What Happened

Thinking Machines Lab, the AI startup founded in early 2025 by former OpenAI CTO Mira Murati, is reportedly in discussions to raise $1 billion at a valuation of at least $40 billion. According to The Information, first reported by TechCrunch on September 3, 2026, existing backer Accel is in talks to lead the round.

The company's annual revenue run rate stands at over $100 million, according to a source with knowledge of Thinking Machines' financials. That revenue is generated through its Tinker platform, which charges usage-based compute fees for adapting open-weight Inkling models on proprietary data — a commercial model the company introduced in July 2026.

This would mark a significant step up from Thinking Machines' prior fundraise: a $2 billion round — among the largest seed financings in history — that valued the company at $12 billion. That round was led by Andreessen Horowitz, with participation from Nvidia, GV, Lightspeed, and Conviction Partners. Investors backed it largely on the pedigree of Murati and the former OpenAI researchers who joined her.

Notably, the $40B figure is below the $50B valuation Thinking Machines reportedly sought late last year, suggesting a recalibration of expectations in a more disciplined funding environment. The company has also experienced several high-profile departures, with co-founders including Lilian Weng and Luke Metz returning to OpenAI.

Neither Accel nor Thinking Machines responded to requests for comment, and the round has not been confirmed as closed.

Why It Matters

A $40 billion valuation on $100 million in ARR represents a roughly 400x revenue multiple — extraordinary even by AI-era standards. For context, this puts Thinking Machines in the same valuation tier as Cognition, which was reportedly in talks for a $40B valuation in August 2026. The signal here is twofold: investors are still willing to pay premium multiples for AI labs with elite pedigrees and real revenue, but the ceiling is being tested.

The gap between the $50B sought and the $40B reportedly landing is the more telling data point. It suggests that even founders with Murati's credentials are facing valuation discipline in late 2026. The era of unlimited upside pricing may be cooling — not freezing, but tightening.

The co-founder departures complicate the narrative. Lilian Weng and Luke Metz returning to OpenAI raises legitimate questions about whether Thinking Machines can retain the technical bench depth needed to compete at the foundation model frontier. Revenue growth is a strong counterargument, but the departure pattern is worth monitoring.

Who Is Affected

AI startup founders should treat this as a valuation benchmark. If you're raising in late 2026, investors will reference the $40B-on-$100M-ARR multiple and the Cognition talks at the same tier. Pedigree plus revenue is the formula — neither alone is sufficient for premium pricing.

Developers and operators building with AI APIs should note that Thinking Machines' Tinker platform is generating real commercial revenue, meaning it has enough adoption to be a production-grade consideration. The Inkling open-weight models and usage-based pricing model offer an alternative to closed API providers.

Enterprise buyers evaluating AI vendors may benefit from the valuation compression trend — as top labs face investor scrutiny, competitive pressure on pricing and contract terms could shift in buyers' favor.

Strategic Implications

For AI startup founders: The $40B-on-$100M-ARR multiple is your ceiling reference for late 2026. Expect investors to benchmark you against this and the Cognition $40B talks. If your revenue is below $100M ARR and your team lacks comparable pedigree, adjust your valuation expectations accordingly. The gap between sought and landed valuations ($50B → $40B) tells you that investors are negotiating harder even at the top of the market.

For developers/operators building with AI APIs: Thinking Machines' Tinker platform and Inkling open-weight models are generating $100M+ in usage-based revenue — enough signal that the platform has real adoption. If you're evaluating open-weight model providers, add Tinker to your shortlist, but validate model quality, API stability, and enterprise support before committing to production workloads.

For non-technical business owners evaluating AI tools: The valuation compression from $50B to $40B suggests even elite AI startups are facing investor scrutiny. This may translate to more competitive pricing and better enterprise terms as these companies need to demonstrate revenue growth to justify their valuations. Don't overpay for AI vendor contracts right now — the market is still settling.

What to Watch Next

Monitor whether the round closes at $40B or if further compression occurs — and watch for any additional co-founder or senior researcher departures, which would signal deeper team instability. Also track Tinker platform adoption metrics and Inkling model performance benchmarks, as these will determine whether the $100M ARR is sustainable or a one-time spike.

Frequently Asked Questions

Q: What is Thinking Machines Lab and who founded it?

A: Thinking Machines Lab is an AI startup founded in early 2025 by Mira Murati, the former CTO of OpenAI. The company builds open-weight AI models (Inkling) and offers a platform called Tinker that charges usage-based compute fees for adapting models on proprietary data.

Q: How much is Thinking Machines raising and at what valuation?

A: According to The Information, Thinking Machines is in talks to raise $1 billion at a valuation of at least $40 billion, with existing backer Accel reportedly leading the round. The company previously raised a $2 billion seed round at a $12 billion valuation led by Andreessen Horowitz.

Q: Why is the valuation lower than what Thinking Machines previously sought?

A: The company reportedly sought a $50 billion valuation late last year but is now settling for $40 billion, suggesting investor pushback and a more disciplined funding environment in late 2026. Co-founder departures, including Lilian Weng and Luke Metz returning to OpenAI, may also be a factor.