NVIDIA Cloud Partners Scale Across Europe — Sovereign, Open and Built for Safe AI Agents
NVIDIA’s European Cloud Partners are expanding sovereign, open, and agent‑centric AI infrastructure, leveraging the DSX AI Factory platform to boost compute efficiency. In tests with NVIDIA GB300 NVL72 systems, DSX MaxLPS delivered 37 % more GPUs and nearly 50 % higher aggregate inference throughput within the same power budget. The rollout includes Vera Rubin training nodes, BlueField adapters, Spectrum‑X Ethernet, and Groq 3 LPX token generators, giving developers low‑latency, serverless endpoints for agentic and edge AI workloads. Engineers can now provision region‑local, power‑optimized clusters to train and serve open models with tighter data‑sovereignty controls.
⚡ Key Takeaways
- 37 % more GPUs and ~50 % higher inference throughput achieved with DSX MaxLPS on GB300 NVL72 under a fixed power budget.
- Architecture combines NVIDIA Vera Rubin, BlueField, and Spectrum‑X Ethernet within the DSX AI Factory platform.
- Trade‑off: Power allocation is coordinated via DSX, requiring careful workload‑power matching to realize efficiency gains.
- Integration uses familiar APIs through managed serverless and dedicated endpoints (e.g., Nebius Token Factory).
- Prerequisite: Access to NVIDIA‑powered DSX infrastructure and compatible hardware (GB300 NVL72, Groq 3 LPX).
By delivering up to 50 % more inference throughput per watt, these regional clouds let production teams scale agentic AI services cost‑effectively while keeping data and compute within European jurisdictions.
✅ Practical Steps
- Deploy workloads on NVIDIA DSX AI Factory platforms (e.g., GB300 NVL72) to leverage DSX MaxLPS power‑allocation features.
- Use Nebius Token Factory with NVIDIA Groq 3 LPX and Spectrum‑X Ethernet to enable low‑latency token generation for agentic applications.
- Configure your model training or inference pipelines to run on Vera Rubin nodes for frontier‑model workloads across the supported European markets.
Want the full story? Read the original article.
Read on NVIDIA Blog ↗