Nvidia has agreed to pay Poolside, an AI coding startup, about $6 billion — and it is not buying the company. That distinction is the whole point.

The deal

The structure has three parts. Nvidia pays roughly $6 billion for a non-exclusive licence to Poolside's model-development software — its "model factory." It separately invests about $1 billion in what remains of Poolside, at a $12 billion valuation. And Nvidia is hiring 109 of Poolside's employees, while the three co-founders — including former GitHub CTO Jason Warner — stay behind to run the independent company.

Add the licence and the investment together and it is roughly a $7 billion commitment. What it is not, on paper, is an acquisition.

Why it's built this way

An acquisition of a $12 billion AI startup by the world's most valuable chipmaker would walk straight into an antitrust review on two continents. A licence plus a minority stake plus a hiring spree does not — because no single company is being bought, no controlling stake changes hands, and each piece, viewed alone, is an ordinary commercial transaction.

The result is that Nvidia gets the technology, the team and the momentum of Poolside, and Poolside's backers — Bain Capital Ventures, eBay, Citi Ventures and others — get paid out. Everyone gets the substance of a sale. Nobody has to file for one.

This is a playbook, not a one-off

It is the third time in a year Nvidia has run this exact move. In September 2025 it spent more than $900 million to hire the CEO of networking startup Enfabrica and license its technology; in December it paid around $20 billion for a non-exclusive licence to Groq's inference chips, with Groq's founder and staff joining Nvidia. Each time, the target stayed technically independent. And the pattern is bigger than Nvidia: the 2024 wave of "reverse acqui-hires" — Microsoft and Inflection, Amazon and Adept, Google and Character.AI — used the same shape to absorb a startup's people and technology without triggering a merger filing.

What began as a workaround is now a standard instrument. When the deal is worth billions and the target is a leading AI lab, the licensing-plus-hiring structure has become the default way to do it.

Why it matters

US lawmakers have started calling these deals what they suspect they are: an end-run around merger review. The concern is straightforward — if the biggest AI players can absorb the most promising startups without any regulator getting a look, then the tools built to keep the market competitive simply stop applying to the part of the economy growing fastest.

There is a real question underneath the legal one. Consolidation in AI is already concentrated in a handful of names. If the mechanism for further concentration is a structure regulators cannot easily touch, the concentration does not slow down — it just stops being visible. Nvidia's $6 billion is not really the story. The template is.