Nscale, a neocloud operator, has signed a multi-year compute agreement with humanoid robotics company Figure. The deal matters because it links two fast-moving AI infrastructure themes: large-scale GPU deployment for model training and the emerging market for physical AI, or robots that operate in human environments. Under the agreement, Nscale will deploy up to 100,000 Nvidia Vera Rubin GPUs for Figure at its data center in Barstow, Texas. The initial contract value is $3.5 billion, with the stated intent to scale to more than $6 billion over time. Nscale will also become a shareholder in Figure through a strategic investment and will serve as Figure's preferred compute provider.
The material mechanics center on a single site. Nscale's Barstow data center is leased from Bitcoin miner Ionic Digital, located at 3013 FM 516, and has 234MW of capacity. The facility is already being used to deploy AI infrastructure for Microsoft, according to the source. Figure will use Nscale's compute to train its AI model, Helix. The agreement is described as multi-year, but the source does not specify the contract duration, payment schedule, or the size of Nscale's strategic investment in Figure. Those details remain unknown based on the available reporting.
The evidence comes from a single full-text article published by Data Center Dynamics on September 4, 2026. The report includes statements from Josh Payne, CEO and founder of Nscale, and Jensen Huang, founder and CEO of Nvidia. Payne described physical intelligence as AI's next frontier and said Nscale is proud to be enabling the future of AI robotics. Huang framed the partnership as activating a robotics flywheel: training Figure's models on Nvidia Vera Rubin through Nscale's AI cloud, validating them in Nvidia Isaac Sim, and deploying them on Nvidia GPUs in Figure's robots. The source also reports that Nscale is nearing an initial public offering in the US, from which it is reportedly aiming to raise $3 billion. No additional sources were available to corroborate the contract value, the GPU count, or the IPO target.
The sector implications are significant but still early. For data center operators, the deal shows how neoclouds are competing for large, concentrated AI training workloads rather than only serving inference or general cloud demand. For robotics, the agreement ties Figure's model development to a specific compute provider and to Nvidia's hardware and simulation stack. The reported scale of up to 100,000 Vera Rubin GPUs also suggests that training humanoid robot models is becoming a major compute-intensive category, comparable in ambition to large language model training. However, the source does not provide Figure's current valuation, Nscale's ownership stake, or the timeline for deployment.
Several limitations should be noted. The dossier contains one source read in full, so the analysis cannot independently verify the contract terms, the strategic investment, or the reported IPO raise. The phrase "intent to scale to more than $6 billion" is forward-looking and may not reflect committed spending. The Barstow facility's 234MW capacity is stated, but the source does not say how much of that capacity will be dedicated to Figure versus Microsoft or other customers. What to watch includes any regulatory or exchange filings from Nscale ahead of its reported US IPO, further disclosure from Figure about Helix training milestones, and confirmation of whether the full 100,000-GPU deployment is actually executed.