Hedge Funds Are Betting Billions on AI-Driven Companies

On a weekday morning in early 2026, stroll around the trading floor of any major lower Manhattan firm, and you’ll notice that the conversation is especially serious. Not quite panic. It is more akin to the focused awareness of individuals who perceive a substantial sum of money moving in a single direction and are attempting to make a snap decision based on insufficient information about whether to be in front of or behind it. Artificial intelligence is clearly the direction. Since Goldman Sachs began monitoring hedge fund positioning in 2016, the rate at which capital is pursuing it—through semiconductor stocks, new hedge funds, and fundraising rounds that would have seemed unthinkable two years ago—is unprecedented.

Hedge fund exposure to AI-related tech hardware had reached its highest level in almost ten years, driven mostly by long positions in semiconductor and chip equipment stocks across US and Asian markets, according to Goldman’s data, which was released in late 2025 and confirmed what many in the market had been sensing for months. Quietly starting in September 2025, the move away from the so-called Magnificent Seven—the small number of mega-cap tech companies that dominated portfolio construction for years—and toward more focused semiconductor bets picked up speed in the fall. It implies that the investors who are transferring this money aren’t just following the AI theme in general; rather, they are making a more focused argument and placing their bets on the physical infrastructure layer of AI rather than the software programs that are built upon it. This distinction is more important than it may first appear.

It is really challenging to put the fundraising figures that were circulating in early 2026 into perspective. $122 billion was raised by OpenAI. $30 billion was earned by Anthropic. $20 billion was raised by Elon Musk’s xAI. The autonomous car startup Waymo made $16 billion. These are capital raises at a scale comparable to the GDP of small countries, flowing into businesses whose revenue base, according to at least one reliable estimate, amounts to roughly $60 billion against $400 billion in annual spending across the sector. They are not incremental funding rounds from a young industry still proving itself. More attention should be paid to the disparity between what AI companies are producing and what they are consuming than is usually the case in the breathless coverage of individual fundraising milestones.

Leopold Aschenbrenner, a 23-year-old former OpenAI researcher who gained some notoriety by publishing a widely read manifesto on AI development, entered this environment and decided to start a hedge fund despite having no prior experience with investing. Over $1.5 billion was raised by him. quicker than the majority of renowned portfolio managers who quit well-established companies to launch their own. That fact reveals something about investor appetite for AI exposure, not necessarily about Aschenbrenner in particular. The market is hot and the standard filters aren’t working when the quickest route to a $1.5 billion mandate passes through a 23-year-old who has never handled money professionally.

It gets really complicated on the other side of the AI trade. Hedge funds have been constructing some of the biggest short bets against software stocks in recent memory while simultaneously building long positions in semiconductor hardware. As the software industry lost about $1 trillion in total market value by February 2026, short sellers had made $24 billion from falling software stocks. By early February, the iShares Expanded Tech-Software ETF, a helpful stand-in for the industry, had dropped more than 21% year to date and was 30% below its peak in September 2025.

Companies that were once thought to be among the more enduring names in the technology industry, such as Microsoft, Oracle, Salesforce, Adobe, and ServiceNow, were all down 15 to 21%. According to hedge fund sources, the specific reasoning behind the short positions revolves around businesses that provide basic automation services that more advanced AI tools can easily duplicate, negating the need for a separate platform or subscription. This analysis might be accurate. Additionally, it’s possible that it’s being used too widely, with indiscriminate selling generating short-term opportunities that patient longer-term holders will eventually benefit from.

According to reports, Michael Burry, whose short position against mortgage-backed securities prior to the 2008 financial crisis became the focus of a book and a movie, has been accumulating a position betting against the AI industry to the tune of about $1 billion. After disclosing this, he quietly closed his fund. Even though the framing seems performative, Burry’s track record demands attention despite the theatrics. There are other skeptics besides him. The revenue-to-spending ratio of the AI economy, which was reported by The Atlantic to be $60 billion in versus $400 billion out, raised precisely the kind of structural question that is overlooked during momentum cycles but turns out to be prophetic later. Whether the AI investment wave is creating real long-term value at this scale or if some of it is outpacing what the underlying economics can sustain is still up for debate.

Observing all of this from the outside, it seems as though the market is making two concurrent, somewhat contradictory bets: that AI hardware infrastructure will be the trade of the decade and that AI tools will soon render significant portions of the current software economy obsolete. They could both be correct. Both of them could be mistaken. The money has already moved, the positions are already substantial, and the time will soon come for the market to balance the disparity between AI revenue and spending. The funds that recognized the difference between infrastructure and application—between what AI will eventually replace and what it needs to run—will probably be the ones remaining on the right side of the trade when it arrives.