Research led by FAU’s Arslan Munir shows that direct-to-chip liquid cooling keeps high-density GPU clusters significantly cooler and uses less power than traditional air-cooling systems. The gain was up to 17% higher performance and 16% lower node-level energy consumption, translating to potential savings of $2 million to $11 million per year for AI data centers with between 2,000 and 5,000 GPU nodes.

“Energy-efficient AI infrastructure isn’t just an engineering optimization – it’s a national imperative,” says Munir. “By rethinking how we cool and power AI factories, we can dramatically increase performance while aligning with global sustainability and cost-reduction goals.”

• Arslan Munir