Nvidia Bets Big on Mira Murati’s Thinking Machines Lab With Gigawatt AI Deal
The second wave of AI startups with foundational models is rolling in: Following major investments of one billion dollars each in AMI Labs by Yann LeCun (ex-Meta), World Labs by Fei-Fei Li and Safe Superintelligence (SSI) by Ilya Sutskever (ex-OpenAI), the next mega-deal for one of these next-generation startups is now following.
The startup Thinking Machines Lab, founded by Mira Murati, has entered into a multi-year strategic partnership with Nvidia. The agreement includes the provision of at least one gigawatt of computing power through Nvidia’s next-generation Vera Rubin chips. The San Francisco-based company announced the agreement on Tuesday and stated that Nvidia is also making an additional significant investment.
According to estimates by Nvidia CEO Jensen Huang, building one gigawatt of AI data center capacity incurs total costs between 50 and 60 billion dollars, with Nvidia products accounting for approximately 35 billion dollars of that sum. The new investment comes in addition to Nvidia’s participation in Thinking Machines’ previous seed financing round of two billion dollars, which valued the company at ten billion dollars. The semiconductor company, with a market capitalization of 4.4 trillion dollars, is thereby continuing its strategy of investing massively in its own major customers.
Platform for Customized AI Systems
Thinking Machines is developing a platform with its main product Tinker that enables companies to adapt and optimize large language models for specific business applications without having to build complex AI training infrastructures themselves. The partnership aims to expand access to advanced AI for companies, research institutions, and the scientific community. The first phase of collaboration is scheduled to begin in early next year when the deployment of Vera Rubin systems starts.
The agreement is part of a series of investments through which Nvidia finances its own customers. This practice has raised concerns about circular financing structures in the industry, as the semiconductor giant channels its enormous cash reserves into its customer ecosystem. A previously announced strategic partnership with OpenAI worth 100 billion dollars was recently abandoned in favor of a capital investment of 30 billion dollars. Huang hinted last week that this might be Nvidia’s last investment in OpenAI before a possible initial public offering.
Personnel Changes at Young Startup
The company, which is only a year old, has undergone significant personnel changes in recent months. Three co-founders and additional researchers have left Thinking Machines. Andrew Tulloch moved to Meta in October, while Barret Zoph, who served as Chief Technology Officer, and Luke Metz returned to OpenAI in January. Murati herself previously worked at OpenAI on the development of ChatGPT, the image generator Dall-E, and the voice mode. She briefly replaced Sam Altman as interim CEO in November 2023 during a board crisis.
As part of the partnership, both companies will work together on developing training and deployment systems for Nvidia architectures. The stated goal is to create powerful AI systems that are understandable, adaptable, and collaboratively usable. The collaboration is intended to create the technological foundation to make advanced AI technology accessible to a broader user base.
