Infinium has revealed the Infinium Edge, a platform designed for high-density artificial intelligence (AI) and high-performance computing (HPC) workloads. The system uses immersion cooling technology to address the data centre challenge of efficient heat management.
The surge in power densities of GPUs, CPUs, and supporting components has made cooling an important factor impacting the performance, efficiency, and scalability of data centres. This is where Infinium Edge Immersion Fluids are used. These custom-engineered dielectric fluids have been crafted for AI and HPC data centres, aiming to remove heat directly at its source, thereby sustaining high-efficiency thermal performance.
This means operators can deploy next-generation AI hardware at greater densities, strengthening reliability while reducing infrastructure intricacies. Notably, where traditional air-cooled centres can support only around 10–20 kW per rack, and direct-to-chip liquid cooling achieves 40–80 kW, immersion solutions present an opportunity for even higher densities, helping demonstrate their role in future data centre designs.
As AI workloads push past the capabilities of current cooling technologies, immersion-based techniques have emerged as a viable strategy for enhanced thermal efficiency and density. Infinium Edge's platform is built on chemistry and industrial process engineering, designed to facilitate large-scale, immersion-cooled data centre operations.
Chief Executive Officer, Robert Schuetzle, emphasised the impact of the platform, highlighting the capability of Infinium Edge to help advance in AI infrastructure. He emphasised its dual advantage of enabling faster computing and sustainable operations whilst reducing the power and water consumption typical of conventional cooling methods.
Infinium Edge fluids are synthetic and do not contain petroleum-derived contaminants, which can affect reliability and long-term performance. They are designed to maintain consistent performance under the high-temperature conditions common in AI and HPC environments.
