What Happened

According to The Information, reported via The Next Web on 6 September 2026, Thinking Machines Lab is in talks to raise $5bn to $6bn at a pre-money valuation of at least $40bn. Accel, which participated in the seed round, is reportedly in talks to lead. Nvidia would supply approximately $2.5bn to $3bn — roughly half the total capital.

The sourcing is unusual. The Information's Amir Efrati stated on X that the details came from Andreessen Horowitz, which had briefed its own limited partners on the deal. This is a backer informing its LPs, not a company announcing a closed round.

This figure corrects the initial version of the story, published on 3 September, which put the raise at over $1bn. The updated number is five to six times larger and names Nvidia as a major participant. Most aggregation that followed carried the original $1bn figure before the update landed.

For context: the seed round in July 2025 raised $2bn at $12bn post-money, led by Andreessen Horowitz with Accel and Nvidia among the investors. By late 2025, the company reportedly sought $50bn or more; those talks did not close. The current $40bn pre-money is at least 20% below that ask and more than three times the seed valuation.

Why It Matters

The revised raise size changes the infrastructure equation. Thinking Machines and Nvidia announced a multiyear partnership in March 2026 to deploy at least one gigawatt of Nvidia's Vera Rubin systems, with the first deployment targeted for early 2027. At $1bn, the lab would have covered a fraction of that cost. At $5-6bn — with the chipmaker supplying half — the math works differently.

Nvidia is effectively financing its own demand. Every job that runs on Thinking Machines' Tinker platform (GA since December 2025) runs on Nvidia hardware. The company gives away its open-weight models — Inkling, a 975-billion-parameter mixture-of-experts system shipped in July 2025, earns nothing on the weights themselves. Revenue comes from Tinker, which sells GPU cluster access for training and fine-tuning open models. Nvidia funding the round closes a circle: it invests in the company, the company buys Nvidia systems, and customers pay Tinker to use those systems.

This pattern extends beyond Thinking Machines. Two days after this story broke, Nvidia appeared as a backer of Nscale's pre-IPO financing. The market has so far priced this circular dynamic as growth rather than risk.

The valuation multiple is steep by any measure. Self-reported annualised revenue is in the hundreds of millions, but that figure is unverified — TechCrunch, working from a different source, put it at over $100m. Against $40bn pre-money, the multiple lands somewhere between roughly 80x and 400x revenue. Efrati called the price lower than the company wanted but still high for a lab this young.

Who Is Affected

AI startup founders should note that Nvidia's strategic capital is available at scale if your business commits to Nvidia infrastructure. The $40bn pre-money for a lab with self-reported hundreds of millions in revenue represents the current ceiling for frontier lab pricing — not the floor.

Developers and GPU cloud customers using or evaluating Thinking Machines' Tinker platform should expect accelerated infrastructure rollout if this round closes. The 1GW Vera Rubin deployment targeted for early 2027 becomes materially more feasible at $5-6bn than at $1bn.

Enterprise buyers evaluating open-weight models should weigh the funding signal against leadership risk. Multiple co-founders and senior researchers departed in late 2025 and early 2026 — CTO Barret Zoph left mid-all-hands in January and landed at OpenAI (then Google), co-founder Luke Metz departed, and Lilian Weng stepped down in July. Chief scientist John Schulman stayed, and Soumith Chintala (PyTorch co-creator) was brought in to run technical work. Murati retains voting control over the board on major decisions, meaning investors at $40bn are buying her judgement with limited recourse.

Strategic Implications

For AI startup founders: Nvidia's pattern of funding its own customers is now a repeatable financing pathway. If your startup commits to Nvidia infrastructure at scale, strategic capital from the chipmaker is a realistic component of your round. Price knowing that $40bn pre-money for a lab with self-reported hundreds of millions in revenue is the current ceiling.

For developers building with AI APIs: Tinker's expanded GPU capacity — backed by $5-6bn and a 1GW Nvidia deployment — could make it a more competitive option for training and fine-tuning open models by early 2027. If you're evaluating open-weight alternatives to closed APIs, this round materially improves Tinker's viability.

For non-technical business owners: The funding scale signals serious capital behind open-weight AI, but leadership churn means execution risk remains high. Don't commit to a single model provider based on funding headlines — evaluate actual product reliability and roadmap delivery.

What to Watch Next

Monitor whether the round closes at the reported $5-6bn figure or whether terms shift before finalisation. Watch for Nvidia's next strategic investment in an infrastructure-buying startup — if the pattern continues, it signals a deliberate channel strategy. Track Tinker's capacity announcements in Q1 2027 against the Vera Rubin deployment timeline.

Frequently Asked Questions

Q: How much is Thinking Machines Lab raising?

A: According to The Information (reported 6 September 2026), Thinking Machines Lab is in talks to raise $5bn to $6bn at a pre-money valuation of at least $40bn. Accel is reportedly in talks to lead, with Nvidia supplying approximately $2.5-3bn. The round has not closed.

Q: Why is Nvidia investing in Thinking Machines Lab?

A: Nvidia is supplying roughly half the capital in the reported round. The commercial logic is that Thinking Machines buys Nvidia's Vera Rubin systems for its 1GW deployment, and every job run on its Tinker platform uses Nvidia hardware. Nvidia is effectively financing its own customer — a pattern it has repeated with other infrastructure-buying startups.

Q: What happened to Thinking Machines Lab's leadership team?

A: Several key figures departed: CTO and co-founder Barret Zoph left in January 2026 (to OpenAI, then Google), co-founder Luke Metz departed, and Lilian Weng stepped down in July 2026. Chief scientist John Schulman stayed, and Soumith Chintala (PyTorch co-creator) was brought in to lead technical work. Founder Mira Murati retains voting control over the board.