AI is driving unprecedented change in the data center, forcing operators to rethink infrastructure design and management. One area getting special attention is cooling. AI chip clusters run much hotter than other equipment, requiring liquid cooling systems to dissipate the added heat.
Air cooling, traditionally the prevailing method of heat dissipation, supports systems of up to 50kW per rack. With AI workloads reaching 150kW and higher, it’s necessary to add liquid cooling to data center infrastructure. Server components, other than the main processing unit such as graphics processing units (GPU), still will be cooled with air, so air cooling will remain necessary – at least for the foreseeable future. But determining how much of each cooling method data centers should deploy is tricky because it requires guessing future needs.
For guidance, data center operators should work with partners that provide the equipment and expertise to meet the following challenges:
- Future-proof data center facilities to accommodate air, liquid, and hybrid cooling methods as requirements evolve
- Set up the infrastructure to scale operations as AI adoption accelerates
- Service new equipment, especially liquid cooling systems
Many decisions must be made. For instance, how should operators connect liquid cooling systems for AI clusters with in-room cooling? Regardless of the mix, the heat removal infrastructure is likely the same. It comes down to having the flexibility to adapt to different types of workloads as they are added. Operators may need more access to prefabricated cooling kits that fit right into the infrastructure like pieces of a puzzle, accelerating deployment from weeks to months with consistency and predictability.
Schneider Electric boosts liquid cooling portfolio
Schneider Electric is better positioned than ever to help data centers with evolving cooling needs. By acquiring a controlling stake in Motivair Corp., we gained expertise, processes, and solutions to provide an end-to-end infrastructure portfolio, from grid to chip and chip to chiller cooling portfolio.
Motivair’s AI-ready offerings range from traditional air-cooled environments to fully liquid-cooled or hybrid implementations. This mix enables us to deliver the agility and flexibility data centers need for the transition. We can deploy equipment that has been pre-validated and pre-approved to work with servers from various manufacturers.
Motivair solutions also provide the right fit for different Nvidia AI clusters. Motivair has worked with the chipmaker for eight years to match cooling solutions to the technology stack, which helps customers accelerate deployment. When Nvidia releases a new cluster, the size of the cooling system already has been aligned, giving customers ready-to-use building blocks as they set up new infrastructure. Schneider has developed designs for accelerated compute cluster implementations. With Motivair, we add another key piece of the puzzle to the designs for straightforward, accelerated deployment.
Another benefit of Schneider’s expanded cooling solutions portfolio is geographic reach. Motivair has manufactured its systems based in the United States, but now we are expanding manufacturing to our plants in Europe and Asia, making it easier for customers around the globe to scale their operations with consistent access to Motivair technology and expertise.
Comprehensive data center infrastructure solutions
With Motivair, Schneider provides a full complement of cooling solutions critical to AI adoption. Schneider and Motivair technicians are receiving training on each other’s technologies, which will allow us to increase the breadth and depth of our service and product offerings globally. Schneider is a full infrastructure company, providing comprehensive solutions from grid to chip and chip to chiller to help data centers smoothly transition to the AI era. Discover how Motivair’s cooling technologies can address your organization’s AI data center cooling challenges.
Add a comment