Karman Industries has launched a carbon dioxide-based heat processing unit (HPU) which it says can cool data centers and recycle waste heat.

The modular 10MW CO2 cooling system can consolidate “massive heat management infrastructure into high-density modular packages,” the vendor said.

According to Karman, HPUs deployed in an AI data center can take high-temperature waste heat from GPU clusters and turn it into cooling fluid at less than 30ºC without the need for large amounts of water.

In cool ambient conditions, HPUs can process waste heat from compute racks to provide electricity for operations or high-grade heat for district heating and other use cases.

CO2 cooling works in largely the same way as a traditional air cooling system, except that the gas used is carbon dioxide rather than an F-Gas-based refrigerant, which can be highly damaging for the planet. Some vendors are pursuing CO2 as an environmentally-friendly alternative to traditional cooling, and DCD took an in-depth look at this emerging market in 2024.

Karman’s system uses a proprietary technology the firm calls the “Kompressor,” which it says can save large amounts of space compared to traditional air cooling systems, reducing the cooling system footprint by up to 80 percent.

“We applied an aerospace systems-engineering approach to data center thermodynamics,” said CJ Kalra, CTO and co-founder of Karman Industries. “Our team designed HPUs to process the extreme heat of gigascale racks using a first-principles-based approach.

“HPUs enable heat reuse and PUE ratings quickly approaching one without water consumption or PFAS chemicals for AI factories. Keeping up with the latest in technology, the HPU leverages 800V DC architecture while borrowing the latest innovations in rocket turbomachinery like metal 3D-printing, and electric vehicles innovations like high-speed motors and silicon-carbide power electronics technology."

Los Angeles-based Karman raised $20 million in a Series A funding round led by Riot Ventures in September 2025, bringing the total funding the company has taken to over $30 million. Other major participants included Sunflower Capital, Space VC, Wonder Ventures, and former Intel and VMware CEO Pat Gelsinger.

Though it hasn’t disclosed any customers, it says it will begin deliveries of its system to clients in Q3 2026. The company has a manufacturing factory, known as GigaWerx, in Los Angeles, which it says can handle 1GW of annual production, with the potential to scale to 4GW.

“In the race to stand up AI capacity, time is the most expensive variable,” said David Tearse, CEO and co-founder of Karman Industries. “We’ve moved beyond the era of legacy chillers to HPUs. By shrinking the footprint of the mechanical yard by 80 percent, we don’t just save land; we eliminate the ‘snowball effect’ of infrastructure complexity, allowing hyperscalers to move from 'shovels in the ground' to 'chips in the rack' many months faster while unlocking additional compute.”