# Figure AI Secures $3.5B Compute Partnership with Nscale to Power Helix Humanoid Brain

Source: TechNewsList (https://technewslist.com)
Canonical URL: https://technewslist.com/en/article/figure-ai-partners-nscale-multibillion-compute-deal-helix-2026-09-10-night
Section: Drones & Robots (https://technewslist.com/en/drones-robotics)
Author: TechNewsList
Language: en
Published: 2026-09-10T20:00:35.524+00:00
Updated: 2026-09-10T20:00:35.681601+00:00

> Humanoid robotics pioneer Figure AI partnered with Nscale in a $3.5 billion infrastructure agreement, securing 100,000 next-generation NVIDIA GPUs to train its Helix physical AI model.

## TL;DR
- Figure AI signed a multi-billion-dollar compute agreement with AI cloud provider Nscale to power physical foundation models.
- The initial $3.5 billion contract scales up to $6 billion, providing dedicated access to 100,000 next-generation NVIDIA GPUs.
- Compute resources will train Helix, Figure's proprietary multimodal brain enabling general-purpose humanoid dexterity and locomotion.
- As part of the strategic agreement, Nscale acquired a significant minority equity stake in Figure AI.

## Key points
- Figure AI and Nscale formally announced their strategic compute alliance on September 8, 2026.
- The dedicated cluster will be deployed at Nscale's high-efficiency liquid-cooled data center campus in Barstow, Texas.
- The computing cluster will ingest petabytes of multimodal sensor data, teleoperation demonstrations, and physics simulations.
- Figure founder Brett Adcock stated physical AI models require compute clusters comparable in scale to frontier language laboratories.
- Nscale's equity participation aligns long-term cloud infrastructure incentives with Figure's commercial humanoid deployment milestones.
- Industry analysts view the deal as the largest dedicated compute procurement in the history of commercial humanoid robotics.

## What happened

Humanoid robotics developer Figure AI announced a historic multi-billion-dollar infrastructure partnership with AI cloud hyperscaler Nscale on September 8, 2026, securing dedicated supercomputing capacity to train the next generation of embodied physical intelligence. The initial agreement commits $3.5 billion in compute procurement, with scalable expansion options extending beyond $6 billion over the next several years.

Under the terms of the agreement, Nscale will supply Figure with dedicated access to up to 100,000 next-generation NVIDIA graphics processing units hosted within its high-density data center facilities. The computing cluster will be hosted at Nscale's liquid-cooled data center campus in Barstow, Texas, which operates on dedicated renewable energy feeds designed for high-thermal-density artificial intelligence training workloads.

![Nscale high-density data center GPU cluster infrastructure earmarked for Helix humanoid model pretraining.](https://rkhynbcsbnkkcwgexzwg.supabase.co/storage/v1/object/public/media/api/1789070425881-lf3y4e-figure-ai-partners-nscale-multibillion-compute-deal-helix-2026-09-10-night-inside-1-65666468d3.webp)

As part of the commercial arrangement, Nscale has also acquired a strategic minority equity stake in Figure AI. The collaboration directly supports the rapid development and scaling of Helix, Figure's proprietary multimodal physical foundation model designed to serve as the unified cognitive and motor cortex for its Figure 02 and future humanoid robot iterations.

## Why it matters

The partnership underscores a fundamental shift in how the robotics industry approaches general-purpose humanoid autonomy. Historically, robotics companies developed isolated control systems for specific kinematic functions—such as balance, walking, or object grasping—relying on classical trajectory planning and localized reinforcement learning routines.

Figure AI is pioneering an alternative paradigm that treats physical embodiment as an end-to-end foundation model problem. By training large-scale vision-language-action models on massive datasets spanning video demonstrations, high-fidelity physics simulations, and continuous teleoperation feeds, Figure aims to achieve emergent physical reasoning, enabling robots to understand natural language instructions, adapt to dynamic factory environments, and execute complex manual tasks without custom programming.

However, training embodied foundation models requires massive amounts of compute that rival the infrastructure budgets of leading frontier language model laboratories. Securing 100,000 GPUs ensures that Figure has the compute capacity needed to train models with hundreds of billions of parameters, positioning the company as the primary challenger to Tesla's Optimus program and Boston Dynamics in the race to commercialize humanoid labor.

