Become a TechRadar Insider
- Tiny bubbles could significantly reduce cooling demands inside AI facilities
- Researchers adapted nuclear reactor science for modern computing infrastructure
- Ferveret claims 15% efficiency gains over existing liquid cooling
Artificial intelligence is driving a rapid expansion of computing infrastructure, creating fresh concerns about electricity consumption and long-term sustainability.
Industry estimates suggest data centers could account for between 9% and 17% of total United States electricity usage by the end of this decade.
Roughly one-third of that power currently goes toward cooling the processors that run AI tools and other demanding workloads.
Nuclear reactor principles find a new role in data center cooling
Now, startup Ferveret believes a technology adapted from nuclear reactor research could significantly reduce the energy required to cool modern computing systems.
Founded by former MIT postdoctoral researcher Reza Azizian and MIT professor Matteo Bucci, the company developed a cooling approach called Adaptive Phase Cooling, or APC.
Rather than relying on traditional fans, the system submerges servers inside a specialized liquid that removes heat more effectively than air.
The distinguishing feature involves the formation of very small bubbles on chip surfaces during operation.
According to the founders, those bubbles separate more frequently and rapidly recondense within the surrounding liquid, accelerating heat removal.
Ferveret adapted the concept from a nuclear engineering process known as subcooled boiling, which has been studied extensively for improving heat transfer efficiency inside reactors.
Air-cooling is associated with noise, bulk, and inefficiency — three things Azizian decided he wanted no part of when he walked into his first data center in 2017.
"I thought, 'Holy crap, this is not how you cool facilities,'" he recalled, noting that air-cooling alone can consume as much as 40% of a data center's total power supply.
“It was not an efficient way of doing things, but since it wasn’t hurting the performance, no one cared that the cooling technology was 50 years old.”
The company says its liquid does not contain PFAS chemicals, often associated with certain advanced cooling technologies.
Ferveret also delivers its APC platform through compact modular units, with each enclosure designed to hold a single server.
According to Azizian, “The physics enable us to get to form factors that weren’t possible in the past.”
Efficiency gains could ease pressure on growing AI infrastructure
Ferveret recently collaborated with researchers from the University of California, Los Angeles, to evaluate the performance of its technology.
According to the company, the APC approach produced a 15% improvement in computational power efficiency compared with leading liquid cooling alternatives.
Ferveret further claims that combining APC with its control software enables operators to generate 35% more tokens from AI workloads using the same power supply.
The company also supplies racks, cooling distribution equipment, sensors, and monitoring software that continuously adjust operating conditions.
Bucci explained that the software analyzes temperature and pressure measurements in real time to reduce unnecessary energy consumption.
“Liquid is a better heat transfer medium than air. That’s why when you stick your hand into room temperature water it still feels cold,” Bucci explains.
“When liquid is boiling, it becomes even better at removing heat because the phase change requires a lot of energy, which is the energy you remove from the chip...”
The founders argue that lower power demand and zero water consumption could make new facilities practical in regions where electricity and cooling resources remain limited.
That possibility could prove significant for parts of Africa, the Middle East, and the United States, where solar energy is abundant while water availability remains constrained.
Ferveret is currently testing its technology with organizations including CleanSpark, FuriosaAI, and Switch, while also participating in Nvidia's Inception startup program.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
