Meta and Thinking Machines Are Playing Musical Chairs With AI Talent

Talent wars cut both ways.

That’s the uncomfortable truth sitting at the center of the Meta vs. Thinking Machines Lab story right now. Yes, Meta has been pulling researchers out of TML. But TML has been doing the same thing right back — and while Meta is a $1.4 trillion company that can absorb the hits, a startup landing multibillion-dollar cloud deals with Google can afford to play offense too.

This isn’t a story about one company winning and one losing. It’s a story about how the AI talent market has gotten so hot that even the biggest players can’t hold onto everyone, and how a well-funded startup with the right founding story can punch well above its weight.

Who Is Thinking Machines Lab, Exactly?

If you haven’t been tracking TML closely, here’s the short version: it’s the $12 billion AI startup founded by Mira Murati, who most people know as OpenAI’s former CTO. That founding pedigree matters enormously in this space. When someone with Murati’s track record starts a company, researchers pay attention. Equity upside at a $12 billion valuation is a very different conversation than equity at a company that’s already gone public or plateaued.

That valuation, combined with the Google cloud deal — which reportedly gives TML access to Nvidia’s latest GB300 chips — puts the startup in a genuinely strong position. Getting early access to that kind of compute infrastructure isn’t just a nice-to-have. It’s a signal that TML is being taken seriously at the infrastructure level, which is where serious AI development actually happens.

Meta’s Poaching Problem Is Also TML’s Poaching Problem

The talent flow between these two organizations is worth examining honestly. Meta has pulled names from TML — including Mark Jen and Yinghai Lu, two researchers who made the move recently. That stings for any startup. Losing experienced people to a company with Meta’s resources and scale is a real cost, not just a PR inconvenience.

But the reverse is also true. TML has been pulling talent away from Meta. Researchers are leaving a company with enormous stability and resources to join a startup, which tells you something about what people in this field are actually optimizing for right now. It’s not just salary. It’s proximity to the frontier, equity upside, and the chance to work on something that feels like it’s still being defined.

Meta, for all its investment in AI, is still a social media company at its core. That framing matters to researchers who want their work to feel like it’s shaping the direction of the field rather than optimizing ad delivery or content ranking. TML, built from scratch by someone who was inside OpenAI during its most consequential years, carries a different kind of energy.

What the Google Deal Actually Signals

The multibillion-dollar cloud deal with Google is the detail that changes the shape of this story. Startups at TML’s stage often struggle with compute access — it’s one of the most real constraints in AI development, and it’s one that large incumbents use as a structural advantage. If you can’t get the chips, you can’t train the models.

TML apparently can get the chips. GB300 access puts them on a short list of organizations with the infrastructure to do serious frontier work. That’s not a minor footnote — it’s the kind of deal that makes a recruiting pitch much easier to deliver. When you’re trying to convince a top researcher to leave a stable job at a big tech company, “we have the compute to actually build what we’re talking about” is a strong argument.

Google’s willingness to sign that deal also says something about how the broader market views TML’s trajectory. Cloud providers don’t hand out multibillion-dollar agreements to companies they think are going to flame out.

My Take

From where I sit reviewing AI tools and the companies building them, the Meta-TML talent exchange looks less like a crisis for either side and more like a healthy, if chaotic, sign of a maturing market. The best researchers have options. They’re using them. Some go to Meta for scale and stability. Some go to TML for upside and mission.

What I’d watch is whether TML can convert its compute access and founding credibility into actual products that ship and hold up under scrutiny. Valuations and cloud deals are inputs. What comes out the other end is what matters. TML has the pieces. Now it has to build something worth reviewing.

And when it does, we’ll be here with the no-BS take.

🕒 Published: April 24, 2026