Figure targets initial deployment in Barstow, Texas, in the second half of 2027; Nscale will take an undisclosed stake and serve as Figure's preferred compute provider.
By RuntimeWire Staff · Published
Primary source: Figure Newsroom
Why it matters #
Adcock is treating humanoid intelligence like frontier AI: collect proprietary data, reserve massive compute, and own the robot that deploys the model. The test now moves to power, delivery and whether Helix improves enough to justify a multibillion-dollar bill.
Brett Adcock (@brettadcock) has committed Figure to an initial $3.5 billion of compute from Nscale, the AI infrastructure provider, giving his humanoid robotics operation a path to as many as 100,000 NVIDIA Vera Rubin GPUs as he tries to scale its Helix AI models.
The companies said in their September 3 announcement that the first systems are targeted for deployment in Barstow, Texas, during the second half of 2027. Figure and Nscale intend to expand the arrangement beyond $6 billion, while Nscale will make an investment of an undisclosed size in Figure.
Adcock has spent his career making capital-intensive bets progressively larger. He started the recruiting marketplace Vettery at 26 and sold it to Adecco for $110 million in 2018, according to his biography. He then co-founded electric-aircraft developer Archer Aviation before starting Figure in 2022. The Nscale agreement gives his thesis that general-purpose humanoids could address labor shortages and eventually work in homes an infrastructure budget closer to a frontier AI lab than a conventional robotics manufacturer.
Adcock is scaling Helix like a frontier model
Figure says its immediate constraint is the data and compute required to train Helix, the vision-language-action system that controls Figure's robots. The agreement would connect three layers: training Helix through Nscale's cloud, validating the models in NVIDIA Isaac Sim, and deploying the resulting software on NVIDIA hardware inside Figure robots.
NVIDIA's Vera Rubin platform combines Rubin GPUs with Vera CPUs, networking, storage processing and rack-scale systems. NVIDIA says a Vera Rubin NVL72 rack contains 72 Rubin GPUs and 36 Vera CPUs. The Figure and Nscale releases do not say how many GPUs are included in the initial phase or whether every system would use that rack configuration.
The compute deal follows Figure's August 25 launch of Index, a crowdsourced pipeline for collecting videos of people performing physical tasks. Figure reported 264,000 downloads, more than 44,000 weekly active contributors and 16 million uploaded videos. It also said it had paid contributors $15 million and planned to spend more than $1 billion on data and compute over the following 12 months.
Figure's two recent posts give slightly different throughput figures. The Index announcement reported 30 minutes of uploaded video each second; the Nscale announcement nine days later used 35 minutes. Both numbers are self-reported. The direction is still clear: Adcock is trying to build a proprietary data supply large enough to justify a dedicated compute buildout.
That approach extends Figure's vertical integration from robot hardware and manufacturing into data and model training. Figure designed Figure 03 for high-volume production at its BotQ facility, while Helix supplies the learned behavior. Index collects the human demonstrations. Nscale would provide the training infrastructure. Adcock is assembling each scarce input under a structure Figure can control or contract directly.
The 2027 delivery is the real milestone
The headline GPU count is a ceiling, and the installation remains at least several quarters away. Figure calls the $3.5 billion figure an initial compute commitment. Nscale's version of the agreement describes a multi-year partnership and names Nscale as Figure's preferred compute provider. Neither announcement defines payment schedules, minimum capacity, cancellation terms or the portion of the 100,000 GPUs covered by the initial commitment.
Nscale's broader Texas capacity build provides context for the delivery challenge, although its public financing release does not identify the financed Ward County campus as Figure's specific Barstow deployment site. In an August 31 financing announcement, Nscale said it had closed approximately $3 billion in aggregate commitments across delayed draw term loan facilities for campuses in Ward County, Texas, and Madison, North Carolina. It allocated up to $1.85 billion to the Ward County facility.
An Ionic Digital supplemental prospectus says Nscale leases 234 megawatts at Ward County and is contractually obligated to lease another 89 megawatts if that capacity becomes available. Ionic said it expected the additional capacity to be energized in the second half of 2027, subject to regulatory approval. The filing also says Nscale granted Microsoft an option to lease additional power at the property, if available, beginning in late 2027.
Those timelines overlap Figure's initial deployment target, but the documents do not tie the Ward County capacity to Figure or allocate the power among Nscale customers. The eventual size of Figure's first installation therefore remains undefined. Power approvals, construction and hardware delivery will determine the starting point.
The investment adds another layer to the relationship. Nscale will become a Figure shareholder while selling Figure billions of dollars of compute. That alignment gives Nscale a direct interest in Helix's progress and gives Figure a provider motivated to deliver the capacity. The announcements do not specify the investment amount, valuation, ownership percentage or closing date. They also do not describe Nscale's preferred-provider status as exclusive.
Nscale founder Josh Payne has built Nscale around control of power, data centers, GPU supply and cloud software. Nscale raised $2 billion at a $14.6 billion valuation in March, led by Aker and 8090 Industries. The companies also plan to explore using Figure humanoids in Nscale's supply chain, without committing to a deployment volume or date.
From factory work to a robot in every home
Figure currently has a more concrete proving ground in manufacturing. In June, Figure said Figure 03 had arrived at BMW's Spartanburg plant for a logistics workflow requiring the robot to manipulate parts while repositioning its body and pulling a cart. RuntimeWire reported in May that Adcock was staking Figure's progress on intelligence and preparing additional customer deployments.
Adcock's larger objective is household deployment. Figure's website presents Figure 03 as a robot for laundry, cleaning and dishes, environments with far more variation than a structured factory station. That shift explains the scale of Index and the Nscale contract: every new room, object and human habit creates another training problem for Helix.
Figure said in September 2025 that it had secured more than $1 billion in Series C commitments at a $39 billion post-money valuation, led by Parkway Venture Capital. NVIDIA, Brookfield, Macquarie Capital, Intel Capital and several corporate venture funds participated. Figure said at the time that the money would support Helix, manufacturing and GPU infrastructure.
The $3.5 billion compute commitment exceeds that round's disclosed size, though the two figures represent different financial arrangements. Figure has not published the duration, financing structure or expected annual spend behind the compute agreement. Those terms will determine whether it functions as a capacity reservation, a firm multiyear purchase or a framework that grows alongside Helix.
Adcock has put a measurable infrastructure plan behind Figure's household ambition. By late 2027, Figure will need Nscale to deliver the power and Rubin systems, Index to keep supplying useful training data, and Helix to turn both inputs into better robot behavior. The GPU count will matter when the robots show what Figure learned from it.