AI Acquihires: Why Big Tech May Buy Talent Instead of Companies

The largest technology companies are spending more on artificial intelligence (AI) than on almost anything else in their history, yet they are buying very few of the startups building it. Capability is still moving toward the incumbents, just through channels other than acquisition, and one of those channels is people. In this blog, learn more about AI acquihires, why Big Tech may be paying for AI teams rather than the companies around them, and what that pattern may mean for investors.

AI Acquihires: Why Big Tech May Buy Talent Instead of Companies

Buying the Team, Not the Company

The biggest names in technology have spent the past two years acquiring AI capability without doing much acquiring. In July 2025, Google paid $2.4 billion in license fees for nonexclusive use of Windsurf’s technology and hired the code-generation startup’s key staff, without acquiring the company. A month earlier, Meta had committed $14.3 billion for a 49% stake in Scale AI. Meta reportedly offered one researcher a package worth as much as $1.5 billion over at least six years, an offer he turned down.[i]

In December 2025, Nvidia paid roughly $20 billion to license Groq’s AI accelerator technology and hire key members of its engineering team, leaving Groq’s inference business formally intact but no one in the driver seat.[ii] These deals never triggered a formal acquisition, which leaves investors with an open question: what is their equity actually worth once the company’s core technology and people are already gone?

Why the Team Became the Asset

In a talent-focused deal, a buyer typically takes a non-exclusive license to a startup’s technology and extends offers to its founders and core technical staff, while the startup itself continues to exist as a legal entity. No shares change hands, which is why these arrangements are sometimes described as reverse acquihires rather than acquisitions.

The appeal may come down to what has stayed scarce. While overall hiring at major technology companies has run roughly 25% below its 2019 baseline, engineering hiring has held up far better, down about 11% over the same period, and engineers now account for a majority of new hires at those firms. Assembling a team of that kind internally can take considerably longer than bringing over one that already works together.[iii]

Record Spending, Very Few Purchases

The same companies competing for these teams are deploying capital at a scale the industry has never seen, and where that money goes is revealing. Combined AI infrastructure spending by Google, Microsoft, Apple, Amazon, and Meta is projected to reach nearly $600 billion in 2026. Over the same stretch, acquisition activity across the Big Five has stayed well below historical levels, hitting a decade low of seven deals in 2024 before recovering only modestly to 14 in 2025 and 12 as of May 2026.[iv]

Rather than buying AI capability outright, these companies have generally redirected capital toward infrastructure build-out and large strategic minority investments. Traditional mergers and acquisitions have not been the primary way capability has changed hands in this cycle, which has pushed attention toward other structures, including arrangements that pair a technology license with hiring a startup’s team.

What It Can Mean for Investors

The exit market has been active but uneven. Startups on Carta completed 421 exits through M&A in the first half of 2026, the busiest first half on record and a 16% increase year over year, while total venture-backed exit value topped $2 trillion in the same period with a single transaction accounting for most of it. Volume and value have not been distributed evenly across the market.[v]

For investors, the structural detail matters more than the headline figure. Because a talent deal pays a licensing fee to the company rather than purchasing its shares, the proceeds still travel through the capital structure. Liquidation preferences, share class, and the valuation at which an investor entered can all shape what any individual holder receives. What remains afterward is also worth weighing. A company that has licensed its technology and lost its founding team still owns its intellectual property, but it may be a materially different business than the one investors originally backed.

The Regulatory Question

These types of transactions are no longer operating quietly. In September 2026, the New York Times reported that the Department of Justice had opened an antitrust probe into the $20 billion Nvidia and Groq licensing deal.[vi] Congress has taken an interest as well. In February 2026, Senators Warren, Wyden, and Blumenthal asked the Federal Trade Commission (FTC) and the Department of Justice to scrutinize the deals, arguing that they function as de facto mergers that consolidate talent and resources while bypassing the review normally applied to acquisitions.[vii]

Final Thoughts

Record AI spending has not translated into a wave of startup acquisitions, and capability has been moving toward the largest buyers through other routes. For investors in private AI companies, the structure of a transaction may matter as much as its size, since a licensing payment and a share purchase can produce very different results for the same cap table. Investors may want to consider how much of a company’s value sits with its team rather than its product when assessing this kind of risk.

Are you ready to invest in startups? Sign up for a MicroVentures account to start investing!

Want to learn more about investing in startups? Check out the following MicroVentures blogs to learn more:

• Beyond the AI Trade: Sectors That Offer Diversification in a Concentrated Market
• The AI Training-to-Inference Shift: Where the Next Wave of Investment Opportunity May Lay
• Building a Diversified Startup Portfolio
• Inside Austin’s Startup Scene: A Look at the Capital, Talent, and Companies Shaping the Ecosystem
• The Unicorn Graveyard: Famous Startups That Were Valued at $1B+ and Still Failed