Humans&, a three-month-old AI lab, just raised $480 million at a $4.48 billion valuation. The company has no product. Its pitch: the founders used to work at Anthropic, xAI, and Google.
This is not an outlier. Over half of all US seed and Series A funding in 2026 has gone to rounds of $100 million or more. The seed stage - traditionally where companies raise a few million to build their first product - has become a place where investors pay billions for talent and a thesis.
The Numbers
The concentration of early-stage capital in mega-rounds keeps accelerating:
- Humans& raised $480 million at seed with no product
- Merge Labs (Sam Altman’s brain-computer interface startup) raised $252 million at seed
- Ricursive Intelligence raised $300 million in Series A at a $4 billion valuation
- Inferact raised $150 million at seed, reaching an $800 million valuation months after founding
These aren’t exceptions. Over 40% of global seed and Series A investment in 2026 has gone to rounds exceeding $100 million.
What Investors Are Buying
In the case of Humans&, the answer is clear: the team. Co-founders include Andi Peng from Anthropic, Eric Zelikman from xAI, Georges Harik (an early Google employee), plus researchers from OpenAI and Meta. The investors include Nvidia, Jeff Bezos, Google Ventures, and SV Angel.
According to co-founder Peng, most of the capital will go to compute for training models. The company’s stated mission - AI tools where “collaboration and human insight remain central” - is broad enough to mean almost anything.
The investors aren’t betting on a product. They’re betting that these specific people, given enough compute, will build something valuable. In an industry where talent concentration matters more than traditional moats, that bet makes a certain kind of sense.
Why This Is Happening
Three dynamics are driving the seed-stage gold rush:
First, the AI talent pool at the frontier is small. There are only a few thousand people in the world who have trained models at the scale of GPT-4 or Claude. When those people leave to start companies, investors compete aggressively to back them.
Second, compute is expensive. Training a frontier model can cost hundreds of millions of dollars. A $480 million seed round isn’t excessive if you’re planning to train competitive models - it’s table stakes.
Third, investors fear missing the next OpenAI or Anthropic. Those companies reached multi-hundred-billion-dollar valuations within a few years. The cost of being wrong on a seed investment is capped at the investment amount. The cost of missing a winner is unlimited.
The Risks
This funding pattern creates obvious problems.
For founders who aren’t ex-Anthropic researchers, raising capital has gotten harder. When half of seed funding goes to mega-rounds led by proven teams, less capital remains for first-time founders building products the traditional way.
For the mega-round companies themselves, the pressure is immense. A $4.48 billion seed valuation means investors expect eventual returns in the tens of billions. That’s a high bar for a company that hasn’t shipped anything.
For the AI industry broadly, this concentration of capital in a handful of pre-product companies may not be efficient. Some of those billions would likely generate more innovation if spread across hundreds of smaller bets.
Who Wins, Who Loses
The winners are obvious: well-credentialed AI researchers can now command venture backing at levels that were unthinkable five years ago. The pedigree premium has never been higher.
The losers are less visible. Founders without elite AI lab backgrounds compete for a shrinking pool of capital. Investors who want to back early-stage companies at reasonable valuations find themselves priced out. And the companies raising these mega-rounds face expectations that most will never meet.
The AI funding market has decided that in this industry, at this moment, the right team is worth billions before they’ve built anything. Whether that bet pays off will take years to know.