What You’ll Find This Week

Microsoft, Google, and Amazon have spent more than $5.8 billion since 2024 hiring the founders and top talent out of AI startups, without formally acquiring a single one of those companies. No merger. No antitrust review. Just an employment offer to the people who mattered and a licensing check to the shell they left behind.

The label everyone reaches for is consolidation. What's actually happening is closer to disassembly, and the way the money moves tells you exactly who a startup is built to protect.

Here’s what you’ll find:

Don’t Miss Our Latest Podcast

This Week’s Article

In March 2024, Microsoft announced it was hiring the majority of Inflection AI's staff. Mustafa Suleyman, Inflection's co-founder, became CEO of Microsoft AI. Microsoft also struck a deal to license Inflection's technology.

But Microsoft didn’t acquire Inflection.

A few months later, Character.AI's founders, Noam Shazeer and Daniel De Freitas, both ex-Google researchers, returned to Google along with part of their team, under a separate technology licensing arrangement. Amazon ran the same play with Adept, hiring most of its staff directly and paying its investors back through a $25 million licensing fee instead of buying the company. By mid-2025, Google was running the same playbook again: it paid Windsurf $2.4 billion to license its AI coding platform and hire its founders and core team, leaving the rest of the company stranded until Cognition acquired what remained days later.

Four deals, four different companies, the same structure almost every time: hire the people, license the technology, leave the corporate skeleton standing. Between March 2024 and mid-2025, Microsoft, Google, and Amazon spent more than $5.8 billion running this exact playbook across Inflection, Character.AI, Adept, and Windsurf. They didn't formally acquire a single one of the companies whose talent they absorbed.

While $5.8B makes for an eye-catching headline, the more useful story is in the mechanics. Once you see how the money is actually structured, this stops looking like an acquisition in disguise and starts looking like a controlled dismantling, with a very specific list of who gets protected in the process. And who does not.

A normal acquisition has one buyer, one seller, and one price, distributed to everyone on the cap table according to the waterfall they signed up for. A reverse acquihire has two separate transactions happening at once, and they don't touch the same money.

The first transaction is personal: the acquirer extends employment offers directly to the founders and the employees it wants, usually with new titles, new comp packages, and a platform an order of magnitude bigger than what they're leaving. This money goes to individuals, not to the company.

The second transaction is corporate: the acquirer pays a licensing fee to the corporate shell of the former startup for rights to use its technology. This money goes to the company, which means it flows through the cap table, subject to whatever liquidation preferences and pro-rata terms the investors negotiated at each funding round.

Most VC financing docs write their "Deemed Liquidation Event" clause broadly enough to capture an exclusive license of substantially all assets, not just a merger, specifically to close this loophole. How well that protection holds depends on how broadly a given charter defines it.

Structuring the deal this way does two things at once. It gets the acquirer the talent without triggering a change-of-control transaction, the kind of deal that would otherwise have to be reported to the federal government under the Hart-Scott-Rodino Act and cleared by the FTC and DOJ before it could even close.

It also does something that rarely gets said out loud: everyone with a seat at the table (the acquirer, the founders, the investors) gets a reason to want the deal, and the one party that might otherwise object, the government, never gets the chance to weigh in.

It would be easy to tell this as founders quietly out-maneuvering the VCs who'd normally sit ahead of them in a liquidation stack. But the data doesn't really support that read.

The licensing fee in these deals isn't a token gesture to make the paperwork look clean. Inflection's investors landed close to breakeven: Microsoft's $650 million fee gave the earlier $225 million funding round about 1.5x and the later $1.3 billion round about 1.1x, according to reporting on the deal's structure. Character.AI's investors did far better: Google's $2.7 billion fee against the $193 million Character.AI had raised works out to roughly 14x.

Which means the people actually winning here are the two parties who normally split one fixed price through the liquidation preference stack: the founders, who get a title, a platform, and often a fresh equity package at the acquirer, and the VCs, who get a real, sometimes better-than-expected payout via the licensing fee without having to fight for a traditional sale in a market where AI acquisitions were getting harder to close. That's a deal between the two parties with the most leverage in the room, not a founder win over the investors who used to sit ahead of them.

The people without a seat in that negotiation are the ones who feel it.

Employees who don't get the call to follow the founders stay behind at a company that just lost the reason it existed. The technology's been licensed out. The leadership's gone. What's left is a shell with a licensing check on the balance sheet but barely any reason to keep the lights on. Windsurf's employees found out what that looks like in practice: stranded for days with no acquirer and no clear path forward until Cognition bought what was left of the company.

Minority and common shareholders are in a similar position, one level up. The preferred stockholders, meaning the VCs, got structured into the deal by design. Common holders, including a lot of early employees whose equity sits below the preference stack, get whatever's left after the fee is paid out and the people who made the company valuable have already gone.

This is the actual shape of the deal, and it's a familiar one dressed up in new mechanics: risk gets distributed to the parties with the least power to negotiate it away, while the parties with leverage structure their way around that risk entirely.

Last week I wrote about regulation as the one forcing function AI hasn't figured out how to route around: regulated industries are outperforming on real invention precisely because they can't shortcut the hard parts.

The reverse acquihire is what happens when that forcing function has a blind spot. The FTC has reportedly started examining the pattern, sending Amazon a list of questions about its Adept hiring and separately scrutinizing Microsoft's Inflection deal, but the deals kept closing while that scrutiny caught up, because the entire structure was engineered around a specific legal definition of what counts as a merger. Lawyers built it that way on purpose.

When a regulatory framework defines its trigger narrowly enough, sophisticated actors will build precisely to the edge of it. The FTC will likely close this particular gap eventually. The more interesting question is how many billions move through it before that happens, and who's protected by the current terms while it stays open.

If you're a founder or an early employee reading this and wondering what it means for you, don't take the lesson as "try to get acquihired instead of acquired," since most companies never get the call. The real lesson sits upstream of that: know exactly where you sit in your cap table's liquidation waterfall before your company becomes the next one getting the call.

Who gets paid if the company is hired away instead of bought? What triggers protect common holders and non-founder employees in a licensing deal versus a traditional sale? Does your equity agreement even contemplate this structure, or does it assume the only two outcomes are IPO and acquisition? Two years ago, most cap tables didn't have language for this scenario because the scenario barely existed. It exists now, at a $5.8 billion-and-growing scale, and the startups negotiating their next round of funding terms would be smart to ask their lawyers about it before they need the answer.

The reverse acquihire gets described as an exit, and in the sense that money changes hands and people move on, it is one. But it's an exit built by and for the two parties who already had the most leverage in the room, engineered specifically to sit outside the review process that would normally scrutinize a deal of this size.

$5.8 billion has moved through this structure since 2024. Zero companies were acquired. Everyone still standing on the cap table when the founders' offer letters get signed should probably know which side of that sentence they're on.