Women Building the AI Era, Part 1- Mira Murati

In September 2024, Mira Murati sent an email to OpenAI's staff saying she needed "time and space" to explore what came next. She had no company, no funding, and no product plan that anyone outside her closest circle had seen.

Ten months later, investors reportedly valued whatever she was building at $12 billion. Nvidia, AMD, and a clutch of Silicon Valley's biggest venture firms had signed checks totaling $2 billion, one of the largest seed rounds in the industry's history, for a company that had not shipped a single thing.

What Thinking Machines Lab was actually selling wasn't a product. It was a bet that the people who built the last breakthrough are the safest bet on building the next one, and in the current AI economy, that bet is worth billions before anyone can check the math.

The Six-Month Valuation

Murati unveiled Thinking Machines Lab in February 2025, five months after leaving OpenAI. The founding team numbered around thirty people, and roughly two-thirds of them had worked at OpenAI before. John Schulman, an OpenAI co-founder, signed on as chief scientist. Barret Zoph, formerly OpenAI's vice president of research, became chief technology officer.

Murati herself had spent nearly four years running product and engineering at OpenAI, a run that included the launches of ChatGPT, DALL-E, and Codex, the kind of resume line investors treat as a track record even before a new company has built anything of its own.

By the summer, the fundraising conversations that had started around a $10 billion valuation closed considerably higher. The round, led by Andreessen Horowitz, landed at $2 billion raised and a reported $12 billion valuation, with Nvidia, AMD, Cisco, ServiceNow, Accel, and Jane Street all writing checks. It was one of the fastest jumps from founding to double-digit-billion valuation the AI industry had seen, reached months before the company had anything a customer could use.

None of them had a shipped product to evaluate. What they had was a roster.

Did You Know?

Thinking Machines Lab wasn't the only pre-product AI startup commanding a startling price tag that year. Safe Superintelligence, the AI lab Ilya Sutskever started after leaving his post as OpenAI's chief scientist, raised its own $2 billion round around the same period at a reported $32 billion valuation, also with no public product. Both founders had reportedly turned down acquisition interest from Meta before raising instead.

What Investors Were Actually Buying

Nobody wrote a check into Thinking Machines Lab because of a business plan. They wrote it because of who was in the room: engineers who had shipped ChatGPT and DALL-E, led by a founder who had run product and engineering at the company that started the current AI race. In a market this competitive, that kind of team doesn't come up for sale often.

It works something like buying a racehorse before it has run a single race. You can't grade its speed yet, so you price its bloodline instead, on the theory that whoever bred the last two winners is likely to breed the next one too. The wider AI industry has been financing itself on similar bets on the future for a while now, which is part of why bond investors have grown nervous about how much of the AI boom rests on debt against assets that don't exist yet.

The pattern repeats across this series: when the next breakthrough could come from anywhere, the fastest hedge is buying the people most likely to have already seen it coming.

The Product Nobody Expected

Thinking Machines Lab didn't ship a chatbot. Its first release, in October 2025, was Tinker: an API that lets developers fine-tune large language models without managing the distributed computing infrastructure that normally requires. It was a tool for other builders, not a consumer product, a bet on becoming plumbing rather than a storefront.

For a founder whose old job was shipping one of the world's best-known chatbots, choosing an unglamorous developer tool as her first release said almost as much as the product itself.

The company followed that in May 2026 with TML-Interaction-Small, the first release in a new "Interaction Models" line built for real-time audio, video, and text. Executives have said more of the company's own models are coming later in the year. That infrastructure-first instinct echoes the same logic that turned a graphics card company into the hardware backbone the entire AI industry now runs on: the money isn't only in the flashiest product, it's in the layer everyone else has to build on top of.

Interesting Fact: Nvidia backed both Murati's Thinking Machines Lab and Sutskever's Safe Superintelligence in the same period, alongside its own chip business supplying both. A single investor hedging across multiple, competing labs while also selling them the hardware they run on is now a routine feature of how the AI industry finances itself.

The Cost of Being Priced This High

A $12 billion valuation buys attention, but it also buys pressure. Thinking Machines Lab's headcount reportedly grew past 150 within a year of launch, more than quadrupling its founding team. It also lost people just as fast: roughly 13 members of that original team left within about a year, including three of the company's six co-founders, pulled away by pay packages from rival labs that reportedly ran into the hundreds of millions of dollars. Two of the departing co-founders, Barret Zoph and Luke Metz, went back to OpenAI in January 2026. Another, Andrew Tulloch, moved to Meta on a package reported at roughly $1.5 billion over six years.

That churn is the quieter half of the story. A valuation set before a product exists is also a bet the market can revise just as quickly, once there's finally something to measure it against, and in an industry moving this fast, that reckoning may come sooner than investors expect.

The market has already tested that theory once. In November 2025, Thinking Machines Lab reportedly began talks to raise a new round at a $50 billion valuation, more than four times its seed price. By January 2026, those talks had collapsed without a deal. Prospective backers, by most accounts, couldn't square a number that high against a company that still had little more than one developer tool and a handful of months of runway to show for it. The mark held at $12 billion instead, a reminder that pedigree buys a company its first price tag, but not necessarily its next one.

Thinking Machines Lab is a useful place to watch how the AI economy actually allocates capital, because it strips the decision down to its rawest form. There was no revenue to model, no customer base to survey, no working product to test against a competitor. There was only a list of names and the record they had built somewhere else.

That's not unique to Murati's company. It's the logic behind Safe Superintelligence's reported $32 billion valuation too, and it rhymes with the wider AI industry's habit of financing infrastructure against projected returns rather than proven ones. What's different here is how visible the mechanism is: pedigree, converted directly into a price tag, with the product left to arrive later and prove the number right or wrong.

The safest bet in this market still isn't a product. It's a familiar face, priced in the billions before anyone can check the math.

Sources & References

• Mira Murati's Thinking Machines Lab is worth $12B in seed round, TechCrunch (2025)
• Mira Murati's $2 billion seed round, Thinking Machines Lab, OpenAI, and female founders, Fortune (2025)
• Former OpenAI CTO Mira Murati unveils new AI startup with a leadership team stacked with former OpenAI colleagues, Fortune (2025)
• Thinking Machines' first official product is here: meet Tinker, VentureBeat (2025)
• Ilya Sutskever's Safe Superintelligence raises $2B at $32B valuation, with no product yet, CTech / Calcalist (2025)
• Mira Murati's startup Thinking Machines Lab is losing two of its co-founders to OpenAI, TechCrunch (2026)