The AI Acquihire Report

What this report finds

In March 2024, Microsoft agreed to pay Inflection AI $650 million. It did not buy the company. It licensed Inflection’s models and hired almost everyone who built them, including both founders. The structure was new, and within two years it had become the dominant way the largest technology companies absorb artificial intelligence startups.

For two decades, the technology exit had two doors. A company went public, or it was acquired. The reverse acquihire is a third door, and for the most valuable AI startups it has quietly become the main one.

Strip away the variations and the structure is remarkably uniform. Microsoft built the template in early 2024, and the deals that followed are recognizable copies of it. Understanding the four moving parts is the difference between reading these transactions and being surprised by one.

The cycle has a clear beginning and a recognizable rhythm. What follows is the transaction-by-transaction account, in the order the deals arrived, with the terms that are known and the gaps where they are not.

The license is the visible price. It is not the thing being bought. In every one of these deals the object of the transaction is a small number of people, and the price of those people, not the technology, explains the numbers.

A conventional acquisition pays everyone who holds equity, in an order they agreed to in advance. A reverse acquihire does not. It separates the people who built the company into those the buyer wants and those it does not, and it pays them on entirely different terms.

Windsor Drake’s research desk compiled this report from transaction data, public filings, and the firm’s sell-side advisory work in software, fintech, AI, and cybersecurity. It is intended to inform founders, owners, and acquirers evaluating a transaction, and does not constitute investment advice.