From Nuclear Reactors to AI Racks: Ferveret's Cooling Tech Promises a Waterless Future

SAN JOSE, CA – March 26, 2026 – A San Jose-based startup is leveraging thermal management principles from nuclear reactors to solve the artificial intelligence industry’s escalating energy and water crisis. Ferveret, a company founded by MIT alumni, today announced benchmark results showing its waterless cooling technology provides a 15% boost in computational efficiency for the powerful chips that train AI models, a significant leap in performance that could reshape data center economics and sustainability.

The findings, stemming from a study with UCLA’s Computer Science Department, position Ferveret’s “Adaptive Phase Cooling” as a critical solution at a time when the voracious appetite of data centers for power and water is drawing global concern. The technology promises not only to enhance performance but also to eliminate water consumption entirely for cooling, a radical departure from current industry practices.

The Unquenchable Thirst of Artificial Intelligence

The AI boom is placing an unprecedented strain on the world's power grids and water supplies. Data centers, the factories of the digital age, are at the heart of this demand. According to the U.S. Department of Energy, these facilities could consume as much as 12% of the nation's total electricity by 2028, a staggering increase from just 4.5% in 2023. Globally, IDC predicts electricity consumption by data centers will more than double by 2028, driven largely by power-hungry AI workloads.

This insatiable demand for energy is matched by a critical need for water. Traditional cooling systems, which often use evaporative cooling towers, consume vast quantities of water to dissipate the immense heat generated by densely packed servers. Projections indicate that U.S. data centers could soon consume over 700 billion gallons of water annually, placing them in direct competition with agriculture and residential needs, particularly in water-scarce regions. The growing environmental footprint has not gone unnoticed, prompting legislative proposals like the Data Center Water and Energy Transparency Act to mandate disclosure of resource usage.

“AI and neo-cloud are redefining the scale, density and intensity of modern computing, and legacy cooling simply cannot keep pace,” said Reza Azizian, CEO of Ferveret, in a statement. “It is critical that we improve the environmental footprint of data centers as this new era unfolds.”

A Cooling Breakthrough from an Unlikely Source

Ferveret’s solution is born from an interdisciplinary leap, adapting a technique known as “subcooled boiling” used for decades to safely manage extreme heat within nuclear reactors. The company’s co-founders, including CTO Matteo Bucci, a faculty member in Nuclear Science and Engineering at MIT, have translated these principles to the micro-level of a silicon chip.

Unlike traditional liquid cooling where a fluid boils and turns to vapor, Ferveret’s Adaptive Phase Cooling keeps the bulk of its dielectric (non-conductive) fluid in a liquid state. Tiny bubbles form directly on the chip's surface, detach, and quickly recondense in the surrounding cooler fluid. This rapid cycle continuously pulls heat away far more efficiently than other methods, allowing chips to operate at lower temperatures even under extreme loads.

The recent benchmark study, conducted with UCLA’s Intelligent Connectivity Laboratory on NVIDIA H200 GPUs, validated this approach. The results showed a 15% improvement in computational efficiency—measured in tera floating-point operations per second per kilowatt (TFLOPs/kW)—compared to leading direct-to-chip liquid cooling systems. This metric demonstrates that for the same amount of power, Ferveret’s system enables the hardware to perform more calculations.

“Ferveret isn’t just a cooling solution—it’s a performance multiplier that redefines expectations for modern computing infrastructure,” commented Omid Abari, an Associate Professor at the UCLA Computer Science Department who collaborated on the study. “Our recent study shows that Ferveret cooling reduces the time required to train machine learning algorithms by enabling hardware to operate at higher sustained clock speeds.”

The Competitive Edge in a High-Density Market

Beyond its environmental benefits, the technology presents a compelling economic case. The 15% efficiency gain translates directly into faster AI model training and more computational work from the same expensive hardware, shortening the path to revenue for data center operators. Furthermore, the system achieves a Power Usage Effectiveness (PUE) of 1.03. PUE is a measure of how much energy a data center uses for its computing equipment versus auxiliary needs like cooling; a score of 1.0 is the theoretical ideal. A PUE of 1.03 places Ferveret’s solution among the most efficient in the world, promising significant operational cost savings.

This performance positions the company strongly within a rapidly evolving and competitive market. The data center cooling industry is projected to grow into a market worth over $70 billion by the early 2030s. Ferveret competes with established technologies like direct-to-chip (DTC) cooling, where liquid is piped directly to processors, and full immersion cooling, where entire servers are submerged in dielectric fluid. While competitors also claim ultra-low PUEs, Ferveret’s combination of high efficiency, zero water usage, and operation near ambient pressure offers a unique value proposition.

Investor Confidence and the Road to Deployment

Ferveret’s innovative approach has attracted a roster of prominent investors, signaling strong market conviction. Backers include Y Combinator, the famed startup accelerator, along with venture capital firms like TO VC, Cerberus, and Aramco Ventures, the corporate venturing arm of the Saudi Arabian oil giant, which has a dedicated fund for energy transition technologies.

“The data center industry urgently needs breakthrough technologies that can meet the accelerating performance demands of AI infrastructure,” noted Charles Goodwin, Partner at TO VC. He highlighted the founders' unique background, stating they “spent years solving thermal management challenges in the nuclear power industry and are now applying those techniques to address the fundamental limitations of data center cooling.”

Ferveret appears to be moving swiftly from research to reality. The company has indicated it is preparing a 1 MW modular, liquid-cooled compute pod for a customer delivery, a key step in commercializing its technology. As the first units are deployed, the entire data center industry will be watching to see if this nuclear-inspired innovation can truly cool a sector running red-hot.