| Date | 3 Sep 2026 |
|---|---|
| Region | U.S. & Allies |
Figure has published a multi-billion-dollar compute commitment and a 2027 start window for Vera Rubin capacity, while the GPU delivery, the Nscale investment terms, and any humanoid supply-chain work remain prospective.
IN BRIEF: On September 3, 2026, Figure announced a partnership with Nscale to deploy the NVIDIA Vera Rubin platform for up to 100,000 GPUs, citing an initial commitment of $3.5 billion of compute, an intent to scale to over $6 billion, and an initial deployment targeted for the second half of 2027 in Barstow, Texas.
RECORD:
Date: September 3, 2026
Beat: Compute Ledger; Chips & Semiconductors; Frontier Capital
Jurisdiction: United States (Barstow, Texas)
Entity: Figure; Nscale; NVIDIA
Instrument / cite: N/A
Source type: Company announcement (Figure news post)
Confidence: reported
Record ID: N/A
Source: Figure
Figure has announced a partnership with Nscale to deploy the NVIDIA Vera Rubin platform for up to 100,000 GPUs. The September 3, 2026 post on Figure’s news site states an initial commitment of $3.5 billion of compute, with intent to scale to over $6 billion, and an initial deployment targeted for the second half of 2027 in Barstow, Texas. For anyone underwriting physical-AI compute, the post is a published capacity plan. It is not a record of GPUs already installed.
The announcement says the arrangement is an agreement, and that Nscale is also making a strategic investment in Figure. No size, security, valuation, or closing date for that investment appears in the post. The parties will also explore the potential to scale Nscale’s supply chain with humanoids. That clause is an exploration. It is not a stated robot order, a site count, or a date by which humanoids enter Nscale’s supply chain.
Figure locates the constraint on its own model. The company writes that it is entering a phase where it is largely bound by the data and compute needed to train Helix, and that the Nscale partnership provides the compute runway needed to train the next generation of AI models to solve general robotics. Brett Adcock, Founder and CEO of Figure, said: “Our AI model, Helix, becomes more capable the same way every learned system does: with more data and compute. Today we’re excited to partner with Nscale to bring the compute required to make this vision a reality.”
The same post ties that compute need to a dataset Figure says it announced the prior week. Index, described as Figure’s effort to build the most diverse humanoid training dataset ever assembled, is generating 35 minutes of data every second, the company states. Figure then says data alone cannot solve this problem, and that scaling physical intelligence will require an immense amount of compute. The post closes by saying that through this partnership Figure is securing access to reliable compute to bring humanoid robots into every home in the world. The household line is an objective stated by Figure. No unit forecast, price, or shipment timeline accompanies it.
Jensen Huang, founder and CEO of NVIDIA, is quoted in the announcement rather than attached to a separately stated NVIDIA supply commitment. “Humanoid robots extend physical AI into the world designed for people—opening a major new industry,” Huang said. “Nscale and Figure have activated the 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. This is the physical AI flywheel that will accelerate the path from models to robots in the world.” That quote is the only technical sequence the post provides: training on NVIDIA Vera Rubin through Nscale’s AI cloud, validation in NVIDIA Isaac Sim, and deployment on NVIDIA GPUs in Figure’s robots. No power allocation, hall count, or split of the 100,000 GPUs between training and on-robot inference is given.
Josh Payne, CEO and Founder of Nscale, said: “We’re excited to partner with Figure as physical intelligence becomes AI’s next frontier.” He continued: “We’ve seen incredible growth with inference and agentic AI and Figure is pushing the boundaries of AI even further. We’re proud to be enabling the future of AI robotics together.” Payne’s remarks place Nscale on the inference and agentic-AI side of demand and cast Figure as the extension into robotics. They do not allocate the $3.5 billion across training, simulation, and on-robot compute, and they do not say what share of the “up to 100,000” figure is firm.
Read as an instrument, the post is a company announcement of a partnership, a dollar commitment, a scale intent, a named platform, a named site, and a start window. It is not a regulatory filing, a reproduced purchase order, or a construction notice. The deployment language is “targeted for” the second half of 2027. The larger figure is an “intent to scale to over $6 billion.” Screening of frontier compute commitments turns on those verbs: targeted, intent, explore, up to. The announcement supports a judgment that Figure is publicly tying Helix and Index to a multi-billion-dollar Vera Rubin plan with Nscale. It does not support a judgment that the GPUs are contracted in full, financed, or scheduled beyond the window Figure has published.
WHY IT MATTERS: Figure has put a $3.5 billion initial compute commitment, a scale intent above $6 billion, and a Barstow, Texas start window on the record against Helix and the Index data rate it cites. Investors and enterprise buyers can treat the partnership as announced. They cannot yet treat the GPUs, the undisclosed Nscale investment, or the humanoid supply-chain exploration as executed.