---
title: "NVIDIA explains how 45°C liquid can cool AI servers"
description: "A June 21 report describes warm-liquid cooling for Rubin. Coolant temperature should not be confused with room temperature or the chip's operating limit."
url: https://kurddevelopers.com/en/news/nvidia-45c-liquid-cooling-ai-factories-june-2026
language: en # English
type: news
published: 2026-06-21T00:00:00.000Z
updated: 2026-11-02T21:13:55.000Z
author: "Osamah Ismail"
author_url: https://kurddevelopers.com/en/author/1
section: "Computers"
tags: ["NVIDIA","Rubin","Liquid Cooling","Data Centers"]
sources: ["NVIDIA — 21 June 2026 — https://blogs.nvidia.com/blog/liquid-cooling-ai-factories/","NVIDIA — Vera Rubin architecture","5 January 2026 — https://developer.nvidia.com/blog/inside-the-nvidia-rubin-platform-six-new-chips-one-ai-supercomputer/","ENERGY STAR — Data-center temperature guidance — https://www.energystar.gov/products/data_center_equipment/5-simple-ways-avoid-energy-waste-your-data-center/raise-temperature"]
image: https://kurddevelopers.com/uploads/2026/11/45cliquidcooling-f8d0dfd6.jpg
translations:
  ku: https://kurddevelopers.com/news/nvidia-45c-liquid-cooling-ai-factories-june-2026
  ar: https://kurddevelopers.com/ar/news/nvidia-45c-liquid-cooling-ai-factories-june-2026
publisher: Kurd Developers
publisher_url: https://kurddevelopers.com/
---

# NVIDIA explains how 45°C liquid can cool AI servers

> A June 21 report describes warm-liquid cooling for Rubin. Coolant temperature should not be confused with room temperature or the chip's operating limit.

![Image: NVIDIA](https://kurddevelopers.com/uploads/2026/11/45cliquidcooling-f8d0dfd6.jpg)

## Warm liquid, hotter equipment
On June 21, 2026, NVIDIA explained how Rubin's cooling design can work with coolant reaching 45°C. In suitable climates, the company says this can enable outdoor-air heat rejection and reduce the need for mechanical chilling.

The figure describes the coolant. It is not the server-room temperature or the chip's own operating-temperature limit. Keeping those measurements separate is essential to understanding the claim.

## A platform-level design
NVIDIA's Vera Rubin technical documentation describes direct liquid cooling alongside coordinated decisions about computing, networking and power. The engineering objective is to improve the complete system rather than optimize each component in isolation.

In simple terms, a cooling loop carries heat away from components and transfers it to somewhere that can reject it. The useful question is whether the temperature difference and heat-transfer capacity suit the workload, rather than whether the liquid feels cold.

## Colder is not automatically better
ENERGY STAR's data-center guidance explains that excessive cooling can waste energy. It also stresses monitoring temperature, humidity and hot spots: raising settings without measurements can create problems instead of savings.

That guidance is not permission to run any server at any temperature. Equipment limits and the way air or liquid reaches individual components remain important constraints.

## What does it mean for Kurdistan?
Our interpretation is that planning a data center in a hot region requires more than comparing processor specifications and purchase prices. It also requires asking how heat will be removed, how reliable the electricity supply is, and how the system behaves during the hottest summer conditions.

A saving demonstrated in a cooler climate should not simply be copied into a proposal for Duhok or Erbil. A useful assessment would measure the actual site and compare the entire facility's energy use, rather than selecting one component's improvement as a proxy for everything.

The practical significance is that sustained performance depends partly on thermal engineering. A powerful processor cannot tell the whole operational story on its own. Any claimed local saving still needs a site-specific assessment; this announcement is not an independent measurement of a Kurdistan facility.

## Sources
- [NVIDIA — 21 June 2026](https://blogs.nvidia.com/blog/liquid-cooling-ai-factories/)
- [NVIDIA — Vera Rubin architecture, 5 January 2026](https://developer.nvidia.com/blog/inside-the-nvidia-rubin-platform-six-new-chips-one-ai-supercomputer/)
- [ENERGY STAR — Data-center temperature guidance](https://www.energystar.gov/products/data_center_equipment/5-simple-ways-avoid-energy-waste-your-data-center/raise-temperature)

**Sources:** NVIDIA — 21 June 2026 — https://blogs.nvidia.com/blog/liquid-cooling-ai-factories/, NVIDIA — Vera Rubin architecture, 5 January 2026 — https://developer.nvidia.com/blog/inside-the-nvidia-rubin-platform-six-new-chips-one-ai-supercomputer/, ENERGY STAR — Data-center temperature guidance — https://www.energystar.gov/products/data_center_equipment/5-simple-ways-avoid-energy-waste-your-data-center/raise-temperature

---

Published by [Kurd Developers](https://kurddevelopers.com/) on 2026-06-21. Cite as: “NVIDIA explains how 45°C liquid can cool AI servers”, Kurd Developers, https://kurddevelopers.com/en/news/nvidia-45c-liquid-cooling-ai-factories-june-2026