AI’s next bottleneck is not the model. It is heat.

The most important computing-power story this past week was NVIDIA’s new 45°C liquid-cooling design for next-generation Rubin AI infrastructure.

That sounds backwards at first. Why would hotter cooling be better? Because if AI servers can safely run with warm liquid instead of chilled air or evaporative cooling towers, data centers can cut cooling energy, reduce onsite water use, and pack more compute into the same physical footprint.

NVIDIA says its Rubin-generation AI infrastructure is designed for fully liquid-cooled systems, including chips and networking components, using a closed loop. Reports from The Verge and Tom’s Hardware note the design could reduce onsite cooling water use to nearly zero in favorable climates by relying more on dry coolers instead of traditional water-intensive cooling towers.

Why it matters: AI data centers are running into power, grid, and cooling limits at the same time. Gartner recently projected worldwide data center power demand will rise sharply in 2026, while the IEA expects global data center electricity consumption to roughly double by 2030. That means future AI growth will depend not only on faster GPUs, but on better power delivery, better heat removal, and smarter site selection.

This cooling shift is a practical sign of where computing power is heading: from “more chips” to “more usable watts.” The winners in AI infrastructure may be the companies that can turn electricity into compute with the least wasted energy, water, and grid stress.

Researched and written by Peter Jonathan Wilcheck

Searchable References

1. NVIDIA Blog: https://blogs.nvidia.com/blog/liquid-cooling-ai-factories/
2. The Verge: https://www.theverge.com/tech/954139/nvidia-data-centers-rubin-liquid-cooling
3. Tom’s Hardware: https://www.tomshardware.com/tech-industry/data-centers/nvidia-announces-liquid-cooling-system-that-runs-hotter-than-a-hot-tub-promises-to-reduce-electricity-consumption-and-cut-water-use-by-up-to-100-percent-but-sustainability-challenges-remain
4. Gartner: https://www.gartner.com/en/newsroom/press-releases/2026-06-10-gartner-says-data-center-electricity-demand-to-grow-26-percent-in-2026
5. IEA Energy and AI: https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai