European Ecosystem Builds World’s Most Efficient AI Factories With NVIDIA DSX
NVIDIA launched the DSX AI Factory Platform at GTC Berlin, a full-stack reference architecture codesigning compute, networking, power and cooling for AI factories, with Europe's leading infrastructure partners adopting it. DSX reference designs use closed-loop liquid cooling and dry coolers to cut facility cooling water use to near zero — 54 million liters a day saved at 2-gigawatt scale — while DSX MaxLPS software delivers the same token output with up to 30% less power. Cloud partners including CoreWeave, Google Cloud, Lambda, Nebius and Nscale are exploring the software; Nscale demonstrated 49% higher inference throughput within a fixed 264-kilowatt power budget using DSX.
Key figures
- Nscale power budget
- 264 kilowatts
- Maxlps power savings
- up to 30%
- Lambda gpu capacity gain
- 19%
- Nscale gpu capacity gain
- 37%
- Grid power draw reduction
- up to 40%
- Reference deployment scale
- 2-gigawatt
- Lambda token throughput gain
- 24%
- Water savings liters per day
- 54000000
- Nscale inference throughput gain
- 49%
AI analysis
The rest of the AI analysis, red flags and sentiment are part of Signal8 Pro.
AI-generated analysis of a public disclosure. Not investment advice; verify against the original document.