In June 2025, Meta agreed to pay $14.3 billion for a 49% stake in Scale AI — and pointedly did not acquire it. Founder Alexandr Wang and a small group of senior staff left for Meta’s superintelligence team. Scale kept operating as an independent company under a new interim CEO. Meta got the person it wanted and preferred access to the data pipeline it needed, while taking a stake with no voting power. What it carefully avoided was buying the company outright.

That structure now has a name — the reverse acqui-hire — and it has happened at least five times in two years.

The pattern

The template is consistent. A large incumbent hires a startup’s founders and a slice of its top engineers, licenses the underlying technology for a headline fee, and leaves the corporate shell standing with its remaining staff and a cash payout for investors. No merger is filed.

Microsoft ran it first. In March 2024 it took Inflection AI’s co-founders and roughly 70 employees, paying about $650 million — $620 million to license the models and $30 million to settle any legal claims, money Inflection used to repay its investors. Amazon did the same with Adept; Google did it with Character.AI in a reported $2.7 billion deal that licensed the startup’s models and brought its founders back in-house. Then in July 2025, Google paid $2.4 billion for Windsurf’s CEO, co-founder, and key researchers plus a non-exclusive license — a coding startup reportedly at $82 million in annual recurring revenue with over 350 enterprise customers.

The Windsurf deal exposed the mechanics most clearly. Google’s $2.4 billion split roughly in half: $1.2 billion to investors and $1.2 billion in compensation for around 40 departing employees, with the largest shares going to the founders. Days later, the rival startup Cognition bought what remained of Windsurf — the product, the IP, and the employees left behind. The company that built the asset was, by then, a leftover.

Why it matters

For founders, this is a genuinely new liquidity path: you can “exit” your key people for nine or ten figures without selling the company, and without the regulatory friction of an acquisition. For early and preferred investors, it can return capital quickly. But it fractures the cap table’s interests. The people who leave get paid in compensation packages; the people who stay inherit a business stripped of its founders and its best engineers, with a licensing cheque and an uncertain future.

It also looks, to regulators, like a way to dodge merger review. The FTC has openly questioned whether these structures are designed to avoid antitrust scrutiny, and the DOJ began examining Google’s Character.AI arrangement on the same logic. When a deal walks, talks, and pays like an acquisition but is filed as a licensing agreement, the question of what was actually bought becomes a live legal one.

The Charaka View

In Manthan Intelligence’s knowledge graph, these transactions don’t fit the “acquired_by” relationship type that normally closes out a company’s history. We tag them as a distinct edge — talent-and-license transfer — precisely because the company keeps existing while its value walks out the door. That modelling choice reflects the deeper lesson: in frontier AI, the defensible asset is increasingly the founder-engineer cluster, not the product or the corporate entity. The model can be licensed; the people are what’s actually being bought.

For anyone assessing an AI startup, the pattern reframes key-person risk from a footnote into a central variable. A company whose entire value can be extracted by hiring eight people and licensing a model is not a company with a moat — it’s a roster with a runway. The strongest AI businesses will be the ones where the value survives the departure of the founders, because the data, the distribution, or the workflow lock-in stays behind. Everything else is an acqui-hire waiting for its bid.

This analysis draws on CNBC’s reporting on the Meta–Scale AI deal, AI Business on the Microsoft–Inflection FTC inquiry, PYMNTS on Google–Character.AI, CNBC, TechCrunch, and Built In on Google–Windsurf, and the American Action Forum on FTC scrutiny of reverse acqui-hires. Human editorial oversight applied.

This analysis is informational and does not constitute investment advice, a research report, or a recommendation to buy, sell, or hold any security.

Charaka Notes by Manthan Intelligence. Subscribe