GTC Berlin 2026: Live Updates From NVIDIA CEO Jensen Huang’s Keynote
NVIDIA CEO Jensen Huang announced the DSX POD reference design, a modular containerized AI factory that can be deployed in 14 days versus six months, delivering up to 30 % lower power consumption and up to 100 % water‑cooling reduction. He positioned agents as the new computing platform, highlighting the Vera Rubin stack that adds a harness, memory, sandbox and tool integration to LLMs. The announcement underscores NVIDIA’s full‑stack AI factory—from hardware to CUDA‑X libraries—aimed at accelerating AI production in Europe.
⚡ Key Takeaways
- DSX POD reduces on‑site deployment time from six months to 14 days once utilities are ready.
- The platform includes NVIDIA’s Vera Rubin AI agent stack (model + harness + memory + sandbox + tools).
- NVIDIA DSX AI Factory Platform cuts water‑based cooling by up to 100 % and power use by up to 30 % for the same AI output.
- Engineers can leverage the full‑stack stack (hardware, system software, CUDA‑X libraries, models) to build end‑to‑end AI factories.
- Deployment assumes site utilities (power, networking) are already provisioned; otherwise the 14‑day timeline does not apply.
Faster, lower‑cost AI‑factory deployments let enterprises spin up production‑grade agentic AI workloads in weeks rather than months, dramatically improving time‑to‑value for AI‑driven products.
✅ Practical Steps
- Order the NVIDIA DSX POD reference design and plan for modular container delivery to a prepared site.
- Integrate the Vera Rubin agent stack (model, harness, memory, sandbox, tools) using NVIDIA’s CUDA‑X libraries to enable full‑agent capabilities.
- Configure site utilities (power, networking) to meet DSX POD requirements, then schedule the 14‑day on‑site installation window.
Want the full story? Read the original article.
Read on NVIDIA Blog ↗