- Schneider Electric and Nvidia join to build ready-to-deploy AI data centres using liquid cooling and prefab parts.
- The design helps operators manage the rising power and heat needs of AI chips like Nvidia’s GB200.
What happened: Schneider and Nvidia unveil fast-build AI data centres with liquid cooling and energy monitoring software
Schneider Electric and Nvidia have built a new plan for modular AI data centres. These centres use Nvidia’s MGX servers, GB200 NVL2 chips, and Liquid-Cooled AI Pods. Schneider adds the power and cooling parts. These are made ahead of time in a factory. The goal is to make it fast to build new AI data sites that use a lot of power and give off a lot of heat.
There is also a software tool called EcoStruxure Resource Advisor Copilot. It runs on Nvidia software. This tool shows how systems work and helps data centres track their energy use. The companies showed all of this at Schneider’s Innovation Summit in Paris. Nvidia’s VP Charlie Boyle said this work lets companies “bring AI to data anywhere.”
Schneider says the design uses liquid cooling that is up to 30 percent more efficient than air cooling. It can also scale quickly.
Also Read: AI data centres to surpass Japan’s power use by 2030
Also Read: DOE identifies sites for AI data centres
Why this is important
AI models like GPT-4, Gemini, and Claude need a lot of power and space. Some systems use more than 100kW per rack. Most old data centres cannot handle that. Cooling is also hard. Fans and air systems no longer work well. So new centres need liquid cooling.
Schneider’s modular plan may fix this. It lets firms set up fast. The design uses closed-loop cooling, which means it does not waste water. The materials used can also be reused or recycled. These features make it easier to run centres in more places.
This plan may help cloud platforms, colocation centres, and others. They can set up systems fast without building from scratch. Schneider gives the power and cooling. Nvidia gives the AI chips. It is a full package.
Big firms are doing the same. Meta uses immersion cooling. Microsoft is testing liquid-based cooling too. All of this shows that data centres must change. The heat and power needs of AI are now too high for old ways to work.