Analysis
Look at where the biggest single checks went this week and a pattern jumps out immediately: both of the two largest venture rounds funded physical constraints on AI growth, not AI applications. Sequoia led a $1 billion Series B into Valar Atomics, tripling the nuclear-reactor startup's valuation to $6 billion to fund mass production of factory-built microreactors for data centers. Days later, London's OLIX raised $312 million at a $3.3 billion valuation to build inference chips that sidestep the High-Bandwidth Memory shortage currently rationing the entire GPU supply chain.
Two Rounds, One Thesis
Neither round is really about AI models. Valar is a power company; OLIX is a materials-and-packaging bet. Both are underwriting the theory that the actual bottleneck on the next several years of AI progress isn't smarter models -- it's electricity and memory, two inputs no software startup can conjure with a better prompt.
“## Two Rounds, One Thesis Neither round is really about AI models.”
This isn't a one-week blip. Venture allocation broadly has been rotating toward 'hard' infrastructure and deep tech -- energy, chips, biomanufacturing, defense hardware -- for several quarters now, a reversal from the 2021-2023 cycle when software-multiple economics justified funding nearly anything with a subscription model and an API.
For GPs, the implication is that the highest-conviction AI-adjacent bets increasingly require domain expertise venture firms didn't need five years ago -- nuclear engineering, materials science, semiconductor packaging -- rather than just an ability to evaluate a product demo and a growth curve.
What to watch: whether this rotation pulls capital away from application-layer AI funding rounds over the next two quarters, or whether the two pools of capital are simply expanding in parallel as the overall AI investment total keeps growing.