Three Prices for Three Different Things
Nvidia is not writing Poolside one check. It is writing three: six billion dollars to license Poolside's Model Factory technology on a non-exclusive basis, one billion dollars to buy equity in Poolside itself at a twelve billion dollar pre-money valuation, and individual salaries for the 109 Poolside employees who received direct job offers from Nvidia. Newcomer first reported the numbers on August 20, 2026, citing a letter Poolside sent its investors, and The Information corroborated the licensing-and-hiring structure independently the same day.
The licensing fee covers Poolside's Laguna family of coding models, the software Poolside has spent 2026 building specifically to write, debug and optimize code inside a customer's own infrastructure. Because the license is non-exclusive, Poolside is technically still free to sell that same technology to Nvidia's competitors, a detail that matters for how regulators will eventually classify the deal.
The 109 Number Is the Part That Looks Like an Acquisition
One hundred and nine Poolside staff received individual job offers from Nvidia as part of the agreement, a headcount large enough to hollow out a startup that reportedly employs only a few hundred people. Commentators who followed the Newcomer story described the structure as a 'reverse acquihire': instead of buying a company and inheriting its employees, Nvidia is hiring the employees directly and licensing the company's output separately.
Poolside's founders pushed back on that framing in their own investor letter, writing that 'this is not an acquisition and it is not an acquihire.' The distinction matters commercially as much as semantically: Poolside keeps its brand, its remaining team and its right to sell to other customers, while Nvidia gets the people and the technology it specifically wanted without listing Poolside as an acquired subsidiary on any regulatory filing.
Why This Shape and Not a Straight Purchase
A licensing fee plus a minority equity stake plus a wave of individual job offers achieves most of what an acquisition would, without triggering the merger review an acquisition would face. Nvidia does not take a majority stake, does not appear as the acquirer in a change-of-control filing, and does not need antitrust clearance to hire 109 people one contract at a time, even though the practical result concentrates a large share of one startup's engineering talent inside Nvidia.
The template is not new to 2026, but this is its highest-value use yet. Microsoft ran a similar play in 2024, when it licensed technology from Inflection AI and hired the bulk of its staff, including co-founder Mustafa Suleyman, while leaving a much smaller Inflection behind. Nvidia had already invested in Poolside once, leading part of a 500 million dollar round in October 2024, so this deal builds on an existing relationship rather than starting one.
What Twelve Billion Dollars Is Actually Buying
The equity half of the deal values Poolside, post-hiring-wave, at twelve billion dollars pre-money, a valuation that has to be earned by whatever remains of the company once 109 of its people leave for Nvidia payrolls. Existing Poolside investors are set to receive 76.20 dollars per share by the end of 2027 under the terms reported, giving early backers a defined exit path regardless of how the smaller Poolside performs afterward.
That leaves Poolside's founders, Jason Warner and Eiso Kant, running a company that is smaller in headcount but larger on paper, still selling a non-exclusive license to the same technology Nvidia just bought a slice of. Whether a twelve billion dollar valuation holds up depends on whether the remaining team can keep building without the 109 people Nvidia just hired away.
The Negotiating Template Every EU AI Startup Should Read Twice
European antitrust regimes are built to catch share purchases and full mergers, not licensing agreements paired with a minority stake and a stack of individual job offers, so a deal shaped like Poolside's routinely clears the merger-notification thresholds that a straight acquisition would trip. Any EU AI startup being courted by a hyperscaler or chipmaker as a 'strategic partner' should recognize this shape before signing anything: a licensing term sheet with an equity kicker is not automatically safer for the target company than a takeover offer, even though it looks that way on the cap table.
The real exposure sits with the team, not the balance sheet. Once a chipmaker or hyperscaler is paying enough to license your technology and invest in your equity, it already knows exactly which engineers it wants, and it can offer them a salary and a name-brand employer directly, without needing your board's approval or your retention plan to hold. Founders negotiating this kind of deal should price retention terms, non-solicit language and board control as hard as they price the headline number, because the headline number is the easy part to agree on.
Read next: Nvidia Hired the Founder, Then Groq's Value Was Halved | Volta Infra: $10bn Anthropic Deal at 7 Months
