Accel may lead a $1 billion round in Thinking Machines at a $40 billion valuation, according to reports. That is more than three times the $12 billion the company was worth at its previous raise. Annualized revenue runs above $100 million, a source familiar with the company's financials said — which puts the proposed price at roughly 400 times revenue. Accel and Thinking Machines did not respond to requests for comment.
The previous round was $2 billion, one of the largest early-stage raises on record. Andreessen Horowitz wrote the first check, and Nvidia, GV, Lightspeed and Conviction Partners followed. What they bought, mostly, was a roster: Mira Murati's reputation and the former OpenAI researchers who left with her. There was no product to underwrite at the time.
There is one now. In July the company released Inkling, a model with open weights, and it earns money through usage-based compute charges when customers adapt models on their own data through its Tinker platform. That is a real business with a real meter on it, and $100 million annualized is a genuine number rather than a projection. It is also the first time Thinking Machines has had to be valued against anything other than its founders' résumés.
The shape of this round is what stands out. The company is reportedly asking for half as much money as last time at more than three times the price. A smaller raise at a much higher valuation is not what a company does when it urgently needs capital; it is what a company does when it wants a new number on the board with minimal dilution. Whether the $1 billion is needed for compute or exists mainly to establish the $40 billion is not something the reporting settles.
The 400x multiple is the part that should not be waved through. Usage-based compute revenue is the kind of revenue that carries a cost of goods sold underneath it, and nothing in the reporting says what Thinking Machines keeps after paying for the GPUs its customers' fine-tuning runs on. No gross margin, no customer count, no growth rate — just an annualized figure above $100 million and a valuation 400 times larger. At that multiple, the revenue is not what is being priced. The team is.
Which makes the departures more than gossip. Lilian Weng and Luke Metz, both founders, have already left and both went back to OpenAI, along with several other notable staff. Investors in the $2 billion round paid for a specific concentration of people; some of that concentration has since dispersed to the company it came from. The $12 billion valuation had a straightforward justification even without a product. The $40 billion one requires believing that what remains is worth more than what assembled it.