic to deepen their ties further with the king of AI infrastructure.</p><h3>Closing Thoughts</h3><p>With that, this wraps up another Big Tech earnings season and another edition of the Mag 7 Megapost. As an investor, I get a ton of joy and satisfaction out of writing about developing trends and staying a couple steps ahead of the market. And while I’m still bullish on the software sector as the best distribution channel for enterprise AI, the conviction behind the CAPEX commitments and renewed focus on infrastructure despite investor sentiment illustrates that we’re still in the beginning stages of the biggest tech investment cycle we’ve ever seen.</p><p>I get why investors want to get off the rollercoaster, but personally, I’ve always preferred off-season visits and no lines at the amusement park. I remain a believer in the technology, and I still think we’re just starting to see what AI can do. I can’t wait to bring you my take on the SaaSpocalypse selloff in a couple of days.</p><p>Thank you for reading. Until next time!</p><h4>New Here? Me Too! Let’s Connect.</h4><p>If you’ve made it this far, let me know what you think of the article by leaving a comment below! Investing has been a solo journey for me for many years, and while it’s been fun, I’m hoping to cultivate a new community of like-minded investors here on Medium. No matter your opinion, I genuinely want to hear it. Thanks again for reading!</p><p><em>This blog is for informational and entertainment purposes only. I’m sharing my personal thoughts and experiences, not financial advice. Always do your own research and consult with a qualified financial professional before making any investment decisions.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3a155711e46b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-investors-handbook/mag-7-megapost-winners-waiters-watch-outs-part-2-3a155711e46b">Mag 7 MegaPost: Winners, Waiters, &amp; Watch Outs (Part 2)</a> was originally published in <a href="https://medium.com/the-investors-handbook">Investor’s Handbook</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>