Ex-OpenAI startup eyes $4B raise

Frontier AI fundraising has stopped behaving like normal startup finance. A lab can barely get incorporated, hire a few researchers, and sketch a thesis — and investors are already discussing multibillion-dollar valuations. That is the real news here. The specific company matters, but the bigger shift is that ex-OpenAI talent has become its own asset class, and money is moving at a speed that would have looked absurd even a year ago. (theinformation.com) ### What changed this week? A report from The Information says a startup founded by former OpenAI researchers, only six weeks old, is seeking funding at roughly a $4 billion valuation. The piece ties that jump to a market where investors are no longer waiting for product launches, revenue, or even much public detail before making huge bets on teams they think can build the next OpenAI or Anthropic. (theinformation.com) ### Why would investors pay up that fast? Because in frontier AI, investors are buying people first and plans second. The logic is simple — if a small group of elite researchers has already helped train or ship major models, backers assume that talent is scarce enough to justify extreme prices before the compan(theinformation.com)Ilya Sutskever’s Safe Superintelligence. (techcrunch.com) ### Why is Nvidia part of this story? Nvidia is not just selling chips into the boom. It has also been investing directly in AI startups, which helps validate young companies and pulls traditional venture firms into the same deals. TechCrunch, citing PitchBook data, says Nvidia participated in nearly 67 venture deals in 2025, more th(techcrunch.com)e access and strategic backing may come bundled together. (techcrunch.com) ### Is this only about chatbots? No — and that is an important twist. Another former OpenAI researcher, Liam Fedus, is raising for Periodic Labs, which is trying to build “AI scientists” and autonomous laboratories for physics and chemistry. Forbes says that company is pursuing a $500 million round at a $7.5 billion valuation, after emerging (techcrunch.com) into AI-for-science bets that need both models and physical experimentation. (forbes.com) ### Where does DeepSeek fit in? DeepSeek shows the same capital frenzy on the China side — but with a state angle layered on top. Reuters reported this week that DeepSeek could be valued at up to $50 billion in its first fundraising round, with China’s state-backed semiconductor investment arm expected to lead and Tencent also in talks. SCMP then said the round could close soon at as much as that same valuation. (money.usnews.com) ### Why does that matter beyond one company? Because it turns frontier AI into a strategic industry, not just a hot software category. In the U.S., the money is coming from VCs, hyperscalers, and Nvidia’s ecosystem. In China, the same race is starting to look more explicitly national-champion driven. Once rounds get this large, they do more than fund payroll — they buy compute, data-center access, researchers, and time. (theinformation.com) ### What’s the catch? Valuation is not capability. A famous founding team can raise billions and still fail to train a differentiated model, ship a product, or build a durable business. But the market clearly believes the downside of missing the next breakout lab is worse than the downside of overpaying early. That is why six-week-old companies can now command prices that used to require years of proof. (theinformation.com) ### Bottom line? This story is really about a new rule in AI: capital now chases elite research talent almost instantly. And with DeepSeek’s round rising in parallel, the bidding war is starting to look global. (theinformation.com)