FOLLOW US

SAN FRANCISCO — Poolside is at the center of a major shift in the artificial intelligence industry after Nvidia agreed to invest $1 billion in the startup. And pay another $6 billion to license its technology.

They will also bring more than 100 Poolside employees into its own AI development efforts. The deal could significantly influence how advanced AI models are developed, particularly as Nvidia seeks to strengthen the U.S. position in the growing competition over open-weight models.

Poolside Agreement is not Traditional

The agreement comes as Nvidia increasingly looks beyond its traditional role as the dominant supplier of graphics processing units, or GPUs, for AI systems. The company is pursuing a larger role in developing the models that run on those chips, including open-weight systems that can be downloaded, modified and deployed by developers.

Under the arrangement, Nvidia will license Poolside's technology and hire roughly 109 employees, according to reporting on the company's investor letter.

Poolside's founders, including Eiso Kant and Jason Warner, are expected to remain outside Nvidia and continue working on unspecified research projects. The arrangement is not structured as a traditional acquisition.

Nvidia Targets the Open-Weight AI Market

Nvidia plans to use Poolside's technology and engineering talent to strengthen its Nemotron family of open-weight AI models.

Open-weight models differ from traditional proprietary AI systems because their trained numerical parameters, or weights, can be made available for developers to download and modify. That can make them less expensive to deploy and easier to customize for specific businesses, researchers and governments.

The strategy also puts Nvidia in competition with some of its largest customers, including OpenAI and Anthropic, which have invested heavily in proprietary AI models.

At the same time, Nvidia is responding to growing competition from China.

Models such as DeepSeek and other Chinese open-weight systems have intensified pressure on American companies to develop models that are both powerful and broadly accessible.

Poolside to Join the Nemotron Coalition

Nvidia launched its Nemotron Coalition earlier this year with several AI companies, including Mistral, Thinking Machines Lab and Perplexity, in an effort to share expertise, data and computing resources.

Nvidia CEO Jensen Huang has also publicly argued that open-weight AI will be important to maintaining American leadership.

Poolside's latest model, Laguna S 2.1, illustrates the company's progress. Released in July, the model has about 118 billion parameters and was designed for agentic coding tasks.

What Is an AI Cluster?

One of the most revealing portions of Poolside's investor letter involves something that is becoming increasingly important to the AI industry: the computing cluster.

In simple terms, an AI cluster is a large collection of interconnected servers and GPUs that operate together as a single computing system. Instead of one computer processing a massive AI training job, thousands or tens of thousands of GPUs can divide the work among themselves.

For frontier AI models, the scale can become enormous. Training requires enormous amounts of data to be repeatedly processed through neural networks.

The GPUs perform the mathematical calculations, while high-speed networking allows the individual machines to communicate and exchange information.

Eric Swalwell Faces FBI Search as Sexual Misconduct Investigation Expands - USA Herald  

Poolside had 2025 Funding Issues 

Poolside said it had a six-week period at the end of 2025 to raise $2 billion to pay for a 40,000-GPU GB300 cluster that was scheduled to come online in January. The company failed to close the financing in time and lost access to the cluster.

The episode illustrates how quickly the economics of AI development are changing. Access to talented researchers is no longer enough. Companies also need enormous amounts of computing power, electricity, cooling capacity, networking equipment and physical data-center space.

Poolside Project Horizon on Hold

The shortage of computing infrastructure became a particularly serious problem for Poolside.

The company had been pursuing Project Horizon, an ambitious data-center project in West Texas that was originally planned as a 2-gigawatt AI campus.

The project was envisioned on hundreds of acres and was designed around access to large amounts of electricity and computing capacity. CoreWeave was expected to serve as an anchor tenant and provide access to more than 40,000 Nvidia GPUs.

But building an AI data center is considerably more complicated than constructing an ordinary office or server facility.

The Data-Center Problem

AI data centers require extraordinary amounts of electricity. They also require specialized cooling systems because thousands of high-performance GPUs generate substantial heat.

Power transmission, substations, fiber-optic connectivity, land, construction schedules and equipment availability can all become bottlenecks.

Poolside's experience demonstrates that capital and physical infrastructure are closely linked. Even if an AI company can raise billions of dollars, it still needs somewhere to put the computers and enough electricity to operate them.

Reporting in April indicated that Poolside was seeking new partners for its Texas data-center ambitions after its previous arrangement with CoreWeave fell apart.

The proposed facility had been designed around an initial 250-megawatt phase with the potential to expand.

With Poolside Nvidia Gains Talent and Technology Scale

The Nvidia-Poolside arrangement could therefore solve several problems simultaneously.

Nvidia obtains access to Poolside's model-development technology and a large group of engineers who have been working on open-weight AI. Poolside gains a partner with substantially greater access to capital, computing infrastructure and Nvidia's own hardware ecosystem.

For Nvidia, the transaction also provides a way to accelerate Nemotron without starting entirely from scratch.

The company's challenge is that AI development is becoming a competition not only over algorithms but also over compute, energy and infrastructure.

The most advanced models require increasingly large training clusters, while AI companies compete for limited supplies of GPUs and data-center capacity.

Poolside's failed attempt to secure the 40,000-GPU cluster provides a glimpse into that problem.

The company said that a cluster containing 10,000 to 20,000 GB300 GPUs could potentially produce a model competitive with the then-current frontier, but that future systems would require dramatically larger clusters.

The constraint, Poolside said, is not simply money but also physical data-center space and contracted computing capacity.

That reality could make Nvidia's position even more influential. The company supplies much of the hardware needed to build AI systems while increasingly developing the software and models that run on that hardware.

A New Phase of the AI Race

The Poolside transaction highlights a broader transformation in artificial intelligence.

The first phase of the AI boom was dominated by access to data, algorithms and researchers. The next phase increasingly depends on industrial-scale infrastructure.

Companies must secure chips, power, cooling, networking and data centers while simultaneously developing models capable of competing with rapidly improving systems from both the United States and China.

Nvidia's decision to put billions of dollars behind Poolside's technology signals that open-weight AI is becoming a major strategic battleground.

The company is betting that the future of AI will not be controlled exclusively by a handful of closed-model providers.

Instead, Nvidia appears to be positioning itself to supply the chips, infrastructure, engineering talent and open-weight models that developers around the world can build upon.

For Poolside, the agreement marks a dramatic change from its attempt to independently finance one of the world's largest AI computing operations.

For Nvidia, it represents another step toward becoming, not simply, the company powering the AI revolution, but one of the companies helping determine what that revolution produces.

Related Resources

• Nvidia's official website — Information on Nvidia's AI platforms, GPUs and software.
• Poolside's official website — Information about Poolside's AI models and technology.
• CoreWeave's official website — AI cloud infrastructure and GPU computing information.
• Laguna S 2.1 on Hugging Face — Technical information about Poolside's latest open-weight model.
• CoreWeave and Poolside data-center announcement — Details on the planned 40,000-plus-GPU infrastructure.