Mira Murati‘s Thinking Machines Lab has signed a multibillion-dollar cloud infrastructure agreement with Google, even as the 14-month-old AI startup faces a relentless talent drain from Meta. The deal, announced at Google Cloud Next in Las Vegas on April 22, signals that Murati is moving to fortify her startup’s technical foundation regardless of who is walking out the door.
Under the agreement, Thinking Machines Lab will expand its use of Google Cloud’s AI infrastructure, including systems powered by Nvidia’s latest GB300 GPUs, with the deal valued in the single-digit billions. The GB300 chips can double training speeds compared to previous GPU generations, and the deal makes Thinking Machines Lab one of the first startups to access these systems.
The timing matters. The contract makes Thinking Machines Lab the third frontier AI developer lining up for Google’s Blackwell and TPU capacity this month, behind Anthropic and Meta. In a sector where compute access is increasingly a competitive moat, being in that cohort matters — especially for a company still in early-stage model development.
What Tinker needs this infrastructure for
Wednesday’s deal provided some insight into what Thinking Machines is building. Google noted in a press release that it can support the startup’s reinforcement learning workloads, which Tinker’s architecture relies on. Tinker, launched in October 2025, is a tool that automates the creation of custom frontier AI models — the startup’s first commercial product after nearly a year of secrecy following its founding.
The deal is not exclusive, so Thinking Machines may continue to use other cloud providers over time. Earlier this year, the lab had also partnered with Nvidia in a separate deal that included an investment from the chipmaker.
Also Read: Mira Murati, Ex-CTO of OpenAI launches AI startup Thinking Machines Lab
A company building under pressure
The Google deal arrives as Thinking Machines Lab contends with a sustained talent exodus. Of the startup’s original founding group, five have gone to Meta, three have returned to OpenAI, and one has joined Elon Musk’s xAI. The Next Web Meta’s recruitment push followed Mark Zuckerberg‘s reported attempt to acquire the startup outright — an offer Murati declined.
Despite that churn, the company has taken active steps to rebuild. It brought in Soumith Chintala, creator of PyTorch, as CTO, and has expanded to roughly 130 employees within a year of its launch. The Google Cloud deal now adds a critical compute layer to that rebuilding effort.
Thinking Machines Lab was reportedly in talks for a new funding round at a $50 billion valuation by November 2025 — a figure that, if realised, would rank it among the most valuable private AI companies in the world. Whether the infrastructure bet and the fundraising narrative can survive the ongoing attrition at the top is the question that now hangs over the startup.
