🤖 AI Summary
Lambda and ECL have brought the industry’s first production-grade, hydrogen-powered NVIDIA GB300 NVL72 systems online at ECL’s Mountain View MV1 modular data center. The Supermicro-built systems deliver 142 kW of compute per cabinet and are liquid-cooled with direct-to-chip cooling fed by centralized CDUs that recycle the water produced by the hydrogen fuel cells, enabling a zero-water, zero-emissions off-grid deployment. Lambda doubled its footprint to occupy the full facility and reported a two-hour cabinet integration time for these 4,000 lb systems—an operational benchmark for deploying high-density AI hardware.
This rollout matters because it validates hydrogen fuel cells as a practical, on-site, low-carbon power source for extremely dense AI infrastructure used for inference and foundational model training. The combination of high power density (142 kW/cabinet), advanced liquid cooling, and on-site water recycling demonstrates a pathway to scale “gigawatt” AI factories while reducing grid dependence and emissions. It also highlights constraints: few existing data centers can accommodate the weight, power and cooling demands of GB300 NVL72 systems, so broader adoption will require purpose-built facilities or retrofits. For AI/ML operations planning large-scale training or inference farms, this is an early proof point that sustainable, off-grid high-performance deployments are operationally feasible.
Loading comments...
login to comment
loading comments...
no comments yet