en.Wedoany.com Reported - A startup named Ferveret is applying cooling technology from nuclear reactors to cool chips in artificial intelligence data centers, with its solution improving computing efficiency while consuming zero water. The company was co-founded by Reza Azizian, a former postdoctoral fellow in nuclear engineering at the Massachusetts Institute of Technology (MIT), and Matteo Bucci, an associate professor at MIT. Their adaptive phase-change cooling (APC) solution accelerates heat transfer by generating smaller, more frequently detaching bubbles on server surfaces.
As AI models scale up, cooling energy consumption in data centers is becoming an industry pain point. By the end of this decade, US data centers are expected to consume 9% to 17% of the nation's total electricity, with about one-third of that power used to cool chips running AI models. Ferveret's solution immerses computer servers in a special liquid that absorbs heat far more efficiently than air. What sets it apart from other liquid cooling solutions is that the company's APC technology draws on the subcooled boiling process used in nuclear reactors, employing a liquid with a low boiling point and free of PFAS "forever chemicals." This generates smaller bubbles on chip surfaces, which detach more frequently and rapidly recondense in the surrounding liquid, thereby accelerating heat transfer.
Ferveret has conducted collaborative tests with data center developer and operator CleanSpark, AI accelerator company FuriosaAI, and Switch, one of the largest data center operators in the US. In a recent study conducted jointly with the Samueli Computer Science Department at the University of California, Los Angeles (UCLA), Ferveret's APC solution improved computing power efficiency by 15% compared to state-of-the-art liquid cooling solutions. The company says that when combined with its power control system to optimize operating conditions, data centers can obtain 35% more tokens (small segments of text or data) from their AI models using the same amount of electricity.
Founder Reza Azizian said: "Our goal is to make data centers as sustainable as possible and help them use every watt of electricity to generate tokens, which is the most useful output. Our system can run more powerful chips, help data centers waste less energy, and all of this is achieved with zero water consumption." The company's modular system is delivered in small chassis, each housing a single server, making deployment and maintenance easy. Additionally, Ferveret offers control software that can adjust the power of each server in real time to improve efficiency.
Ferveret is currently part of Nvidia's Inception startup program and is in talks with major cloud computing companies, with plans to announce expanded partnerships later this year. The founder says the technology could help build data centers in sunny but water-scarce remote areas, including parts of Africa, the Middle East, and the US. "The computing industry is facing huge challenges in terms of power availability, and in many regions, there are also issues with water access," Azizian said. "As the industry grows, this will only become more constrained. The main goal of these data center operators is to get more tokens from the power they have. We have proven we can do that."
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
