The talent war between Meta and Thinking Machines Lab just got more interesting. Weiyao Wang, an eight-year Meta veteran who helped build multimodal perception systems and worked on SAM3D, walked out the door last week and joined Thinking Machines Lab (TML), according to TechCrunch AI. He’s the latest in a string of researchers swapping Meta’s seven-figure pay packages for a $12 billion startup that’s released exactly one product.
What stands out here is how aggressively the talent flow runs both ways.
The Google Cloud deal that changes the math
TML just signed a multibillion-dollar cloud agreement with Google, announced this past Tuesday at Google Cloud Next. The deal gives the startup access to Nvidia’s latest GB300 chips and makes it one of the first companies to run on that hardware. TechCrunch AI notes the agreement puts TML in the same infrastructure tier as Anthropic and Meta itself.
That’s a serious upgrade for a company barely two years old, and it follows an earlier Nvidia partnership.
Meta’s poaching, and TML’s counter-raid
Business Insider reported last week that Meta has poached seven of TML’s founding members. Meta even held acquisition talks with TML around this time last year before pivoting to picking off founders one by one.
But TML isn’t sitting still. Based on a LinkedIn review reported by TechCrunch AI, the startup has been hiring more researchers from Meta than from any other single employer. The notable names:
- Soumith Chintala: 11-year Meta veteran, PyTorch co-founder, now TML’s CTO
- Piotr Dollár: another 11-year Meta vet, co-author of Segment Anything, now on technical staff
- Andrea Madotto: FAIR research scientist on multimodal LLMs, joined in December
- James Sun: nearly nine years at Meta on LLM pre- and post-training
- Kenneth Li: Harvard PhD, joined this month after 10 months at Meta
TML has also drawn talent from elsewhere. Neal Wu, a three-time IOI gold medalist and former Cognition founding member, joined early this year. Jeffrey Tao came via Waymo, Windsurf, and OpenAI. Muhammad Maaz left an Anthropic research fellowship. Erik Wijmans arrived from Apple. Liliang Ren spent over two years on Microsoft’s AI Superintelligence team pre-training OpenAI’s coding models. Headcount now sits around 140.
Why researchers pick $12B over Meta’s millions
Meta’s pay packages are well documented. Seven figures, no strings attached. So why are top researchers walking away?
The math, per TechCrunch AI’s read, is simpler than it looks. TML is currently valued at $12 billion. That number would’ve seemed absurd for a one-product company in any previous tech cycle. But measured against OpenAI and Anthropic’s record-breaking valuations, there’s still real equity upside on the table.
For a researcher choosing between a guaranteed paycheck and a stake in what could become a frontier-lab giant, the upside calculation tips toward the startup faster than you’d expect.
What this signals for the AI industry
The Meta-TML talent flow tells us a few things worth tracking:
- Compute access matters as much as cash: TML’s Google and Nvidia deals give researchers reasons to stay that Meta’s checks can’t replicate alone.
- The frontier-lab tier is consolidating: TML now sits beside OpenAI, Anthropic, and Meta on the chips, talent, and capital fronts.
- PyTorch’s center of gravity is shifting: with Chintala, Dollár, and other foundational Meta AI researchers at TML, expect the open source ecosystem they shaped to drift with them.
A TML spokesperson declined to comment to TechCrunch AI. Expect more high-profile moves in both directions before the dust settles. Full breakdown is available at the original source.