D is for Dilution
(Any views expressed here are the personal views of the author and should not form the basis for making investment decisions, nor be construed as a recommendation or advice to engage in investment transactions.)
Want More? Follow the Author on Instagram, LinkedIn and X
Access the Korean language version here: Naver
Subscribe to see the latest Events: Calendar
My eyes open. It’s Saturday morning, Halloween weekend in Tokyo, October 28 2016. The goon squad decamped from Wyndham Gai, boarded a plane, and became yet another rowdy bunch of gaijins on a stag weekend. I rejoice that I stumbled home in one piece to my comfy Roppongi hotel bed. The previous night, we ventured far and wide in Minato-ku. There is a funny picture of my now chief of staff and me at a “reputable” establishment, shitfaced, posing in what looks like a Lethal Weapon poster.
That Friday night, the Zcash genesis block occurred and marked one of the most insane trading days in crypto history. Similar to Bitcoin, to buy physical ZEC meant purchasing it from a miner.[1] The privacy narrative then, as is now, was hot. Everyone wanted to own a piece of the first scalable private digital cash network. The problem was that the supply was zero and only increased block by block as miners earned ZEC and sold to cover their operating expenses.
The dominant shitcoin exchange of the day was Poloniex. I have many fond memories of trading on Polo and coding against their absolutely dogshit API. If you ever got a response from a customer service rep, it was like being spoken to by God. Circle subsequently bought Poloniex for an exorbitant price a few years later and ran it into the ground before Justin Sun picked up the pieces. But I digress. Remember degens, price is what you pay, value is what you get.
To buy ZEC, one had to trade on Poloniex. Therefore, as drips and drabs of ZEC appeared on offer at Poloniex, eager shitcoiners snapped them up at insane prices. As you can see by the below chart, most degens who purchased ZEC in the first days and weeks of trading are still underwater.
Traders might retort that you must adjust for the circulating supply to ascertain the “true” performance of ZEC. While it is correct that the ZEC market cap is at all-time highs, as a trader, you don’t trade the market cap; you trade the price per token. This is a key distinction that many investors fail to appreciate, which results in poor returns as supply increases quickly and dampens the price per token or share even if the market cap increases.
Migrating to AI, do the supply dynamics of the public equities of SpaceX and soon to list dastardly duo, Anthropic and OpenAI, portend to profit or poverty? Poverty you fucking plebe! I argue that, similar to early ZEC investors, if you bought or buy a share of these IPOs at their offering price, you will lose money as the insiders unlock and dump their shares. This presents an issue for the general market environment because the entire AI trade rests on vibes. The vibes surrounding whether you believe the lofty future profitability and ubiquity of AI determine whether you lend money to the industry. These vibes emanate from the only objective measure of AI’s success, whether the stonk price is Number Go Up or Down. Therefore, without any discussion whether the CEOs of SpaceX, Anthropic, and OpenAI are lying or committing accounting fraud, I know the market will assume they are once the stock prices sag below the IPO price because of massive insider unlocks. Once I show you the terrible maths which will drive investor disillusionment, then I can lay out a trading strategy as global liquidity deteriorates before an eventual bailout.
SpaceX and the Dastardly Duo
The three AI companies to focus on are SpaceX, Anthropic, and OpenAI because they consume the vast majority of compute as they produce intelligence tokens. If their demand for compute slows, the entire AI narrative collapses. Let’s dig into their current and anticipated stock performance by looking at the supply of freely tradable shares. But before we do some maths, I want to emphasize why the price of the equity is so important to the AI story.
These companies all must spend hundreds of billions of dollars building capex to power their almost-God silicon creation. Because they were not public companies, raising corporate debt in the sizes required was impossible. They required some publicly listed hyperscaler or semiconductor company to backstop or guarantee the debt. The equity raises, while large, paled in comparison to the debt issued. Once the labs become public companies, they can issue corporate debt. SpaceX is not a pure play AI lab, but listing allowed them to secure an investment grade rating and issue tens of billions of dollars of debt since the IPO. Credit spreads tighten and amounts issued rise as these companies get access to institutional pools of capital. The willingness of investors to accept affordable yields depend on their perception of the future viability of an AI lab. Looking at their publicly traded stock prices is the best way to determine whether the market believes in the AI story and a particular lab’s ability to profit. If the stock slumps, which is almost mathematically assured because of the large notional of stock to be sold, then debt investors demand higher yields if they are willing to invest at all. It is this reflexive process that motivates Elon, Dario, and Sam to take their companies public. They will lie, cheat, and steal (in legal ways, obviously, they are gentlemen, let’s not forget) their way to a public listing because failure to do so means certain death for their capital-intensive debt-fueled business.
