Thinking Machines Lab Reportedly Nears $5 Billion Raise at $40 Billion Valuation
Mira Murati’s Thinking Machines Lab is reportedly about to raise another $5 billion.
Business Insider reported on Friday, November 6, that the San Francisco AI lab is closing in on a new round at a $40 billion pre-money valuation, with Accel leading. Counting the new money, that puts the company at roughly $45 billion.
According to the report, which cited people familiar with the matter, the deal could close within days.
Nothing is final yet. Thinking Machines declined to comment, and Accel didn’t respond to requests for comment. No other investors have been named.
The timing is interesting. The news landed within a day of the company releasing Inkling 2 Small, a new model aimed at one of the most technical jobs in AI: training other AI models.
From $12 billion to $40 billion
Thinking Machines has been an unusual company from the start. Murati, who was OpenAI’s chief technology officer, founded it in San Francisco in February 2025. A few months later, in July 2025, it raised a $2 billion seed round at a $12 billion valuation, before it had shipped a single product.
That was the largest seed round on record, and it drew a remarkable group of backers: Accel, Andreessen Horowitz, Nvidia, AMD, ServiceNow, Cisco and Jane Street. Nvidia went further in March 2026, announcing a multi-year partnership to deploy at least a gigawatt of its Vera Rubin systems with Thinking Machines, starting in early 2027. It also made an undisclosed investment at the time.
If the new round closes as reported, the pre-money valuation would be about 3.3 times the seed price. Thinking Machines would have raised around $7 billion in disclosed funding, not counting Nvidia’s March investment.
That’s a big jump. It’s also lower than it could have been. The company reportedly held talks late last year about raising at around $50 billion, which never came together.
What Thinking Machines actually sells
For a while, the main criticism of Thinking Machines was that nobody quite knew what it was building. That’s changed.
Its first product is Tinker, a cloud platform that lets companies train custom AI models on their own data. Thinking Machines supplies the computing power and charges for usage. A finance team, for example, could train a model on its own earnings call transcripts instead of relying on a general-purpose chatbot.
In July, the company released Inkling, its first open-weight model. And this week came Inkling 2 Small.
A source told TechCrunch in September that the company’s annual revenue had passed $100 million. That’s real money for a lab this young, but it also shows how much of the valuation rests on future growth. At $40 billion, investors are paying something like 400 times revenue.
Inside Inkling 2 Small
Inkling 2 Small is “small” only by frontier-lab standards. According to SiliconANGLE, it has 594 billion parameters in total, with about 17 billion active for any given prompt. That design, known as mixture of experts, keeps it cheaper to run than its size suggests. It also includes a vision encoder, so it can look at images as well as read text.
The interesting part is what it was trained to do. Thinking Machines trained it on hundreds of thousands of long coding tasks, including what it calls “autoresearch”: post-training neural networks, optimising code that runs language models, and other work normally done by AI researchers.
The company shared one example. It asked Inkling 2 Small to fine-tune Qwen3.5-4B, a small open-source model, for forecasting. It says the model doubled Qwen’s forecasting score in four and a half hours.
Thinking Machines also says the model beats several open-source rivals on knowledge-work and financial benchmarks, and outperforms Mistral Large 4 on the Artificial Analysis index. Several Chinese open-weight models still score higher on that same index, though. Users can also trade answer quality for lower cost, depending on the task.
Inkling 2 Small is available now through Tinker. The company plans to release the full weights under the Apache 2.0 open-source license, pending external safety testing.
The fit with the business is obvious. If your product is a platform for training custom models, a model that’s good at training models is exactly what you’d want to build.
The talent problem
Not everything has gone smoothly. Thinking Machines has lost a significant part of its founding team over the past year.
Co-founder Andrew Tulloch left for Meta. Barret Zoph and Luke Metz, both co-founders, went back to OpenAI. Business Insider estimates that around a third of the founding team has now left, many lured away by huge pay packages from Meta and OpenAI.
That kind of turnover would worry investors at most startups.
The fact that Accel is reportedly still willing to lead a $5 billion round suggests it believes the remaining team, and the products it’s now shipping, can carry the company forward. The lab has also kept hiring, recently bringing on competitive programmer Neal Wu.
A busy week for AI money
This report came out on the same day Bloomberg reported that OpenAI had raised $11 billion at a $1.25 trillion valuation. Both deals point to the same thing. Investors are still writing enormous cheques for AI labs, but they’re pushing back harder on price than they were a year ago.
OpenAI reportedly scaled back its own round. Thinking Machines is reportedly raising below the $50 billion it discussed last year. The money is still flowing. It’s just coming with more negotiation.
What happens next
The first thing to watch is whether the round actually closes this week, and who joins Accel. Given its history, Nvidia would be an obvious candidate.
After that, attention shifts to the Inkling 2 Small weights. An Apache 2.0 release would put a very capable model in developers’ hands for free, and it would be a strong signal that Thinking Machines is serious about open models.
Longer term, the question is whether Tinker can turn into a business big enough to justify a $40 billion price tag. Revenue of $100 million is a solid start. Investors are clearly betting it’s only the beginning.
Key facts
Company: Thinking Machines Lab
Headquarters: San Francisco, California
Founded: February 2025, by Mira Murati
Round: not specified (reportedly closing)
Amount reported: $5 billion
Valuation: $40 billion pre-money (about $45 billion post-money)
Lead investor: Accel (reported)
Reported: November 6, 2026, by Business Insider
Status: not confirmed by the company
Previous round: $2 billion seed at a $12 billion valuation (July 2025)
Latest product: Inkling 2 Small, released November 5, 2026
Sources: SiliconANGLE, Dataconomy, Tech Funding News