On August 21, Nvidia struck a deal with AI coding startup Poolside: a $6 billion non-exclusive license for its Model Factory software, a $1 billion investment at a $12 billion pre-money valuation, and hiring offers to 109 of its employees. Poolside's three co-founders, including former GitHub CTO Jason Warner, will remain. The company stays independent.
The $6 billion will be distributed to Poolside's existing investors by the end of 2027. The 109 employees transferring to Nvidia represent the core engineering and research team behind the Laguna coding models. Poolside CEO Eiso Kant described the team's size on the Latent Space podcast last month: "Less than 70 people built this model. Less than 115 between engineering and researchers, like, together did this effort." Nvidia is taking most of them.
Three Deals, Two Years, One Playbook
This structure is now a pattern. In December 2025, Nvidia paid approximately $20 billion for a non-exclusive license to Groq's inference technology, with founder Jonathan Ross joining Nvidia while Groq continued independently under a new CEO. Earlier in 2025, Nvidia executed a similar arrangement with Enfabrica for AI networking hardware at roughly $900 million.
Each deal follows the same template: license the technology, hire the team, take a minority stake, leave the company technically independent. The license is non-exclusive in each case—Poolside can still license its Model Factory elsewhere, just as Groq can license its chips to other customers.
But the result is the same each time: Nvidia absorbs the operational capacity of critical AI startups while avoiding the regulatory scrutiny that typically accompanies full-scale corporate integration.
Nvidia Bought the Factory, Not the Model
Poolside's Laguna coding models are solid but not dominant. Laguna M.1 scored 72.5% on SWE-bench Verified—competitive with Qwen-3.5 and DeepSeek V4-Flash, but trailing Claude Sonnet 4.6. The smaller Laguna XS 2.1 variant hit 70.9%, and the long-horizon Laguna S 2.1 reached 70.2% on Terminal-Bench 2.1.
Nvidia is not paying $7 billion for benchmark rankings. It is paying for the industrial capacity to build coding models at scale. The Model Factory is Poolside's development infrastructure—the system that turns data, compute, and research into an open-weight coding model. By securing the factory, Nvidia ensures it remains the primary beneficiary of the Laguna family's development, regardless of which specific model version eventually captures the most market share.
The letter Poolside sent to investors explains the pressure that drove the deal: "At the end of last year, we had a 6 week window in which to raise $2 billion dollars to pay for a 40,000 GB300 cluster coming online in January. We didn't close it in time, and we lost the cluster." The company could have built a frontier-rivalling model with 10,000 to 20,000 of those chips. Next year's frontier, it argues, needs "far more than an order of magnitude larger cluster."
Poolside needed Nvidia's hardware. Nvidia wanted Poolside's software and talent. The deal solves both problems.
The "Not-an-Acquisition" Loophole
By structuring each deal as a licensing agreement and a minority equity investment rather than a buyout, Nvidia avoids the antitrust scrutiny that would accompany absorbing these companies entirely. The founders remain in place, and the companies continue to operate—providing the appearance of market competition while Nvidia's control over the technical talent and proprietary systems deepens.
Bernstein analyst Stacy Rasgon noted that "antitrust is clearly the primary risk," but the structure "preserves the illusion that competition still exists." Lawmakers have already questioned whether licensing agreements in place of direct acquisitions let companies bypass merger review. As Nvidia's deals of this type multiply, the regulatory pressure is likely to grow.
The Poolside deal is Nvidia's third in this template. It will not be the last. For investors, the deal shows Nvidia's willingness to pay up for AI software and talent rather than build in-house. For the AI industry, it signals a new kind of consolidation—one that leaves the company standing but strips it of its most valuable asset.
P.S. The $6 billion licensing fee will be paid out to Poolside's investors by the end of 2027. By then, Nvidia will have hired the people who built the models, licensed the system they used to build them, and locked in a long-term customer for its chips. Poolside will still exist. But its core capability will have moved to Nvidia. That's not an acquisition. It's a migration.
Frequently Asked Questions
Q: What did Nvidia actually buy from Poolside?
A: Nvidia did not acquire Poolside. It bought a non-exclusive license to Poolside's Model Factory software (the system used to build coding models), invested $1 billion at a $12 billion pre-money valuation, and hired 109 of Poolside's core employees. Poolside remains an independent company.
Q: How much did Nvidia pay?
A: Nvidia paid $6 billion for the license and invested an additional $1 billion. The license fee will be distributed to Poolside's investors by the end of 2027.
Q: Which Poolside employees are moving to Nvidia?
A: The 109 employees transferring to Nvidia represent the core engineering and research team behind Poolside's Laguna coding models. Poolside's three co-founders—including former GitHub CTO Jason Warner—will remain with Poolside.
Q: What is the Model Factory?
A: Model Factory is Poolside's development infrastructure—the system that turns data, compute, and research into open-weight coding models. Nvidia is licensing this system, not the Laguna models themselves.
Q: Is Nvidia the only company that can use Poolside's technology?
A: No. The license is non-exclusive. Poolside can still license its Model Factory to other companies, just as Groq can license its chips to other customers.
Q: Why is Nvidia structuring deals this way?
A: By using licensing agreements and minority investments instead of outright acquisitions, Nvidia avoids the antitrust scrutiny that would accompany buying these companies entirely. The founders stay, the companies remain independent, but Nvidia gains access to core talent and technology.
Q: What other companies has Nvidia done this with?
A: This is the third such deal. In December 2025, Nvidia paid approximately $20 billion for a non-exclusive license to Groq's inference technology and hired founder Jonathan Ross. Earlier in 2025, Nvidia executed a similar arrangement with Enfabrica for AI networking hardware at roughly $900 million.
Q: Does Poolside still have a future after this deal?
A: Poolside remains independent with its co-founders in place. However, the core team that built its models has moved to Nvidia. The company's long-term trajectory will depend on its ability to rebuild its engineering and research capacity.
Q: What are the regulatory risks?
A: Bernstein analyst Stacy Rasgon noted that "antitrust is clearly the primary risk" and the structure "preserves the illusion that competition still exists." As Nvidia repeats this playbook, regulators are likely to scrutinize whether licensing agreements are being used to bypass merger review. Lawmakers have already raised similar concerns about such deals in other tech sectors.
Q: Why did Poolside agree to this deal?
A: Poolside was under pressure to secure computing resources. The company had lost a planned 40,000-chip cluster after failing to raise $2 billion in a six-week window. Nvidia's deal solves that problem by giving Poolside access to hardware while Nvidia gets the software and talent.
