Last Updated on July 7, 2026 by Taya Ziv
Mustafa Suleyman co-founded DeepMind in 2010. Google acquired it for $500 million. He spent years building AI products inside Google, then left to start something of his own. Inflection AI raised $1.5 billion, built a chatbot called Pi with millions of users, and had one of the most decorated AI research teams on the planet.
Then Microsoft called. Not with an acquisition offer. With something stranger.
In March 2024, seventy percent of Inflection’s staff walked out the door and into Microsoft. Suleyman became CEO of Microsoft AI. The price tag was $650 million, but $620 million of that went to investors as a “licensing fee” and $30 million was literally a payment for “waiver of legal rights related to mass hiring.” What remained of Inflection was 12 employees, a chatbot running on borrowed time, and a new CEO no one had heard of.
The word for this is “acqui-hire.” But the old acqui-hire was a failed startup selling for parts. This is something different. This is a $1.5 billion company getting strip-mined while technically still existing. And Inflection wasn’t alone.
Between March 2024 and January 2026, Google, Microsoft, Amazon, and Meta spent over $20 billion doing this same thing. They hired away the founding teams of seven major AI startups through “licensing agreements” designed to avoid one specific legal requirement: the antitrust review that kicks in on acquisitions above $120 million. They got the talent, the technology, and the competitive elimination. They just didn’t have to call it a purchase.
The Pattern That Should Worry Every AI Founder
The playbook is the same every time. Big Tech signs a “technology licensing agreement” with the startup. The founders and core engineers “voluntarily resign” and start at the new employer. Investors get partially compensated through the licensing fee. And the startup continues to exist on paper, with a skeleton crew and a future that’s measured in months.
Google did it three times: Character.AI for $2.7 billion, Windsurf (the company formerly known as Codeium) for $2.4 billion, and Hume AI for an undisclosed amount. What makes Google’s pattern especially revealing is who they kept buying back. Noam Shazeer co-authored the original Transformer paper at Google in 2017, left to build Character.AI, and came right back to Google DeepMind in 2024. Alan Cowen spent years at Google before founding Hume AI, then returned in 2026. Google is spending billions to undo its own brain drain.
Amazon did it twice, and more brutally. Adept AI cost just $25 million in licensing fees. That’s pocket change. Amazon hired 66% of the staff, including the CEO. Today, LinkedIn shows four remaining employees at Adept. A whistleblower filed an FTC complaint alleging that Covariant, Amazon’s other target, had become a “zombie company” existing solely to collect the final licensing payment.
Meta went bigger. Its $14.8 billion deal for a 49% stake in Scale AI was structured as an equity investment, not an acquisition. But Scale AI’s CEO Alexandr Wang joined Meta AI. The company that remained has been, as CNBC delicately put it, “rocky.”
What Happens to the People Who Stayed
Here’s the part that should make you uncomfortable if you’re building an AI startup and asking people to take equity over salary.
At Windsurf, the $2.4 billion deal went roughly $1.2 billion to investors and $1.2 billion to the approximately 40 people Google hired. Employees who had joined in the last year received zero payout. Around 200 people were left behind until Cognition AI picked up the pieces for about $250 million. And if you hadn’t exercised your stock options within the post-termination exercise window? Gone.
At Character.AI, the situation got darker. After the founding team left for Google, the company spiraled into a teen safety crisis. Two teenagers died after developing emotional relationships with Character.AI chatbots. Multiple lawsuits followed. The question nobody wants to answer: when you drain a company’s founding team for their talent, who takes responsibility for the product they left behind?
At Inflection, the 12 remaining employees pivoted to “Inflection for Enterprise.” Around 13,000 organizations expressed interest in the API. But turning interest into revenue with 12 people is, let’s say, a different kind of challenge.
Why This Structure Exists (and Why Regulators Are Playing Catch-Up)
The legal innovation here is almost admirable in its cynicism. Traditional acquisitions above $120 million trigger mandatory Hart-Scott-Rodino premerger notification. The FTC reviews the deal. Regulators can block it. Microsoft learned this the hard way with the Activision acquisition.
The licensing-plus-hiring structure sidesteps all of it. As Columbia Law Review’s analysis put it: Big Tech gets everything it wants, talent, technology, competitive neutralization, while maintaining the legal fiction that no acquisition occurred.
Three senators wrote to the FTC and DOJ in February 2026 calling these deals what they are: “de facto mergers, allowing the companies to consolidate talent, information, and resources, all while apparently attempting to bypass the scrutiny typically applied to mergers and acquisitions.”
The Federation of American Scientists was more direct: “If reverse acqui-hires become the default path for absorbing promising startups, the dynamic competition that has long defined the American technology sector risks being replaced by a cycle of defensive consolidation that suppresses innovation.”
But here’s the thing. Under the current administration, the FTC has taken a lighter touch. The legal infrastructure is building, but it’s building slowly. Meanwhile, the deals keep happening.
What This Actually Means for Founders
I know what you’re thinking. “This is a Big Tech problem. I’m building a $2M ARR SaaS tool. Nobody’s coming to acqui-hire me.”
Maybe. But the acqui-hire wave tells you three things about the market you’re building in.
First, your exit assumptions are wrong. The old playbook was: build something valuable, get acquired, everyone wins. The new playbook is: Big Tech takes your team, pays your investors just enough to avoid a lawsuit, and your remaining employees hold worthless paper. If your business model depends on a Big Tech exit, understand that the exit might not include your company.
Second, revenue is the only real defense. Every company on the acqui-hire list had one thing in common: massive funding, tiny revenue. Inflection had $1.5 billion in funding and no meaningful business model. Character.AI had 28 million monthly users and no path to profitability. The companies that Big Tech can’t strip-mine are the ones where the business is worth more than the people. If your entire value proposition walks out the door when your founding team leaves, you don’t have a company. You have a headhunting opportunity.
Third, the talent scarcity is the whole story. The number of researchers capable of training frontier models is measured in the low thousands globally. That’s it. Big Tech isn’t buying these companies because it wants their products. It’s buying the people because it’s literally faster to acquire an entire company than to recruit the individuals one by one. If you’re a founder with a strong AI research team, you are already in the acqui-hire pipeline whether you know it or not.
The Uncomfortable Historical Parallel
The common comparison is to Facebook buying Instagram and WhatsApp. But those were real acquisitions. The products survived and thrived.
The better parallel is the 2010-2015 AI talent acquisition wave. Google hired Geoffrey Hinton by “acquiring” DNNresearch, his two-person company. Baidu hired Andrew Ng. Facebook hired Yann LeCun. Those deals were pure talent plays too. Nobody cared about the company.
The difference now is scale. The companies being dismantled have hundreds of millions in investor capital, millions of users, and employees whose life savings are tied up in stock options. OpenAI alone has done 14 acquisitions since 2023, and most of those products are already dead.
The collateral damage isn’t measured in two-person research labs anymore. It’s measured in thousands of careers, billions in lost capital, and products that disappear because the person who built them took a better offer.
If you’re building an AI startup today, here’s the question that matters: are you building something that survives when your team gets a $10 million retention package to walk across the street? If the answer is no, you’re not building a company. You’re building a resume.
