Frore Systems, a startup building solid-state cooling technology for AI chips, just closed a $143 million funding round at a valuation of approximately $1.64 billion. The deal underscores a reality that the AI industry is only beginning to reckon with: the biggest bottleneck in AI infrastructure is not compute, not data, not algorithms. It is heat.
As AI chips get more powerful and data centers get denser, thermal management has gone from an engineering afterthought to a billion-dollar problem.
Why AI Chip Cooling Matters Now
Modern AI accelerators from Nvidia, AMD, and custom silicon from Google and Amazon generate enormous amounts of heat. A single Nvidia H100 GPU can draw 700 watts of power, and next-generation chips like the Blackwell B200 push even higher. Pack thousands of these into a data center rack, and you have a thermal engineering challenge that traditional air cooling simply cannot solve.
The problem is getting worse, not better. As models get larger and inference workloads scale, power density per rack is increasing rapidly. Data centers that were designed for 10 to 15 kilowatts per rack are now facing demands of 40 to 100 kilowatts per rack. Without better cooling, chips throttle their performance or fail entirely.
What Frore Systems Actually Built
Frore’s core product is called AirJet, a solid-state active cooling chip that replaces traditional fans. AirJet uses vibrating membranes to create high-velocity airflow across heat-generating components. The technology is silent, has no moving parts that wear out, and fits into spaces too small for conventional fans.
The original AirJet was designed for thin laptops and mobile devices, but the company has since scaled the technology for server and data center applications. The key advantage is that AirJet modules can be placed directly on or near the heat source, providing targeted cooling without the overhead of liquid cooling infrastructure.
This matters because while liquid cooling is effective, it requires specialized plumbing, maintenance, and infrastructure that many existing data centers were not designed to support. Frore offers a path to better thermal management that can be retrofitted into existing facilities.
The AI Cooling Market Is Exploding
Frore is not the only company chasing this opportunity. Liquid cooling companies like CoolIT Systems, GRC, and LiquidCool Solutions are growing rapidly. Immersion cooling, where entire servers are submerged in non-conductive fluid, is being deployed by Microsoft, Oracle, and other hyperscalers.
The global data center cooling market is projected to reach $30 billion by 2028, up from roughly $15 billion in 2024. AI is the primary driver of this growth. Every new AI training cluster and inference deployment comes with a cooling challenge that must be solved before a single computation runs.
The Energy Angle
Cooling is also an energy problem. Traditional cooling systems in data centers consume 30 to 40 percent of total facility power. More efficient cooling directly translates to lower energy costs and reduced environmental impact. With AI data centers already drawing scrutiny for their power consumption, better cooling technology is not just a performance issue but an environmental one.
Frore claims AirJet technology can reduce cooling energy consumption by up to 50 percent compared to traditional fan-based systems, though independent verification at data center scale is still limited.
What This Means for the AI Industry
The Frore Systems funding round signals that investors are looking beyond the obvious AI plays of chips, models, and software. The infrastructure layer that makes AI possible, including power, cooling, networking, and physical space, is becoming its own massive investment category.
For anyone building or operating AI infrastructure, thermal management is no longer a checkbox item. It is a strategic priority that directly impacts performance, cost, and the ability to scale.
The Bottom Line
Frore Systems’ $143 million raise at a $1.64 billion valuation is a clear sign that the AI infrastructure buildout is extending well beyond chips and software. As AI models get bigger and inference demand explodes, the companies solving the physical constraints of compute, especially AI chip cooling, are becoming some of the most important players in the ecosystem.