## Technical details

The supercomputing cluster allocated to Figure will run on custom high-bandwidth networks, pairing NVIDIA Blackwell and future-generation architectures with 800Gbps InfiniBand fabrics. This low-latency interconnect is essential for distributed training across thousands of compute nodes, where gradient synchronization across multimodal transformer layers creates substantial network traffic.

Figure's Helix model processes multimodal sensory inputs simultaneously, including synchronized stereoscopic camera streams, joint torque sensors, tactile finger arrays, and spatial audio. The training pipeline ingests petabytes of data gathered from fleet deployments across BMW manufacturing facilities alongside hundreds of millions of simulated task cycles executed in parallel through GPU-accelerated simulation platforms.

![Executive partnership signing between Nscale leadership and Figure AI celebrating the $3.5 billion physical AI compute pact.](https://rkhynbcsbnkkcwgexzwg.supabase.co/storage/v1/object/public/media/api/1789070429423-ua22q3-figure-ai-partners-nscale-multibillion-compute-deal-helix-2026-09-10-night-inside-2-f82bf512af.webp)

Nscale's Barstow campus utilizes direct-to-chip liquid cooling architectures capable of dissipating up to 100 kilowatts per rack, maintaining optimal operating temperatures while minimizing power usage effectiveness. This dedicated environmental infrastructure ensures continuous compute availability for extended multi-month model training runs without thermal throttling or hardware degradation.

## Market / industry impact

Figure's multi-billion-dollar compute procurement signals that humanoid robotics has entered a capital-intensive industrialization phase. Commercial humanoid development is no longer primarily an electro-mechanical engineering challenge; it is increasingly defined by access to massive AI supercomputing clusters and specialized simulation pipelines.

The deal validates Nscale's emergence as a major player in the competitive AI cloud hosting market. While global hyperscalers like Microsoft Azure, AWS, and Google Cloud have focused on broad enterprise workloads, specialized cloud providers like Nscale, CoreWeave, and Lambda have captured significant market share by offering purpose-built clusters optimized specifically for frontier AI model training.

The commercial implications for industrial manufacturing and logistics are profound. As Figure accelerates the intelligence and reliability of its robots, automotive manufacturers, warehouse operators, and logistics providers are preparing for pilot commercial deployments. Automating repetitive, ergonomically hazardous assembly tasks with general-purpose humanoid workers could mitigate chronic industrial labor shortages across North America and Europe.

## What to watch next

The immediate focus will center on Figure's upcoming technical demonstrations showcasing Helix model updates. Watch for benchmarks demonstrating improved manipulation dexterity, fine-motor coordination, and autonomous tool use across dynamic, unstructured environments.

In the infrastructure domain, track the staged delivery and commissioning of Nscale's Barstow facility. Reaching the initial 50,000 GPU threshold will serve as a vital milestone verifying that supply chain deliveries of advanced AI silicon and liquid cooling hardware remain on schedule.

Finally, monitor competitive responses from Tesla, Boston Dynamics, and Sanctuary AI. As Figure secures unprecedented compute reserves, rival humanoid robotics ventures may be pressured to announce similar mega-scale infrastructure agreements to maintain pace in physical AI foundation model training.

## Sources

* PR Newswire: [Nscale and Figure sign strategic partnership to power next generation of physical AI](https://www.prnewswire.com/news-releases/nscale-and-figure-sign-strategic-partnership-to-power-the-next-generation-of-physical-ai-302868918.html)
* The Next Web: [Nscale Figure 3.5B compute partnership equity stake](https://thenextweb.com/news/nscale-figure-3-5bn-compute-equity-stake)
* Humanoids Daily: [Figure inks multi-billion compute deal with Nscale for 100,000 GPUs](https://www.humanoidsdaily.com/news/figure-inks-multi-billion-dollar-compute-deal-with-nscale-to-deploy-100-000-next-gen-nvidia-gpus)

Mentions: Figure AI, Nscale, Brett Adcock, Josh Payne

## Sources
- [PR Newswire](https://www.prnewswire.com/news-releases/nscale-and-figure-sign-strategic-partnership-to-power-the-next-generation-of-physical-ai-302868918.html)
- [The Next Web](https://thenextweb.com/news/nscale-figure-3-5bn-compute-equity-stake)
- [Humanoids Daily](https://www.humanoidsdaily.com/news/figure-inks-multi-billion-dollar-compute-deal-with-nscale-to-deploy-100-000-next-gen-nvidia-gpus)