ICE Datacenter: Meeting AI's Energy Demands
A full-scale research datacenter and test facility for next-generation digital infrastructure
ICE Datacenter, the leading research facility in Europe, has for a decade served as a testbed in Sweden for next-generation digital infrastructure. As AI’s energy demands surge, ICE is taking on the task of prototyping the solutions tomorrow’s systems will depend on.
AI Needs Power: A Lot of It
The global acceleration of AI development is reshaping the digital landscape. Large-scale model training, foundation models, and high-performance inference workloads require unprecedented amounts of electricity and cooling capacity. Datacenters, already among the fastest-growing energy consumers, now face steep increases in thermal loads and power demand.
Cooling alone can have a large operational energy use. Without new approaches, the energy footprint of AI risks outpacing both grid capacity and sustainability targets. Europe aims to build sovereign and resilient AI capabilities, but achieving this depends on datacenters that are not only powerful, but also radically more efficient and adaptable.
How ICE Is Addressing the Challenge
Advanced Cooling Technologies
ICE is a leading testbed for next-generation cooling systems, from direct-to-chip and rear-door liquid cooling to full immersion test-rig setups. These technologies dramatically reduce energy use while supporting denser, hotter AI workloads. Experiments at ICE use high-power server emulators and investigate how various cooling strategies influence thermal stability, server performance, and overall facility efficiency.
Heat Recovery
Heat from servers is one of the central constraints in today’s datacenters, as rising AI workloads push thermal limits and drive up energy use. ICE addresses this challenge by treating waste heat as a usable resource through pilots that recover and feed excess thermal energy into local district-heating systems, and by prototyping the next generation of power-dense, energy-efficient systems.
Frequency Regulation
ICE is also pioneering research on how datacenters can operate as active components within the energy grid. By dynamically adjusting power consumption, ICE investigates how datacenters can support frequency regulation and help stabilize the electricity system as renewable energy sources become increasingly dominant.
A Platform for the Future
Since its launch in 2016, ICE has evolved into a full-scale, industry-grade test and demo facility where new ideas can be prototyped long before they reach commercial datacenters. With thousands of physical servers, future GPU server emulators, advanced cooling technologies, and a fully controllable smart facility grid, the environment enables researchers and companies to experiment under real operating conditions.
Impact
As AI reshapes industry and society, datacenters must become cleaner, smarter, and more deeply integrated with local energy systems. By combining life cycle assessments, efficient power distribution, high-performance computing, sustainable cooling, heat recovery, and grid-support technologies, ICE offers a glimpse of what tomorrow’s infrastructure can look like.
Interested in ICE Datacenter: Meeting AI's Energy Demands?
Get in touch to learn more about how this initiative can support your organization.