Thinking Machines Seeks $1B at $40B Valuation, Led by Accel

Thinking Machines, the AI lab started early last year by former OpenAI CTO Mira Murati, is in talks to raise $1 billion at a valuation of at least $40 billion, according to The Information. Existing investor Accel is in discussions to lead the round. The company's annual revenue run rate has crossed $100 million, but that figure still puts the proposed valuation at an extremely high revenue multiple. If the round closes, it would mark a step down from the $50 billion valuation the startup reportedly targeted late last year.

What happened

Detail	Figure
Round size (proposed)	$1 billion
Valuation (proposed)	At least $40 billion
Previous valuation target	$50 billion (reportedly sought late 2025)
Prior seed round	$2 billion at $12 billion valuation
Annual revenue run rate	Over $100 million
Round lead (proposed)	Accel (existing investor)

The Information first reported the deal on September 3, 2026. Accel and Thinking Machines did not respond to requests for comment at the time of publication.

Thinking Machines was founded by Mira Murati, who served as CTO at OpenAI before departing to start the lab. The company’s $2 billion seed round, led by Andreessen Horowitz and joined by Nvidia, GV, Lightspeed, and Conviction Partners, was one of the largest seed financings ever recorded. Investors at that stage were betting heavily on the credentials of Murati and the team of former OpenAI researchers she brought with her.

What the company actually sells

In July, Thinking Machines launched Inkling, an open-weight model (meaning the model weights are publicly available for download and modification). Revenue comes from usage-based compute fees charged through its Tinker platform, which lets customers adapt models on their own proprietary data.

That business model is worth noting: Thinking Machines is not selling subscriptions or API credits in the traditional sense. It charges for compute consumed when businesses fine-tune Inkling on private datasets, which keeps them close to customer workflows.

Why it matters

A $40 billion valuation on $100 million in annual revenue works out to a 400x revenue multiple. Even by 2025-2026 AI startup standards, that number is extreme. For context, the company’s seed-stage valuation of $12 billion was already considered aggressive, and it has now tripled on a revenue base that, while growing, is still relatively modest.

The valuation also came down from the $50 billion mark the company reportedly sought just months ago. That compression is a useful signal: even well-funded AI labs with founder pedigree are finding that the market has a ceiling, at least for now. The departure of co-founders Lilian Weng and Luke Metz, both of whom returned to OpenAI, likely did not help the narrative going into this raise.

For businesses watching the AI infrastructure space, rounds like this shape which platforms get the resources to scale and which compute ecosystems become dominant. Thinking Machines’ bet on open-weight models with proprietary fine-tuning infrastructure is a direct competitor to closed-API approaches from OpenAI and Anthropic.

You can follow similar funding signals and model releases in our AI news coverage as the space continues to move quickly.

Our take

A 400x revenue multiple is not a business valuation in any traditional sense. It is a bet on a future that has not arrived yet, and the reduced ask from $50 billion to $40 billion suggests even optimistic investors are applying some friction. That said, the open-weight plus compute-fee model is genuinely interesting: it avoids the API lock-in criticism while still generating recurring revenue tied to usage.

The departure of founding researchers is the bigger concern here. The original seed thesis was almost entirely built on team quality. When that team changes, the justification for the premium needs to shift to product traction, and $100 million in run rate, while real, has to grow fast to support a $40 billion cap table.

If you are a business evaluating whether to build on Thinking Machines’ Tinker platform, the funding trajectory suggests the company is likely to survive and keep investing in the product. But the dependency on compute-fee pricing means your costs will scale directly with usage, so model the economics carefully before committing. If you need help thinking through AI integration options for your stack, that kind of vendor evaluation is exactly where it pays to get specific.

What to do about it

1. Watch for the round to officially close before treating Thinking Machines as a long-term infrastructure partner: a reported deal is not a signed term sheet.
2. If Inkling’s open-weight approach interests you, test fine-tuning on a small, non-sensitive dataset through Tinker to get real cost data before committing to a production workflow.
3. Track founder and leadership stability at any AI lab you depend on. High-profile departures at the co-founder level are worth monitoring over the next 6 months.
4. Compare compute-fee pricing against API-based alternatives from OpenAI, Anthropic, and others using your actual expected token or compute volumes, not list prices.

The round, if it closes, cements Thinking Machines as a well-capitalized player, but the valuation math only works if revenue scales by an order of magnitude in the next few years.

Source: TechCrunch · AI