When Every Company Is an AI Company
We’re retiring the Swift AI Index as it approaches its second anniversary. We launched it in 2024 as a side project with a simple thesis: public companies making serious investments in AI should outperform. We built an AI system to look at hiring, research, open-source contributions and financial results to find companies that were putting real resources behind their AI ambitions.
The index’s tracked performance from January 3, 2024 through this week (September 28, 2026) is up 115%, versus 63% for the S&P 500. The top ten companies in the index added roughly $15 trillion in market cap over that period. Chips drove much of the growth, but software companies like AppLovin and Palantir were part of the story too. We’re pretty happy with that result.
AppLovin is a good example of what we were looking for. Its investment in AI helped advertisers find customers more efficiently, giving them a reason to spend more on the platform. In 2024, its advertising revenue grew 75% to $3.2 billion, while advertising adjusted EBITDA grew 91%. AI was improving the thing customers already paid it to do, and the results were showing up in the business.
What’s changed since 2024 is how widespread that AI investment has become. Back then, we found that 70% of the companies we tracked had hired no AI roles at all. Today, that number is only 15%. And when listening to earnings calls the focus on AI initiatives has radically changed, too. The average number of times CEOs mentioned AI rose from 10 per call to 17, with some reaching 150! There’s certainly no shortage of enthusiasm.
Get Swift Ventures’s stories in your inbox
Join Medium for free to get updates from this writer.
Years ago, we used to tell investors that every company was becoming an AI company. That’s increasingly true, which makes it less useful as a way to decide where to invest. Hiring AI engineers and funding research still matter, but two companies can make serious commitments and get very different results. One might find a new source of growth, while the other spends a fortune just to hang on to its customers.
The more interesting question is what happens to the business as AI gets better. If a software company, for example, can build a better product that helps customers do the same work with fewer people then they might find themselves needing fewer seats. Similarly, a services firm might become more productive and discover that customers expect the savings passed along in lower prices. AI can work exactly as promised without making the company selling it more valuable.
There’s also less need to build everything yourself. A small team using models developed elsewhere can tackle problems that once required a large research budget. That makes the size of a company’s AI team a less reliable guide to its potential. You have to understand what it can do for customers, what they’ll pay and why a competitor can’t easily do the same thing.
Those are the questions we spend our days on as investors in private AI companies. There’s still an enormous gap between what models can do and what most businesses can actually get done with them.
Some of the biggest opportunities may be in work businesses barely do today because it’s too expensive: checking every customer account for a sales opportunity, testing thousands of product variations or constantly comparing suppliers. If agents make that work cheap and reliable, customers can do much more than they could justify paying people to do. That gives startups a chance to build new markets, provided the extra work produces results customers will pay for.
We’re proud of the index’s results. It was a side project that grew out of the work we’ve always done at Swift: figuring out who can turn advances in AI into a great business. We’re retiring the index, but that question keeps getting more interesting and we get to work on it every day with the founders we back.
