Ferveret Lifts Data Center Efficiency 15% With Nuclear-Inspired Cooling as AI Power Demand Surges
Updated
Updated · MIT News · Jun 10
Ferveret Lifts Data Center Efficiency 15% With Nuclear-Inspired Cooling as AI Power Demand Surges
3 articles · Updated · MIT News · Jun 10
Summary
A UCLA-linked study found Ferveret’s Adaptive Phase Cooling improved computational power efficiency by 15% versus current liquid-cooling systems, targeting a major bottleneck for AI data centers.
The system submerges servers in a low-boiling liquid and uses smaller, faster-detaching bubbles to speed heat transfer, cutting electricity use without water and avoiding PFAS chemicals.
With Ferveret’s power-control software added, data centers can generate 35% more AI tokens from the same power budget, the company said.
CleanSpark, FuriosaAI and Switch are already testing the modular one-server boxes, while Ferveret says it is also in talks with hyperscalers and plans broader partnerships later this year.
Data centers could consume 9% to 17% of U.S. electricity by 2030, and roughly one-third of their power now goes to cooling—making water-free efficiency gains increasingly valuable.
Will waterless cooling transform remote deserts into the world's next AI infrastructure hubs?
What is the hidden environmental cost of the specialized liquid enabling this water-free AI boom?
Ferveret’s Adaptive Phase Cooling: Achieving 15% Higher AI Compute Efficiency with 0 Water Use and 1.03 PUE for Sustainable Data Centers
Overview
The rapid growth of artificial intelligence is putting huge pressure on data centers, which now face unprecedented demands for both power and water. Traditional cooling methods are struggling to keep up, leading to significant environmental strain, as seen in places like Phoenix where water use for cooling is soaring. This urgent situation highlights the need for new, more efficient cooling solutions. Ferveret’s Advanced Process Cooling (APC) technology is gaining strong market traction by addressing these challenges, offering a promising path toward sustainable and efficient data center operations in the AI era.