The Infinium Edge Innovation Center isn’t a demo environment — it’s a live inference facility running the same hardware and workloads our customers deploy. Every immersion tank, every GPU node, every telemetry feed operates the same way it would in a production data center.
The difference is that here, we’re running it all in the open: benchmarking against published industry standards and capturing thermal data under sustained load.
The Infinium Edge Innovation Center is a live inference facility running the same NVIDIA Hopper and Blackwell generation GPUs and workloads our customers deploy.
The platform is also running Intel and AMD GPUs, and leading memory architectures from Samsung and Micron — reflecting the silicon-agnostic foundation of the Infinium Edge Thermal Vectoring™ platform. AI infrastructure moves fast, and the cooling layer shouldn’t be the reason you’re locked to a single vendor roadmap.
Edge Thermal Vectoring removes the thermal ceiling that forces air-cooled and direct-to-chip systems to throttle under sustained load. Today, at the Innovation Center, simulated inference loads run GPUs at their full rated power draw — no derating, no throttling, no managed underclocking to protect hardware that can’t shed heat fast enough.
Full GPU utilization with cooling held steady demonstrates the effectiveness of the Edge Thermal Vectoring™ platform. In the Innovation Center, fluid temperature, power draw, and GPU clock speeds are captured continuously throughout every benchmark run.
In an air-cooled environment, a GPU under the kind of sustained load we run would throttle back to protect itself. Junction temperatures climb, power limits kick in, and clock speeds drop — sometimes significantly. The throughput numbers you see in air-cooled benchmarks often reflect a system already in thermal retreat. In our Edge Thermal Vectoring platform, that doesn’t happen. Fluid temperature stays stable, clock speeds hold, and the GPU delivers the same performance at the end of a sustained run as it did at the start. That’s the property we’re measuring. The data shows it.
The Innovation Center serves as the primary validation environment for Infinium Edge dielectric immersion fluids — purpose-built, synthetic hydrocarbons engineered for the thermal and electrical demands of high-density AI compute. Fluid temperature, dielectric stability, and long-term material compatibility are monitored continuously, generating the operational data that supports our fluid performance under real conditions rather than laboratory simulations.
The result is an AI infrastructure platform — cooling system, fluids, and hardware — that has been demonstrated to handle the thermal requirements of today’s frontier silicon and the generations ahead.