Ferveret Unveils Its New Cooling Solution
On June 12 at 2:00 PM, Ferveret introduced a server cooling system that adapts a technique called subcooled boiling, originally developed for nuclear reactors. This novel approach delivers up to 15% greater computational efficiency compared to the most advanced immersion cooling systems today. It also enables data centers to generate 35% more tokens from AI models without raising electricity consumption. The system, branded Adaptive Phase Cooling (APC), is built from compact modules, each housing a single server.
Why This Innovation Matters
Ferveret’s technology is engineered for data center use and relies on a coolant with a low boiling point. In this process, vapor bubbles rapidly condense back into liquid within the cooling medium itself. Company co-founder Reza Azizian explained that “this design makes it simple to integrate the system into existing infrastructure.” He also noted that “most immersion cooling setups depend on large tanks, whereas Ferveret’s solution comes in a compact, modular format.”
This technology could prove especially valuable in regions facing water shortages but abundant in solar energy. Co-founder Matteo Bucci, who is also an MIT associate professor of nuclear science and engineering, emphasized that “it can allow data centers to be located in areas where resource constraints previously made this difficult—such as Africa, the Middle East, and parts of the United States.”
With demand for artificial intelligence surging, the efficiency of this cooling innovation becomes a critical factor for data centers that supply computing power. By the end of this decade, data centers may account for as much as 17% of total U.S. electricity consumption, underscoring the urgent need for energy-efficient solutions in this sector.
The rollout of Ferveret’s Adaptive Phase Cooling could fundamentally reshape server cooling strategies, especially as workloads on data centers intensify. This not only helps cut energy costs but also broadens the possibilities for siting data centers in resource-limited regions. This breakthrough marks a significant step forward in building the infrastructure required to support AI and high-performance computing—essential for the smooth operation of modern information technology.
As the demand for AI technologies continues to rise, the potential energy implications of these advancements cannot be overlooked. A recent report highlights that AI servers may contribute to an impending energy crisis by 2026. This underscores the importance of innovative cooling solutions like Ferveret's, which aim to mitigate energy consumption while enhancing performance in data centers.