via The Verge AI
Nvidia’s Hotter AI Data Centers Slash Water Use to Near Zero
Nvidia has unveiled a new AI data center design that operates at higher temperatures and uses 100 percent liquid cooling, promising to reduce water consumption to “near zero.” This represents a significant shift from traditional air-cooled systems, which rely heavily on evaporative cooling and consume vast amounts of water—a growing concern as AI workloads continue to expand.
By 2026, as global AI infrastructure demands are projected to double, water scarcity has become a pressing issue for data center operators. Nvidia’s approach directly addresses this by eliminating the need for water-based cooling towers. Instead, the company’s direct-to-chip liquid cooling circulates a non-conductive fluid through cold plates attached to GPUs and CPUs, capturing heat more efficiently and allowing servers to run at higher ambient temperatures without risking thermal throttling.
The design leverages Nvidia’s latest GPU architectures (such as the anticipated “Blackwell” successor), which are engineered to operate reliably at elevated temperatures. According to Nvidia, this not only cuts water usage but also reduces overall energy consumption for cooling by up to 30% compared to conventional air-cooled facilities. The company claims that for every megawatt of IT load, the new design can save approximately 3.5 million gallons of water annually.
Nvidia’s initiative aligns with broader industry trends toward sustainable computing. Major cloud providers like Google and Microsoft have also pledged to become water-positive by 2030, but Nvidia’s solution offers a more immediate hardware-level fix. The company plans to make reference designs available to partners in 2025, with deployments expected in new hyperscale data centers by late 2026.
Critics caution that liquid cooling systems require higher upfront capital investment and specialized maintenance. However, Nvidia argues that total cost of ownership (TCO) improves over five years due to lower utility bills and reduced water procurement costs. As regulators in drought-prone regions tighten water usage permits, this design could become a de facto standard for next-generation AI data centers.
In summary, Nvidia’s hotter-running, liquid-cooled data centers represent a strategic pivot toward sustainability without sacrificing performance—a critical balance as AI’s environmental footprint comes under increasing scrutiny.