According to Business Insider, Soumith Chintala, the former Meta AI leader who co-created PyTorch, has joined former OpenAI CTO Mira Murati’s startup Thinking Machines Lab. Chintala announced his move this week after leaving Meta earlier this month following an 11-year tenure where he helped build PyTorch from scratch into the dominant AI framework used by virtually every major AI company. The startup has been aggressively hiring top researchers from Meta, OpenAI, and Anthropic while offering technical salaries as high as $500,000 before bonuses and equity. Thinking Machines raised a $2 billion seed round at a $10 billion valuation earlier this year and is now reportedly in talks to raise funding at a staggering $50 billion valuation.
The talent wars escalate
This isn’t just another executive shuffle—it’s a major signal about where AI talent sees the real action happening. Chintala built the infrastructure that powers modern AI development. And he’s choosing to join a startup that‘s barely launched products over staying at Meta, which has essentially infinite resources. That tells you something about where the excitement is right now.
Thinking Machines is basically assembling the AI Avengers. They’ve got John Schulman (ChatGPT co-lead), Bob McGrew (former OpenAI research chief), and now the guy who built the framework everyone uses. When you’re offering half-million-dollar base salaries and competing for the absolute top tier, you’re playing in a completely different league than most startups.
Meta’s talent exodus continues
Here’s the thing: Meta’s AI brain drain is becoming a real pattern. First Yann LeCun reportedly preparing to leave, now Chintala—these aren’t replaceable mid-level engineers. These are the architects. Meta’s trying to counter by hiring from OpenAI and Google DeepMind into their new Superintelligence Labs division, but losing your foundational people is different than acquiring new ones.
Think about it—PyTorch was born at Facebook Research. Chintala watched it grow from an internal tool to the backbone of AI development worldwide. If he feels like the future isn’t at Meta anymore, that should worry anyone invested in the company’s AI ambitions. The infrastructure expertise he’s taking to Thinking Machines is literally priceless.
The startup valuation craze
A $50 billion valuation for a company that hasn’t even launched its full product suite? That’s absolutely wild. We’re talking about a startup going from $10 billion to $50 billion in valuation within months, based largely on the team they’re assembling and the promise of “human-AI collaboration.”
But honestly, in today’s AI market, does anyone even blink at these numbers anymore? When you’ve got the former OpenAI CTO plus the creator of PyTorch plus the ChatGPT leads all in one place, investors will throw money at the problem. The real question is whether Thinking Machines can deliver products that justify that insane valuation before the AI hype cycle cools down.
What’s actually being built?
We know they have Tinker, their first tool for fine-tuning LLMs, being used at Princeton and Stanford. But the big vision seems to be about human-AI collaboration—whatever that actually means in practice. With this level of talent, expectations are sky-high.
Chintala said he wanted “something small… something new… something uncomfortable” after leaving Meta. Well, joining a startup aiming for a $50 billion valuation with all the pressure that brings? That definitely qualifies as uncomfortable. The entire AI world will be watching to see if this dream team can actually build the next big thing in AI, or if this becomes another case of too much hype and too little product.
