NVIDIA GTC Washington D.C. Announcements | Generated by AI
NVIDIA Announcements on October 29, 2025
NVIDIA is wrapping up its GTC Washington D.C. conference (October 27–29), with several major reveals today focused on AI infrastructure, telecom advancements, and next-gen hardware. Here’s a breakdown of the key ones:
Building America’s AI Infrastructure
NVIDIA announced collaborations with the U.S. Department of Energy’s national labs (Argonne and Los Alamos) and leading companies to deploy massive AI supercomputers and create a blueprint for gigawatt-scale “AI factories.”
- Key systems: Solstice (100,000 NVIDIA Blackwell GPUs, DOE’s largest AI supercomputer, built with Oracle) and Equinox (10,000 Blackwell GPUs, operational in 2026) at Argonne, delivering up to 2,200 exaflops. Additional systems like Tara, Minerva, Janus at Argonne, and Mission/Vision at LANL using the upcoming Vera Rubin platform.
- Partners: Includes Oracle for Solstice, HPE for hardware, Bechtel/Jacobs for digital twins, and power/cooling firms like Eaton, GE Vernova, and Tesla. Cloud providers such as Google Cloud, Microsoft Azure, and xAI are scaling AI factories with Blackwell and GB300 racks.
- Goal: Accelerate scientific discovery, economic growth, and the “next industrial revolution” through American-led AI innovation. NVIDIA is launching an AI Factory Research Center in Virginia powered by Vera Rubin.
This positions the U.S. as a leader in AI, with investments from enterprises like Lilly (drug discovery) and Mayo Clinic (healthcare).
Partnership with Nokia for AI-Native 6G Networks
NVIDIA and Nokia unveiled a strategic alliance to pioneer AI-powered radio access networks (AI-RAN) for 5G-Advanced and 6G, aiming to reclaim U.S. telecom leadership.
- Investment: NVIDIA committing $1 billion to Nokia (at $6.01/share) to speed up innovation.
- New tech: NVIDIA Aerial ARC-Pro platform (a 6G-ready RAN computer combining connectivity, computing, and sensing). Nokia will integrate it into its AirScale baseband for upgrades.
- Collaborators: T-Mobile for 2026 trials on performance/efficiency; Dell for PowerEdge servers.
- Impact: Targets the $200B+ AI-RAN market by 2030, handling AI traffic surges (e.g., ChatGPT on mobile) and enabling new services for drones, AR/VR, and agentic AI. Also exploring Nokia’s optical tech for NVIDIA’s AI data centers.
Reveal of Next-Gen Vera Rubin Superchip
In a major hardware unveiling, NVIDIA showcased the Vera Rubin platform—its successor to Blackwell—for extreme-scale AI.
- Specs: Vera CPU (88 ARM cores, 176 threads) paired with two Rubin GPUs (each with two reticle-sized dies, up to 50 PFLOPs FP4, 288 GB HBM4 memory). Supports 1.8 TB/s NVLink and up to 1.8 TB system memory.
- Systems:
- Rubin NVL144 (2H 2026): 3.6 exaflops FP4 inference (3.3x GB300), 13 TB/s HBM4.
- Rubin Ultra NVL576 (2H 2027): 15 exaflops FP4 (14x GB300), 4.6 PB/s HBM4.
- Timeline: GPUs in TSMC labs now; mass production Q3/Q4 2026 or sooner.
This powers the AI factories and supercomputers mentioned above, with 2–12x performance uplifts in inference, training, and bandwidth.
Other notes: Yesterday’s Palantir integration (CUDA-X and Nemotron into AIP for enterprise AI) ties into today’s infrastructure push, and NVIDIA contributed to open robotics frameworks at ROSCon (ending today).
NVIDIA and Partners Build America’s AI Infrastructure
NVIDIA and Nokia to Pioneer the AI Platform for 6G
NVIDIA Shows Next-Gen Vera Rubin Superchip
GTC Washington D.C. Event Page