SpaceX (Ticker: SPCX)
The above chart shows the percentage change of the stock price vs. the number of shares outstanding from the IPO date to the present. As you can see, the stock price flatlined since the IPO, while the float increased by 50%.
Between now and the end of June 2027, an estimated 8.6 billion additional shares will unlock, which is roughly another 50% dilution; most of the increase is because of Elon’s shares vesting. To add to that, if Elon completes the rumored SpaceX & Tesla merger, it could mean an almost additional 50% dilution of Class A shareholders.
The only thing that could save the bulls would be if the company magically generated an insane amount of free cash flow. Generating profit isn’t enough. Elon must guide the market that this cash would be used to buy back stock or issue dividends. Short of that, there is no reason to own the stock in the face of massive assured dilution.
One done, two to go. Let’s move onto the pure expression of AI, Anthropic.
Anthropic (Ticker: ANT)
Anthropic is a pure play frontier AI lab. It trains closed-weight models and then sells access to this intelligence token by token. It is a one-time consumption business. Spending on past tokens says nothing about future spend. This is something many investors cannot fathom. They want to apply SaaS metrics like ARR when they should think of Anthropic as an oil & gas company.[2]
Earnings quality aside, what are the supply dynamics of the upcoming IPO?
Reading through the recently filed and updated S-1 (honestly speaking my agent did it), I estimate that initial IPO inventors face 83% dilution by June 2027. Let that sink in degen. Eighty-Three Fucking Percent Dilution! Wowzers. By June of next year, Anthropic will not be profitable and even if they were, the S-1 clearly states there are no planned dividends in the near future. All available cash flows will be used to fund the roughly $900bn of compute and possible additional spend on data center buildout outlined in the S-1. Purely driven by this insane dilution schedule, it is almost impossible mathematically for the stock price to go up between listing and June 2027.
But, impossible is what investors believe when it comes to Anthropic. Therefore, if the market cap at IPO is $1.5 trillion, then in order for the stock price to standstill requires ~$1.245 trillion of new capital to invest in Anthropic to absorb all the potential selling. Not impossible, but improbable.
And now ladies and germs, let’s discuss the worst of the bunch, OpenAI.
Get Arthur Hayes’s stories in your inbox
Join Medium for free to get updates from this writer.
OpenAI (Ticker: TBD)
Assuming OpenAI can get their IPO out the door before investors learn their lesson about aping into cash flow negative mega AI IPOs, it will be the worst of the bunch. Its cash burn is worse, product worse, and customer concentration worse than Anthropic. But even as the runt of the litter it still requires prodigious amounts of capital to build its silicon almost-God.
As of right now, OpenAI is reportedly attempting to raise $20bn, down from $30bn, at a $1.25tn valuation down from $1.5tn, and if the recent FT story is to be believed, it’s proving difficult. That situation should improve if Anthropic can successfully complete its IPO and the stock rises post listing.
We don’t have an official S-1 filing from OpenAI, but it’s safe to assume that its offering will mimic SpaceX and Anthropic’s. They will float roughly 5% of the total supply, and allow a significant number of employees and early investors insiders to dump their shares over the next six months. Because this is a dick swinging contest for micro-penises, Sam will probably attempt to value OpenAI around $1.5tn to match his pal Dario.
Let’s assume the IPO happens in January, so that by June 2027, a similar situation as Anthropic unfolds where it requires hundreds of billions if not trillions of dollars of new capital to keep the stock hovering around its IPO price.
Sideways is Down
As I mentioned earlier, the point of the IPO is to gain access to the corporate debt markets so that these companies can raise capital without hyperscaler and semiconductor company backstops. To inspire confidence in potential debt donkeys, the stock cannot just tread water but needs to pump. As I showed earlier, treading water is extremely capital intensive for the broader market. The most likely scenario is that investors aren’t as stupid as these CEOs believe, and refuse to be dumped on by insiders. Therefore, even if revenues (there ain’t going to be any earnings, lol) match expectations, the supply onslaught slowly wears on the stock price as it leaks lower.
Once these companies stand on their own feeble legs, can Sam even squat the bar?, how wide will their debt offerings price to high yield average spreads? There is no way their debt will stay investment grade (SpaceX) or avoid a junk classification (Anthropic and OpenAI). The problem is once the market knows the price to fund the end consumer of compute, everyone upstream will face increasing difficulty selling their AI capex debt at anything other than nose bleed wide levels. And then negative reflexivity destroys what remains of the AI bull story as far as stock price returns go. The wider credit spreads are, the more interest expense paid to finance the build out, which reduces eventual profits.
