Thinking Machines seeks $1B-plus at a $40B valuation after last fall's $50B ask
The lower target still values Mira Murati's AI lab at more than three times its July 2025 post-money price.
By Ryan Merket · Published
Primary source: Stephanie Palazzolo on X
Why it matters
The talks test whether shipped products and early revenue can sustain a frontier AI valuation after founder turnover and an abandoned $50B-plus target.
Mira Murati (@miramurati) is in talks to raise more than $1 billion for Thinking Machines Lab at a valuation above $40 billion, The Information reported on Thursday.
The proposed price is at least 20% below the $50 billion-plus valuation Thinking Machines sought last fall. It would still mark a steep increase from July 2025, when Murati closed a $2 billion seed round that valued Thinking Machines at $12 billion after the investment. Andreessen Horowitz led that round, with participation from investors including Accel, Nvidia, AMD, Cisco, ServiceNow and Jane Street.
Stephanie Palazzolo (@steph_palazzolo), one of The Information reporters behind Thursday's story, wrote that Thinking Machines is generating annual revenue in the hundreds of millions of dollars. That figure is self-reported through people familiar with the business rather than audited public financial statements. Even at the top end of that description, a $40 billion valuation would place a large premium on revenue that Thinking Machines has only recently begun producing.
Murati's ability to command that price begins with her record at OpenAI. Before founding Thinking Machines, she ran OpenAI's research, product and safety functions as chief technology officer and briefly served as interim CEO during the November 2023 boardroom crisis. She left OpenAI in September 2024 and unveiled Thinking Machines in February 2025, recruiting researchers who had worked on ChatGPT, Character.AI, Mistral models and widely used open-source projects including PyTorch.
Revenue arrived after the valuation
Thinking Machines raised its seed financing before releasing a commercial product. Murati has since given investors more to price.
Tinker, released in October 2025 and made generally available that December, sells access to infrastructure for training and fine-tuning open models. Researchers control training data and algorithms through an API while Thinking Machines manages the underlying GPU clusters. The product supports models from Thinking Machines and outside labs, including families developed by Moonshot AI, Alibaba's Qwen group, DeepSeek, Nvidia and OpenAI.
The product points to the likely source of the reported revenue: developers and enterprises paying Thinking Machines to run expensive training workloads without operating their own clusters. Thinking Machines has not published revenue, customer concentration or margins, leaving investors to assess how much of the reported sales growth reflects durable usage and how much passes through to compute providers.
Murati expanded the product line in July 2026 with Inkling, a 975-billion-parameter mixture-of-experts model released with open weights. Thinking Machines describes Inkling as a customizable multimodal model that processes text, images and audio. Thinking Machines also released Inkling-Small, a 276-billion-parameter version with 12 billion active parameters.
Those launches give the financing talks a firmer foundation than last fall's $50 billion proposal, which did not close. The revised valuation also shows that shipping products and producing revenue did not preserve the earlier price.
A billion dollars does not buy a gigawatt
The amount under discussion would support a capital-intensive infrastructure plan. In March, Thinking Machines and Nvidia announced a multiyear partnership to deploy at least one gigawatt of Nvidia's Vera Rubin systems, with the first deployment targeted for early 2027. Nvidia also made an investment in Thinking Machines, although the companies did not state its size.
A gigawatt-scale buildout places Thinking Machines in direct competition with better-capitalized frontier laboratories for chips, power and data-center capacity. Raising more than $1 billion would strengthen Murati's balance sheet, though the round alone would cover only part of an infrastructure program at that scale.
Murati is pursuing the financing after turnover among the founders she recruited from OpenAI. Andrew Tulloch left for Meta in late 2025. Barret Zoph and Luke Metz departed in January 2026, and Lilian Weng stepped down in July. Thinking Machines retained co-founder and chief scientist John Schulman and appointed PyTorch co-creator Soumith Chintala to lead technical work after Zoph's exit.
The proposed financing therefore prices Murati's operating record, the revenue emerging from Tinker and the prospect that Inkling can turn a model-training platform into a broader AI supplier. At more than $40 billion, investors would be paying over three times the price set by last year's seed round. The discount from the abandoned $50 billion-plus target is meaningful. The remaining multiple is the larger statement about how much private markets are still prepared to pay for an OpenAI veteran with products, revenue and an enormous compute bill.