Nvidia Figure Investment: $1B Bet Tests the Circular AI Machine
Talks over a $1 billion Nvidia Figure investment deepen a capital loop that ties the chipmaker's equity stakes to the GPU demand it books as revenue.
Nvidia has been in talks to put roughly $1 billion into Figure, the humanoid robotics developer Brett Adcock founded in 2022, extending a capital campaign that has turned a slice of the chipmaker's data-center profits into equity across the AI stack. The prospective Nvidia Figure investment would deepen a relationship that began with a much smaller check two years ago. Figure was most recently valued at about $39 billion.
The discussions surfaced this week and have not been confirmed, and no terms have been disclosed. They follow a string of completed and pending deals that have made Nvidia one of the largest financial backers of the industry it supplies. Nvidia closed its $12.9 billion purchase of the open-source model platform Hugging Face, has committed about $30 billion to OpenAI across several funding rounds, and is negotiating roughly $3 billion for Perplexity, a company last valued at $35 billion.
Figure's history explains why Nvidia keeps returning to it. The company raised $675 million in February 2024 from a consortium that included Jeff Bezos, Microsoft, Nvidia, Intel, and the venture arms of Amazon and OpenAI, at a $2.6 billion valuation. By September 2025 it had closed a Series C of more than $1 billion that valued it at $39 billion, with Intel Capital and Qualcomm Ventures joining Nvidia on the cap table. Lifetime fundraising stands at roughly $1.9 billion.
The ties run through procurement as well as equity. Figure's robots run on Nvidia hardware, and the stake sits beside a commercial relationship that routes spending back into Nvidia's data-center business. Nvidia supplies the compute first, then buys into the companies that buy the compute.
| Target | Commitment | Valuation or note |
|---|---|---|
| Hugging Face | $12.9B acquisition, completed | More than 80x annualized revenue |
| OpenAI | About $30B across funding rounds | Final tranche closed October 1, 2026 |
| Perplexity | About $3B under discussion | Last valued at $35B |
| Figure | About $1B under discussion | Last valued at about $39B |
Nvidia's acquisition engine now runs past $140 billion, with robotics, autonomous driving, and on-device inference among the stated targets. The DGX Spark line, built to run AI agents locally, ties those bets back to Nvidia's own silicon. The logic is that every new place AI gets deployed becomes a new place Nvidia hardware gets sold, and the Nvidia Figure investment fits that pattern: fund the customer, supply the compute, book the revenue.
The pattern predates this round. Nvidia's first check into Figure came in 2024, when the startup was worth $2.6 billion, and the position has since been measured against a valuation roughly fifteen times higher. That track record explains why the chipmaker keeps writing larger checks into the same companies rather than spreading its bets across new names.
The $1 billion under discussion is earmarked for the compute and training-data work that turns demonstrations into dependable labor. Adcock founded Figure to build general-purpose humanoid robots. Humanoid makers are targeting industrial and service roles that employers struggle to staff, particularly overnight shifts, which is the commercial case Figure is pitching to manufacturers.
Figure's technical progress gives the equity stake a rationale beyond financial engineering. The company set aside $1 billion to pay people over 12 months to film themselves folding laundry, making beds, and picking objects off the floor. Within a month, that data had produced a sixfold gain in its robots' success at those chores. Reaching 95% reliability for home use remains the next threshold, and Figure intends to deploy humanoids in automotive manufacturing facilities along the way.
Competitive pressure shapes the timing. Tesla has pursued its own humanoid program, and Figure's roughly $39 billion valuation places it among the largest private developers in the category. Holding that premium requires shipping robots that work outside controlled demonstrations, and the data-collection spending is the mechanism Figure is betting on to close the gap. ABB, Agility, KUKA, Teradyne, and Yaskawa work with Nvidia's simulation tools, giving the chipmaker exposure to the category regardless of which developer wins.
The Circular Logic Behind the Nvidia Figure Investment
The question for enterprise buyers and investors is what the loop does to reported demand. Nvidia funds a customer, that customer commits to buying Nvidia silicon, and the resulting purchases can appear as data-center revenue in the same period the equity stake is written up. Growth produced this way is real, but it is partly generated by the supplier rather than independently by the market.
Figure's valuation history shows how fast those revaluations can travel. A $2.6 billion valuation in early 2024 became $39 billion by late 2025, roughly a fifteenfold increase in under two years, on cumulative fundraising of about $1.9 billion. The distance between capital raised and valuation assigned reflects expectations about physical AI rather than revenue already booked.
Hugging Face illustrates the same tension on the software side. Nvidia agreed to pay more than 80 times the platform's annualized revenue, a multiple that reflects strategic urgency after rival interest from Salesforce. Hugging Face hosts the open-source ecosystem Nvidia now owns, and ownership changes the incentives for the developers who build on it.
The strategic risk cuts both ways. If Figure's valuation falls, Nvidia records a write-down on its stake at the same time that Figure's GPU commitments become harder to finance, squeezing the revenue line. If Figure succeeds, Nvidia books a gain on the equity and sells more silicon, but the customer is also a portfolio company whose largest shareholder benefits from its spending.
Rivals face difficult arithmetic. A competing chipmaker cannot easily match a bundle that pairs silicon with an equity check, and startups that accept Nvidia's money may find their procurement options narrowed. Figure's decision to end its OpenAI partnership in 2025 shows how fluid those alliances remain even as the capital ties tighten.
For enterprise buyers, the practical consequence is that vendor risk and supplier risk are merging. A company evaluating Nvidia hardware is also, indirectly, taking a position on the startups Nvidia has funded, because a correction in those valuations would affect both the chipmaker's investment returns and the demand forecasts behind its roadmap.
What to Watch Next
Confirmation of the Nvidia Figure investment and its terms is the first signal to track. The final tranche of the OpenAI commitment closed on October 1, 2026, and the Perplexity discussions remain open, with that company valued at $35 billion. Each of those closing, or stalling, will test how durable the model proves. Nvidia's Isaac and Cosmos robotics platforms, which simulate environments and generate training data for partners including ABB, Agility, KUKA, Teradyne, and Yaskawa, are the products that must justify the strategy commercially.
Why this matters
Nvidia now buys stakes in the companies doing the AI buildout, in addition to selling them the hardware. For enterprise buyers, that means the vendor's fortunes and its customers' fortunes increasingly move together, which complicates any procurement or due diligence that treats them as separate risks. Investors should read the equity stakes and the data-center revenue side by side, because the two numbers now reinforce each other in ways that can flatter demand as much as create it.
Photo by Brecht Corbeel on Unsplash
Related Articles
- Nvidia OpenAI Circular Financing: $250 Billion Guarantee Reshapes AI Infrastructure Risk
- Nvidia Pitches AI Compute as an Investable Asset Class
- Nvidia Hugging Face Acquisition: What $12.93B Buys at the Open Model Hub
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.