Nvidia's $6bn Poolside licence is the third run of the same play

Nvidia has agreed to pay Poolside $6bn to license the software the startup used to build its AI models, to offer jobs to 109 of its staff, and to invest a further $1bn in what remains at a $12bn pre-money valuation ^[1]. The structure matters more than the headline number: this is the third time Nvidia has bought a non-exclusive licence, hired the staff, taken a stake, and left the target standing as an independent company ^[2].

The terms come from a letter Poolside sent its investors, first reported by Newcomer ^[3]. The letter says the deal is "not an acquisition and it is not an acquihire" ^[4], and the mechanics are consistent with that: the licence covers a system Poolside calls the Model Factory ^[5], it is non-exclusive, leaving Poolside free to license the same software elsewhere ^[6], the three founders stay ^[7], and Poolside intends to distribute the $6bn to its investors by the end of next year ^[8]. Nvidia was already an investor, having committed up to $1bn to the company last October ^[15].

The precedent is Groq. Nvidia paid $20bn for its inference technology and took its top engineers, Groq appointed new leadership, and the company raised $650m for what was left ^[16][17]. This month Groq closed a $350m round at $3.5bn, with Nvidia participating ^[18]. On those two numbers, the technology-and-people package was priced at roughly 5.7 times the continuing business ^[27]. Nvidia struck a similar arrangement with the hardware startup Enfabrica for about $900m, according to The Decoder, which notes that such deals secure staff and know-how without buying a company outright and avoid the regulatory review a purchase can bring ^[19][20].

The head count is where the Poolside version gets specific. Chief executive Eiso Kant said on the Latent Space podcast last month that fewer than 70 people built the company's model, and fewer than 115 worked across engineering and research in total ^[21]. Offers to 109 staff therefore reach about 95 percent of that stated technical organisation ^[26]. Latent Space called it a reverse-execuhire ^[22]. The licence alone works out at roughly $55m per person hired ^[28].

Poolside's letter explains why the seller was willing. At the end of last year the company had a six-week window to raise $2bn to pay for a 40,000 GB300 cluster coming online in January, did not close it in time, and lost the cluster ^[9]. It argues it could have built a frontier-rivalling model with 10,000 to 20,000 of those chips, that next year's frontier needs "far more than an order of magnitude larger cluster", and that the constraint "is not only capital, it is physical data center space and contracted compute" ^[10].

What is not explained is the price. Nvidia has not said what makes the licence worth $6bn, has not commented publicly, and Poolside has not published the letter ^[11]. The Information's Amir Efrati wrote that it is not clear why Nvidia is paying such a large licensing fee ^[12]. Nvidia builds open models in the Nemotron line and is working towards a trillion-parameter open model ^[23]; Poolside's Laguna, built by the same team now receiving offers, was pitched as the West's answer to DeepSeek and Qwen and trained on Nvidia server chips ^[24]. Nvidia also halved a $250bn commitment to OpenAI this month ^[25].

Three things to watch. Whether a fourth deal of this shape appears, since the template now has a track record and no merger filing attached to it ^[2][20]. Whether any of the non-exclusive licences is ever exercised elsewhere, which is the only test of whether "non-exclusive" is substantive or structural ^[6]. And what Poolside becomes: the founders say they are "not ready to share the updated vision" ^[13], while Poolside Infrastructure Company, spun out in January, is building a 1.2GW data centre in Texas and appointed a chief executive two months ago ^[14].