Figure's $3.5B Nscale deal: capacity lands in 2027
Figure committed $3.5B to Nscale for up to 100,000 NVIDIA Vera Rubin GPUs. First deployment is second half of 2027. Read the delivery date, not the headline.
Figure and Nscale signed a compute partnership for up to 100,000 GPUs on the NVIDIA Vera Rubin platform, per Figure's own announcement. Initial commitment is $3.5 billion, with stated intent to scale past $6 billion. Initial deployment starts in the second half of 2027, in Barstow, Texas. That last number is the one that should change how you plan.
What actually happened
The compute is earmarked to train Helix, Figure's proprietary foundation model for humanoid robotics. Nscale is also making a strategic investment in Figure — the press release does not disclose the stake — and the two say they will explore using humanoids in Nscale's own supply chain, which is either vertical integration or a very good slide.
This lands the same week Nscale is reportedly raising up to $3.5B in pre-IPO financing. We wrote about that separately: backlog is not revenue. A signed multi-year compute commitment shows up in an announcement today and in a data center eighteen months from now.
Why 2027 delivery matters for your business
Nothing in this deal changes what you can build this quarter. That is the point worth internalizing, because the headline economics of AI infrastructure are being written on a timeline that has nothing to do with yours.
Two practical consequences.
First, physical AI is not a 2026 vendor category. If a supplier is pitching you humanoid-adjacent automation with delivery promises inside the next twelve months, the compute to train the underlying models is being poured into concrete in Texas next year. Ask what model they are actually shipping on today.
Second, this is what real capacity procurement looks like. Dollar figure, GPU count, named platform, named site, named quarter. When you buy AI capability — hosted inference, an agent platform, a per-seat tool — you are entitled to the same specificity. What model version. What region. What happens at renewal. Vendors who cannot answer at Figure's level of detail are selling you an allocation they have not secured.
The operator's read: the frontier is capital-intensive and slow, and your advantage is that you are neither. Build on models that exist today, with contracts you can exit, on data you own. When the 2027 capacity comes online, it will be a cheaper input to the same systems — not a reason to have waited.
Key takeaways
- Figure committed $3.5B initially to Nscale, with stated intent to exceed $6B, for up to 100,000 NVIDIA Vera Rubin GPUs
- Initial deployment begins in the second half of 2027 in Barstow, Texas — none of this capacity exists yet
- The compute trains Helix, Figure's humanoid robotics foundation model; Nscale is also taking an undisclosed equity stake
- Treat vendor claims about physical AI delivery inside twelve months with the training timeline in mind
- Demand Figure-grade specificity from your own AI vendors: model version, region, term, exit
You don't need 2027 hardware to ship in 2026. We build automation on models that are generally available today, with vendor-agnostic wiring so a cheaper option next year is a config change, not a rebuild. See what we've shipped or run the numbers on your workflow.
Sources: Figure and Nscale Sign Strategic Partnership, Nscale press release via PR Newswire.
- #figure
- #nscale
- #gpu
- #compute
- #robotics
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
AI agents ran a full ransomware breach in under 10 hours
Unit 42 documented an AI-driven ransomware intrusion that took under 10 hours instead of two weeks, used 50+ ATT&CK techniques, and left an 80-page audit.
Read itStop Rogue AI Act wants a machine-readable agent inventory
The Stop Rogue AI Act directs NIST to set AI agent security standards — machine-readable inventory, tamper-proof action logs, and teeth for federal contractors.
Read it