While everyone obsesses over who builds the smartest AI, Frore just raised $143 million because someone has to keep the damn things from melting.
The Signal
Frore Systems hit a $1.64 billion valuation making liquid cooling tech for AI chips. Not the chips themselves. The plumbing. This tells you three things about where we actually are in the AI infrastructure buildout.
First, the thermal wall is real. Modern AI accelerators push 700+ watts through a postcard-sized chip. Air cooling topped out around 300 watts. Liquid cooling isn't a nice-to-have anymore, it's the prerequisite for running frontier models at scale. Frore's tech matters because it works at the chip level, not just the data center level.
Second, Fidelity backing a cooling company signals institutional money finally understands infrastructure. Not the sexy stuff. The unglamorous physics problems that determine whether your AI agents actually run or throttle themselves into uselessness. Data centers are redesigning around thermal management. Cooling is now a primary cost and performance constraint.
Third, $1.64 billion for a cooling startup means the market expects AI compute to 10x from here, minimum. Nobody pays that multiple betting on steady state. They're betting on orders of magnitude more inference, more agent workloads, more always-on AI systems generating heat 24/7. Frore is selling shovels, and the gold rush math just got real.
The Implication
Watch the infrastructure layer. The companies solving physics problems, not just software problems, will capture more value than most people expect. If you're building AI agents or deploying models, your thermal constraints matter more than your token efficiency. Plan accordingly.
Source: Bloomberg Tech