Thinking Machines reportedly raising $1B at $40B valuation: what builders should know

Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, is in talks to raise $1 billion at a valuation of at least $40 billion, with Accel expected to lead the round. For builders, the key takeaway is that the market is willing to pay a premium for open-weight model platforms that monetize through usage-based compute fees on proprietary data.

The reported terms and what they mean

According to The Information and TechCrunch, Thinking Machines is seeking at least $1 billion in new funding at a valuation of around $40 billion, down from the $50 billion target it pursued late last year. The company's annual revenue run rate exceeds $100 million, which at a $40 billion valuation implies an extraordinarily high revenue multiple. The round would be led by existing backer Accel, with other investors likely to follow.

The startup previously raised a $2 billion seed round led by Andreessen Horowitz, with participation from Nvidia, GV, Lightspeed, and Conviction Partners. That round valued the company at $12 billion, largely on the strength of Murati's reputation and the team of former OpenAI researchers.

In July, Thinking Machines launched Inkling, an open-weight model that generates revenue by charging usage-based compute fees for adapting models on proprietary data via its Tinker platform. This monetization approach is central to the company's strategy and a key reason investors are willing to pay a high multiple.

Why this matters for AI tooling builders

The potential round signals strong investor appetite for AI labs that combine open-weight models with a platform for custom adaptation on proprietary data. If completed, it would set a precedent for revenue multiples in the AI tooling space, showing that the market values not just model performance but also the infrastructure for fine-tuning and deployment.

For builders evaluating model platforms, Thinking Machines' approach offers an alternative to API-only models: you get open weights and pay for compute when adapting them to your data. This could be attractive for teams that need control over model customization without committing to a full self-hosting setup. The involvement of Nvidia as an existing backer also suggests potential hardware-level optimizations for the Tinker platform.

However, the company has seen several high-profile departures, including co-founders Lilian Weng and Luke Metz returning to OpenAI. These leadership changes introduce execution risk and could affect the product roadmap. Builders should monitor whether the company can retain top talent while scaling.

Caveats to watch

The fundraising is still in discussions and may not close as reported. The valuation, while lower than the earlier target, remains extremely high relative to revenue. The company's reliance on usage-based compute fees means its revenue growth depends on customer adoption of the Tinker platform, which is still early. Additionally, the talent drain to OpenAI raises questions about long-term research and product velocity.

If you're building AI products that require fine-tuning on proprietary data, Thinking Machines' model is worth watching. The Inkling model is open-weight, so you can inspect and modify it, but the Tinker platform charges for compute usage. This could be cost-effective for teams that need occasional adaptation but expensive for high-volume retraining. The reported round, if completed, would give the company significant runway to prove its model at scale.