Izood MAG

AI — explainer

Thinking Machines $40B round: what Accel talks mean for AI valuations

4 min read
A black-and-white photograph of a man in a plain suit, seen from the waist up, stands at the left edge of the frame, his body turned slightly toward the center.

The reported $1B round at a $40B valuation

Thinking Machines is in talks to raise $1 billion at a valuation of at least $40 billion. The company was founded by Mira Murati, formerly OpenAI's CTO. It's an early-stage raise by the standards of what it would value the company at — which is the point worth staring at.

Existing investor Accel is said to be in talks to lead the round. That matters less for the money than for the signal: a backer with access to the books choosing to put more in at a number this large, rather than a new name setting the price.

Skip this if what you needed was deal size and lead investor — that's both sentences above. The rest is what the number implies, and it implies a lot.

Why the valuation dropped from $50B to $40B

The new round, if it closes, lands below the mark Thinking Machines set for itself late last year, when it reportedly sought $50 billion. A ten-billion-dollar haircut between the target and the reported number, and the company took the meeting anyway.

That gap is the story. Founders don't reprice ambition downward unless the alternative is worse — and the alternative here was likely no round at all on the original terms. Skip this if you want the hero narrative. The quieter read is that leverage moved from the seller to the buyer sometime in the last two quarters.

The revenue math behind a $40B price tag

Thinking Machines' annual revenue run rate stands at over $100 million. At that revenue figure, a $40 billion valuation reflects an extraordinarily high revenue multiple. Run the arithmetic and you land somewhere around 400x forward revenue.

The company's annual revenue run rate reportedly exceeds $100 million — consistent across both reports, which matters when every other number in this deal gets disputed. A 400x multiple is not a typo. Public SaaS companies, even the ones growing fast, rarely trade above 20-30x revenue. Venture-stage AI has pushed that boundary before, but not to four hundred.

Skip this if you're looking for a conventional valuation framework. There isn't one that makes this make sense. What the multiple actually tells you is that investors are pricing the round on something other than current revenue — model performance, compute position, founder pedigree, or the simple fear of missing the next OpenAI.

Take that with some caution — revenue run rates are self-reported, and a "run rate" extrapolates the most recent month or quarter forward. A strong month becomes a flattering annual number. The same caveat applies to every AI startup touting a run rate right now, not just this one.

What Inkling and Tinker mean for future revenue

In July, Thinking Machines shipped Inkling, an open-weight model. The model itself is free. The money comes from Tinker, the company's platform for adapting Inkling on proprietary data.

You pay usage-based compute fees while that customization runs. That is the entire monetization story right now: not seats, not subscriptions, not enterprise licenses. Compute metered by consumption.

The economics are different from closed-model APIs. There is no per-token markup on inference — the open weights mean customers can run Inkling anywhere. Tinker's revenue is tied to the training and fine-tuning workload, which is bursty. A customer might spend heavily for two weeks, then nothing for a quarter. That is a hard revenue profile to value at $40B.

From $2B seed to $12B valuation: the funding history

Thinking Machines' prior raise was a $2 billion round that valued the company at $12 billion. That round stands among the largest seed financings in history — the kind of number that used to signal a company's final private round, not its first.

Andreessen Horowitz led the investment, with participation from Nvidia, GV, Lightspeed, and Conviction Partners. Nvidia's presence is the tell — chipmakers don't write checks like that unless they see a customer who'll buy silicon at scale.

If the reported $40 billion number holds, investors who got in at that seed round are looking at a 3.3x markup in roughly a year. That's the multiple everyone else in the AI infrastructure space is now being measured against, whether their fundamentals justify it or not.

Founder pedigree and departures: risk to the valuation

The single biggest driver of Thinking Machines’ valuation is Mira Murati, and investors have said so with their money. The previous round was backed largely on her pedigree and that of the former OpenAI researchers who joined her — not on revenue, which sits around $100 million.

That is the problem. Co-founders Lilian Weng and Luke Metz have left and gone back to OpenAI. A startup valued at 400x revenue because of who works there gets re-priced fast when those people walk out the door.

Murati remains, and she is the name the valuation rests on. But the departures raise a fair question: if the round’s thesis was the team, what happens to the multiple as the team shrinks? I would not want to be the investor underwriting the answer.