A startup founded by two Massachusetts Institute of Technology researchers is adapting nuclear reactor science to reduce the energy and water needed to cool AI data centres.
Ferveret’s Adaptive Phase Cooling system submerges computer servers in a specialised liquid that absorbs heat more efficiently than air.
The company says the system uses no water and significantly less electricity than conventional cooling approaches.
The technology targets one of the fastest-growing sustainability challenges in digital infrastructure: how to keep high-powered AI chips cool as data centre demand rises.
Why is data centre cooling under pressure?
Artificial intelligence is driving a major expansion in data centre capacity.
Data centres are projected to account for 9 to 17 per cent of total United States electricity use by the end of the decade.
Around one-third of today’s data centre electricity is used to cool the chips that run AI models.
Traditional air cooling relies on large fans to move heat away from servers.
Ferveret co-founder Reza Azizian said he saw that limitation when he first walked into a data centre in 2017.
“I thought, ‘Holy crap, this is not how you cool facilities,’” Azizian said. “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.”
- Want more information on what Inside Water is up to? Sign up for our weekly email, landing in inboxes with the latest news.
- Do you want the magazine delivered to your letterbox? Sign up here to subscribe.
What makes Ferveret’s system different?
Liquid is already more effective than air at transferring heat.
That is why many data centre operators have moved toward liquid cooling, including immersion cooling, where chips are submerged in liquid.
Ferveret’s approach goes further by adapting a nuclear reactor process known as subcooled boiling.
The system uses a liquid with a low boiling point and no toxic PFAS forever chemicals.
At the surface of the chip, Ferveret’s liquid produces smaller bubbles than other immersion cooling approaches.
Those bubbles detach more frequently and recondense more quickly in the surrounding liquid.
That accelerates the bubble-rewetting cycle at the chip surface and improves heat transfer.
Why do the bubbles matter?
In boiling liquid systems, phase change improves heat removal.
When a liquid becomes a vapour, it absorbs a large amount of energy from the heat source.
MIT Associate Professor Matteo Bucci said that makes boiling a powerful cooling mechanism.
“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,” Bucci said. “That lets you transfer large quantities of heat with minimal temperature differences between the chips and the liquid.”
The challenge is that boiling systems can add complexity.
Operators must manage bubbles, pressure, temperature and fluid inventory.
Ferveret’s subcooled boiling approach is designed to improve heat transfer while keeping the vapour phase local and controlled.
How does it compare with existing systems?
Ferveret said its Adaptive Phase Cooling system delivered a 15 per cent improvement in computational power efficiency compared with state-of-the-art liquid cooling solutions.
That result came from a study conducted in collaboration with the Samueli Computer Science Department at the University of California, Los Angeles.
The company said combining those savings with its power control system allows data centres to get 35 per cent more tokens from AI models with the same amount of power.
Tokens are small pieces of text or data produced by AI models.
Azizian said the company’s goal is to make data centres use power more efficiently.
“Our goal is to make data centres as sustainable as possible and help them use every single watt of power to generate tokens, which are the most useful outputs,” Azizian said. “Our system enables the operation of more powerful chips, it helps data centres waste a lot less energy, and it accomplishes all that with zero water consumption.”
How does the design support deployment?
Many immersion cooling systems use large tanks that hold multiple servers.
Ferveret has taken a modular approach.
Its system is delivered in small boxes, each housing one server.
The founders said that this makes deployment easier and simplifies maintenance.
“The physics enable us to get to form factors that weren’t possible in the past,” Azizian said. “Most immersion cooling solutions are large tanks that people submerge the servers in. We have a smaller, modular rack-mounted solution that makes it adaptable to the current infrastructure, so it’s easier for people to deploy our technology.”
The company also offers control software that adjusts the power going to each server in real time.
Bucci said the company delivers full-stack systems, including the cooling box, rack, cooling distribution units and sensors that measure temperature and pressure.
“Our software monitors those sensors and optimises the operating condition inside each box to ensure that energy consumption is minimised in the system,” Bucci said.
Why does water-free cooling matter?
Water use has become a major constraint for data centre development.
Cooling demand can limit where facilities are built, especially in regions with abundant renewable energy but limited water.
Bucci said water-free cooling could open up locations that are well suited to solar power but poorly suited to water-intensive cooling.
“The sun shines in places where you don’t have much water, so the advantage of us being water-free is we allow you to build data centres where you have solar energy but nothing to cool the data centre down,” Bucci said. “This technology can help deploy data centres in regions where normally you wouldn’t have the resources to do so, including Africa, the Middle East, and, of course, parts of America. It’s a huge unlock.”
Who is testing the technology?
Ferveret is testing its technology with CleanSpark, a data centre developer and operator.
It is also working with FuriosaAI, an AI accelerator company, and Switch, one of the largest data centre operators in the United States.
The company is part of Nvidia’s Inception program for startups and is in talks with major cloud computing companies.
Azizian said data centre operators are trying to get more computing output from limited power and water resources.
“The computing industry is facing a huge challenge in the form of access to power, and they have a problem with access to water in many regions,” Azizian said. “That will only become more limiting as the industry grows. The main goal for these data centre operators would be to get more tokens from the power they have. We’ve shown we can do that.”
