Nvidia has found a way to acquire the most valuable part of an AI startup without acquiring the startup. On August 20 the chipmaker agreed to pay Poolside roughly $6 billion to license the software the company uses to build its models, and to invest a further $1 billion in a growth round, as first reported by Newcomer and confirmed by Bloomberg. Poolside keeps its three co-founders, keeps operating on its own, and keeps its name. Nvidia gets the machinery and, it hopes, many of the people who run it.
The licensed asset is Poolside's Model Factory, the pipeline it uses to train its Laguna family of open-weight coding models. The deal is non-exclusive, so Poolside can keep using and licensing the system elsewhere. Alongside the license, Nvidia is making offers to 109 Poolside employees who work on Laguna. The $1 billion investment values the remaining business at around $12 billion pre-money, according to the reporting.
Why not just buy it
The obvious question is why Nvidia structured the deal this way instead of writing a bigger check for the whole company. Part of the answer is regulatory. A string of large AI acquisitions has drawn antitrust scrutiny, and a license-plus-hiring arrangement avoids the formal review that a takeover would trigger. Part of it is focus. Nvidia does not need Poolside's business; it needs Poolside's ability to turn raw compute into working models, because that is the capability it wants to offer its own customers on its own chips.
Seen that way, the deal is less about coding models and more about Nvidia's push to sell more than silicon. If the company that makes the hardware can also hand you the software that builds a model on that hardware, it locks customers in one layer higher up the stack. The Laguna models are a bonus. The Model Factory is the prize.
The pattern is becoming familiar
This structure, a giant licensing fee paired with a mass hire and a minority stake, has quietly become one of the defining moves of the current AI cycle. It lets a large company absorb talent and technology while leaving the startup's investors a clean payout. The reported $6 billion fee is expected to reach Poolside's backers by the end of 2027, which turns what would have been an acquisition into something closer to an early exit for everyone who bet on the company.
For Poolside the arithmetic is generous and a little bittersweet. The founders stay in charge of a business that has just handed its crown jewels to a partner large enough to be a future competitor. For Nvidia it is another way of spending its enormous cash position to make sure the AI boom keeps running through its products. The company has already been watching new chip architectures that could one day challenge its dominance, and deals like this one keep the software gravity pulling toward Nvidia hardware.
What makes the Poolside deal worth watching is not the price, large as it is. It is the shape. If licensing a startup's core technology and hiring its team becomes the standard way big players grow, the traditional acquisition may start to look like the slow, expensive option nobody reaches for first.
Sources
- i. www.bloomberg.com
- ii. www.newcomer.co
- iii. thenextweb.com
- iv. www.pymnts.com
Commentarii · 0