The Thinking Machines Lab funding round is nearing its close, with Mira Murati's startup reportedly raising $5 billion at a $40 billion pre-money valuation and Nvidia among the investors. Business Insider reported the deal on November 6, saying it could close as soon as this week. Nothing has been announced. The company declined to comment, and Accel did not respond to requests for comment.
The timing is hard to ignore. The same week, Thinking Machines released Inkling 2 Small, a 594-billion-parameter open-weight model, and said it will publish the weights under the Apache 2.0 license next week. Investors are reportedly prepared to pay more than triple the company's last known valuation, even after a recent AI stock selloff.
Thinking Machines Lab funding round at a $40 billion valuation
Business Insider's reporting puts the round at $5 billion, led by Accel, at a $40 billion pre-money valuation. If the deal closes on those terms, the post-money valuation would be about $45 billion, according to AI Weekly's roundup of the report. That is a steep climb from the company's last public round. In July 2025, Thinking Machines raised a $2 billion seed round at about $12 billion, led by Andreessen Horowitz, according to the company's Wikipedia entry.
The pre-money figure is about 3.3 times the seed valuation, which is why the story is drawing attention from investors. Business Insider also reported that the company abandoned talks last year to raise at a $50 billion valuation. The price has moved up and down in a short period, and the latest figure is the one that matters most to current backers.
The round has also grown in size. In September, The Information reported that Thinking Machines was seeking about $1 billion. It later updated that figure to $5 billion to $6 billion, according to TheNextWeb's summary of the report. Readers who saw the earlier number should treat it as outdated.
How Nvidia fits into the deal
Nvidia's role carries the most money. Business Insider reported that Nvidia is in talks to invest about $2.5 billion. The Information's earlier reporting said Nvidia would supply about half of the $5 billion to $6 billion round, which would put its share at roughly $2.5 billion to $3 billion. The figures point in the same direction, but they are still reports, not company disclosures.
The two companies already have a longer relationship. In March, Thinking Machines and Nvidia announced a multiyear partnership to deploy at least one gigawatt of Nvidia Vera Rubin systems, with first deployment targeted for early 2027. Nvidia invested at the same time, but neither side disclosed the amount.
That arrangement ties a large share of Nvidia's future chip sales to one lab's growth. For chip buyers, it shows how frontier labs and their hardware suppliers are becoming bound together. Nvidia's dealings with AI companies have also drawn attention in Washington. Lawmakers recently asked Nvidia CEO Jensen Huang to brief them on the company's ties to Megaspeed.
Inkling 2 Small: what the new model is
Thinking Machines launched Inkling 2 Small on November 5. It is a mixture-of-experts model with 594 billion total parameters and 17 billion active, built for autonomous coding, visual reasoning and knowledge work. The company makes it available through its Tinker platform.
The model follows Inkling, a larger open-weight family the company released in July 2026 with 975 billion parameters under Apache 2.0. The company then released a distilled Inkling Small on July 31. Inkling 2 Small is the newest step in that line.
The benchmark picture is mixed. One report, from Techstrong.ai, says Inkling 2 Small beat Claude Sonnet 5.5 on one test, trailed it on another, and used substantially more tokens. The same coverage cites one analysis finding that the model generated roughly four times as many output tokens as a rival model. USA 24 did not independently verify those results, so treat them as a starting point for your own testing.
Why the valuation matters beyond the lab
For investors, the reported price signals that private AI valuations are still climbing even as public AI stocks have stumbled. AI Weekly's daily roundup placed the Business Insider report alongside an AI stock selloff and a midterm-election backlash against data centers. Those market details come from that roundup and were not verified in this reporting. Still, the contrast is the story: late-stage investors appear willing to pay up while public markets reprice.
That matters for people who do not own lab shares directly. Retirement accounts and funds with private AI exposure can carry these marks, and private rounds often set the values those funds report. USA 24 has also reported that Anthropic's $1.5 trillion IPO faces valuation pushback after an AI rout. Read together, the two stories show how far opinions differ on what AI companies are worth.
The talent picture complicates the valuation story. Business Insider reported in May that about one-third of the founding team had left for rivals, and the company has lost people to Meta and OpenAI. A high valuation buys compute and hiring power, but it also raises the bar for what the company has to deliver.
What the numbers still don't settle
Several questions remain open. The round has not been announced as closed. Sources describe it as 'set to close,' and Thinking Machines declined to comment.
Nvidia's share is also unsettled. Business Insider's $2.5 billion figure and The Information's earlier estimate of about half of a $5 billion to $6 billion round point in the same direction. Neither has been confirmed by the company.
The final terms could also differ from the early reports. The $1 billion figure from September circulated widely before The Information updated it, which is a reminder that early numbers in this market often shift.
What developers should check before using Inkling 2 Small
Consumers have no action to take here. Developers and companies that evaluate open-weight models have a few concrete steps.
- Wait for the weights and read the license. The company says it will publish them under Apache 2.0 next week, so confirm the released license text matches that claim before you build on it.
- Compare cost per completed task, not the listed price per token. The token-usage findings suggest a model can cost more overall if it writes much longer outputs.
- Run your own evaluations. Benchmark results have been mixed, so test coding and reasoning tasks on your own data.
- Plan the hardware. A 594-billion-parameter model needs memory for all of its weights, even though only 17 billion parameters are active for each token, so self-hosting is a serious infrastructure decision.
What to watch next
Three events will show whether the story holds. The first is a formal announcement of the round, including the investor list and Nvidia's actual check. The second is the promised release of Inkling 2 Small weights under Apache 2.0, which will let outside researchers test the benchmark claims. The third is the Nvidia compute deal, with first Vera Rubin deployment targeted for early 2027.
If the round closes near $5 billion at a $40 billion pre-money valuation, it will reinforce a pattern in which the largest AI labs raise at prices far above their last rounds. The harder test is whether investor appetite for frontier labs survives the next AI stock selloff.
