Nvidia's reported $1 billion Figure bet tests AI supply loop

Nvidia is in advanced talks to invest an additional $1 billion in Figure AI at a $39 billion valuation, TechTimes reported on Oct. 13, citing an Oct. 7 report by The Information.

The reported valuation would place Figure AI above Goldman Sachs’ $38 billion forecast for the global humanoid robot market in 2035, TechTimes said, framing the scale of Nvidia’s potential bet.

Neither Nvidia nor Figure AI has confirmed the reported transaction, according to the TechTimes article. The reported follow-on would come after Nvidia joined Figure AI’s Series C round in September 2025, when the robotics company raised more than $1 billion at a $39 billion post-money valuation, meaning the value after new funding.

Nvidia Chief Executive Jensen Huang cast the relationship as an integrated robotics system at GTC 2026. “Nscale and Figure have activated the robotics flywheel,” Huang said, referring to Figure’s model training on Nvidia’s Vera Rubin graphics-processing units, Nvidia Isaac Sim simulation software and Nvidia chips inside robots.

The structure is drawing attention because Nvidia already has exposure to another part of the same chain. TechTimes said Nvidia invested about $500 million in Nscale, a U.K.-based GPU cloud provider, while Nscale agreed on Sept. 3 to provide Figure AI with at least $3.5 billion of compute capacity, potentially rising above $6 billion.

Under that arrangement, Figure AI would use Nvidia’s Vera Rubin GPU platform through Nscale’s cloud infrastructure, initially at a new data center in Barstow, Texas. GPU deployments are targeted for the second half of 2027, according to the article.

That creates a structure in which Nvidia supplies chips to Nscale, owns equity in Nscale, may add direct equity in Figure AI and benefits when Figure buys compute capacity built on Nvidia hardware. TechTimes said critics, including short seller Jim Chanos, call Nvidia’s broader structure circular financing; the article also said Nvidia disputes that characterization.

Figure AI’s strongest operating evidence cited by TechTimes comes from its BMW work. The company’s Figure 02 humanoids ran 10-hour weekday shifts at BMW’s Spartanburg, South Carolina, plant over 11 months, loading more than 90,000 parts across more than 1,250 runtime hours and contributing to more than 30,000 BMW X3 vehicles, according to Figure AI material cited by TechTimes.

The company also uses a Robot-as-a-Service model, a subscription structure for robot deployment, software, maintenance and support. TechTimes cited pricing of about $1,000 per robot per month, which would turn a large upfront equipment purchase into a recurring operating cost for customers.

The commercial risk is that technical progress does not yet equal broad reliability. TechTimes said Figure’s Helix 2.5 system recorded a 56% task-completion rate across 30 previously unseen homes in September 2026, a zero-shot result, meaning the system attempted unfamiliar environments without location-specific programming.

A 56% completion rate supports the claim that the system can generalize, but it also shows failures remain common. TechTimes reported no Figure AI revenue, profit or customer-retention figures, leaving investors to weigh production milestones against an aggressive private valuation.

What to watch

In the second half of 2027, the checkable milestone is whether Nscale begins the targeted Vera Rubin GPU deployments for Figure AI at the Barstow, Texas, data center described in the Sept. 3 compute agreement.

The same second-half 2027 window will test whether Figure AI starts using Nscale’s Nvidia-based cloud capacity at the scale implied by the $3.5 billion minimum commitment, with any move toward more than $6 billion dependent on the contract terms reported by TechTimes.

Source: trade-press article, TechTimes, Oct. 13, 2026