Too Late
All you AI bulls with your knee pads on listening to Sweet Tooth by R Kelly please take note. I don’t give a fuck whether AI is useful, or the revenues surprise to the upside, or some productivity statistic you made up yesterday points to more corporate adoption. Because I can now see in real time what the market thinks of SpaceX, Anthropic, and OpenAI via their share price, and that’s disastrous for the bull narrative. It is disastrous because they are dumping so much stock on the market it is too difficult mathematically for Number Go Up. Debt investors will take cues from the equity market and demand insanely high yields if they are even willing to participate at all. In addition, the debt covenants will be onerous; e.g. mandating a certain buffer of cash on the balance sheet or acceleration of repayment.
In short, God appeared after the pilgrims fled the pews.
This is bad news bears for all other holders of AI related debt. If the credit markets freeze due a deluge of shitty high yielding debt issues, then the entire industry vertical suffers. The rating downgrades begin, and it’s a classic Wiley Coyote moment.
Bank Balance Sheet Contraction
Why do we care as crypto degens if the AI story will implode because of SpaceX, Anthropic, and OpenAI’s slumping share prices? We care because fiat liquidity, specifically the dollar flavor, will decline as AI credit spreads widen and issuance dries up.
Either the Fed or US commercial banks must together or by themselves grow their balance sheets to provide the liquidity needed to pump our crypto bags. The Fed, at least for now, is on strike. Warsh the Weasel and Buffalo Bill Bessent agree that the banking system should create credit and not the Fed. Therefore, an increase in the US banking system’s balance sheet created the favorable liquidity backdrop for the mini crypto pump that began in August. But, as of last week, the banking system’s balance sheet fell WoW. I know this is only one print, but set against the backdrop of negative AI credit market vibes, you can see why I’m worried.
Banks’ balance sheet asset growth helps the AI bubble. Lending: banks lend to all things AI and AI-adjacent. Securities financing: banks provide margin loans to stonk speculators. Proprietary trading: banks also buy stocks and bonds on their own account. Globally, no matter where you invest, unless you are war profiteering, you are speculating on the AI bubble. Therefore, if the AI vibes sour because of falling stock prices of these three companies and the credit markets subsequently freeze or cool, then banks’ balance sheets will at best flatline and at worse fall. This is the transmission channel from AI to crypto.
Clarity Sucks
To repeat, the problem isn’t that these companies light investors’ money on fire to create a silicon almost-God. The problem is that a liquid stock price that mathematically will probably fall because of aggressive dilution becomes the common knowledge messenger. Pulling up a real time price that is dumping shatters the belief that the AI trade is up only. Watching these pitiful charts inspires introspection about the stupendous assumptions built into the AI bull narrative. Irrespective of whether the labs eventually turn a GAAP profit or AI becomes a general-purpose technology, the fact remains that today the only objective measure of AI’s success is trending down and to the right.
If these companies could shirk the searing gaze of public market scrutiny for another few years while AI earns its keep, they would. But they cannot raise the hundreds of billions to single digit trillions of dollars necessary without access to the corporate debt markets. Entrance to this arena requires one to list publicly. Time is up, and the two seven off suit bluff with a broken straight on fourth street comes up short against a pair of pocket rockets that hit trips on the river. The ace is the scrutiny and transparency of public markets, and will always triumph over the bluff of a scallywag.
Absent any dramatic change in the fiat liquidity picture, Maelstrom will chill on the sidelines and watch the implosion of AI mega IPOs under the weight of dilution. The bailout will come. When and how are the pertinent questions to be answered in subsequent essays. Changing the behavior of the Fed and all other central banks requires the glimmer of a financial crisis. The AI bubble is global; either you are building data centers, or exporting the raw materials to construct them and or the semiconductors bring them to life. The global economy is figgity fucked without the AI build out. Therefore, every major central bank, monetary, and fiscal authority will print money to forestall a crisis. It will be glorious for Bitcoin and crypto hodlers. But be patient.
While we wait, compute will flood the spot markets as more data centers begin operations. Anthropic and OpenAI have no choice but to buy compute even if training and serving AI models feature negative unit economics. This powers an exponential increase in AI agentic workflows as their raw food, compute, falls in price. The FLOP Network will benefit, alongside another AI project Maelstrom advises, called Genlayer. Watch out for their upcoming TGE, where if you so choose, you may own a piece of the adjudication layer of the agentic economy.
Toodoloo you sexy mother fuckers. Enjoy the trenches.
Want More? Follow the Author on Instagram, LinkedIn and X
Access the Korean language version here: Naver
Subscribe to see the latest Events: Calendar
