Thinking Machines Lab Intensifies Talent War with Meta Amid $12 Billion Valuation and Google Cloud Deal

Thinking Machines Lab recruits Meta veterans and PyTorch co-founder Soumith Chintala, leveraging a $12B valuation and a massive Google Cloud hardware deal.

By: AXL Media

Published: Apr 26, 2026, 7:37 AM EDT

Source: Information for this report was sourced from TechCrunch

The Great AI Talent Reallocation

A high-stakes recruitment battle has intensified between Meta and Thinking Machines Lab (TML), as the startup successfully lures long-term veterans from the social media giant’s premier research divisions. Weiyao Wang, an eight-year Meta veteran who contributed to the SAM3D open-world segmentation project, is the latest to join TML’s expanding roster. This move follows a period of fluid talent movement where Meta previously attempted to acquire TML and has recently poached seven of the startup’s founding members. Despite these losses, TML has responded by raiding Meta’s research talent pool more than any other single employer, signaling a fierce rivalry for the specialized workforce required to build next-generation perception systems.

Acquiring the Architects of Modern AI

The most significant blow to Meta’s talent retention is the departure of Soumith Chintala, who now serves as TML’s Chief Technology Officer. Chintala spent over a decade at Meta and is widely recognized as the co-founder of PyTorch, the open-source framework that currently underpins the majority of global artificial intelligence research. Joining him is Piotr Dollár, another 11-year Meta veteran and former research director who co-authored the influential Segment Anything Model. The migration of such high-level leadership suggests that TML is positioning itself not just as a competitor, but as a primary destination for the architects of the foundational tools currently used by the entire industry.

Infrastructure Parity via Multibillion Dollar Cloud Deal

To support its growing team of 140 researchers, Thinking Machines Lab has entered into a massive cloud partnership with Google. Announced at Google Cloud Next, the deal provides TML with access to Nvidia’s latest GB300 hardware, placing the startup on the same infrastructure tier as Meta and Anthropic. This agreement follows an earlier collaboration with Nvidia and ensures that TML has the computational power necessary to train large-scale multimodal language models. Access to this cutting-edge hardware is a critical component of TML's strategy to maintain a competitive edge against better-funded incumbents.