Thinking Machines Lab has raised billions in capital while assembling one of AI's most elite technical teams. Now it's watching some of that talent walk out the door.

Loading audio narration...

As Big Tech rivals dangle eye-popping pay packages and early employees unlock their first slice of equity, the buzzy, 1-year-old startup has become ground zero for tech's escalating talent wars.

Nearly a third of its 42-person founding team — 13 people — have left since its launch, a Business Insider review of LinkedIn profiles and conversations with four sources found. Those departures include three of its six co-founders.

Thinking Machines Lab's head count has more than quadrupled to over 150 people since it launched, a person familiar with the matter told Business Insider.

The departures have been driven by a mix of factors, though a major one is the huge packages offered by rival AI labs like Meta and OpenAI, as well as the passage of an important milestone in startup compensation: the one-year cliff, four sources familiar with the matter said.

It's not unheard of for early members of a startup to jump ship over time, though Thinking Machines Lab was supposed to be different. It emerged from stealth last year as one of the highest-profile AI labs thanks to its leadership's deep ties to OpenAI. CEO Mira Murati previously served as OpenAI's chief technical officer, while other key figures helped train the early versions of ChatGPT.

Investors eager to back the next breakout AI company flocked to the startup, which raised $2 billion before launching a product.

The startup's reputation as a talent hub has put a target on its back.

Thinking Machines Lab declined to comment.

An opportunity they couldn't refuse

Some of the offers to founding members have reached well into nine figures — hundreds of millions of dollars in cash and stock over several years — according to one defector who shared notes with peers.

"I got an opportunity that I couldn't turn down," the person said.

Meta — which considered buying Thinking Machines Lab last year, The Verge reported — has been the most aggressive poacher, quietly luring away seven founding team members, plus a star AI researcher.

Such was the case of veteran software engineer Joshua Gross. Last year, Gross was deep in the trenches of AI development at Thinking Machines Labs, helping launch the flagship product, Tinker.

Gross "had a blast" shaping Tinker's early vision, he wrote in a LinkedIn post at the time. After helping build and ship the product from "zero to one," he's no longer at Thinking Machines Lab. Wired first reported on details of the compensation packages.

Subject line: $1.5 million cash 'and up'

The race for elite AI talent has become hotter than ever across tech, said Sam Agre, cofounder of recruiting firm People In AI.

While the pool of people with hands-on experience building leading AI models and products is small, it runs deep at Thinking Machines Lab. It's so difficult to poach rank-and-file workers from there that Agre said he has resorted to sending them LinkedIn messages with subject lines touting $1.5 million in cash compensation "and up."

That's more than three times the annual salary for software engineers offered by Thinking Machines Labs on its careers page, which mentions a range of $350,000 and $475,000 a year.

While the nine-figure offers are reminiscent of deals for major athletes, Big Tech firms like Meta covet people who can help them pull ahead in AI, Agre said.

"You don't want to fall behind in that type of arms race," he said.

Meta isn't the only tech giant circling Thinking Machines Lab. OpenAI has hired five founding members, and Elon Musk's xAI has poached one more. OpenAI and Meta declined to comment, and xAI didn't respond to a request for comment.

One-year cliffs opened the floodgates

Meta first began trying to lure away Thinking Machines Lab talent when it launched its superintelligence team last year. It scored a major coup by bringing on cofounder Andrew Tulloch. OpenAI brought on two more of Murati's cofounders, CTO Barret Zoph and researcher Luke Metz.

Then this year, the floodgates opened.

Because the company launched a little over a year ago, many founding employees started hitting their "one-year cliffs" — a standard startup threshold that unlocks the first chunk of equity after 12 months, after which it vests monthly.

Once they get their shares, employees can leave without walking away empty-handed, making it a natural moment for Thinking Machines Lab staffers to entertain outside offers, according to three people familiar with the matter.

One-year cliffs are supposed to help retain talent, though they have become something of an Achilles' heel for startups ever since the AI boom began, said Dan Walter, an independent compensation consultant. He told Business Insider that retention rates in tech startups are the lowest he's ever seen due to the ease of building a startup and rampant poaching.

"The entire model is being challenged," he said.

Some in the compensation industry are now arguing for the cliff to start at five years, not one, to lock in early team members, he added.

Thinking Machines Lab still attracts top talent

Despite the rampant poaching, Thinking Machines Lab has scored wins.

Its CTO — Soumith Chintala, the creator of the open-source AI project PyTorch — left Meta to join Murati's startup. Thinking Machines Lab regularly brings on research talent from Meta, including Kenny Yu, a member of the tech giant's hotshot TBD lab — a small team of AI all-stars launched for its superintelligence push.

Murati's startup has been putting that talent to use. On Monday, it announced that it had built a new type of AI model that interacts with people seamlessly, handles interruptions, and translates languages in real time. The model will launch more widely later this year.

The company is also taking steps to stem the losses. It's hiring a person to build a fresh framework for doling out equity and to implement systems that "attract and retain highly sought-after talent," a job listing says. The position, which pays between $250,000 and $425,000 in base salary, was posted in March.

For all the departures, Murati's reputation as a talent magnet hasn't faded.

When Business Insider asked Carina Hong, CEO and founder of Axiom Math — already valued at $1.6 billion — which company tries to poach most of her researchers, she didn't hesitate: Thinking Machines Lab.