Nvidia's Jensen Huang: Vera Rubin Seven Chips Fully in Mass Production, Foreseeing One Trillion Dollar Computing Power Orders

Gate News Report, March 17 — NVIDIA Founder and CEO Jensen Huang officially announced at GTC 2026 that the Vera Rubin platform is now in full production, integrating seven new chips and covering five types of rack systems. The entire design is a supercomputer built specifically for AI.

The core rack, Vera Rubin NVL72, integrates 72 Rubin GPUs and 36 Vera CPUs, interconnected via NVLink 6. Compared to the previous Blackwell platform, the number of GPUs needed for training large hybrid expert models has been reduced to a quarter. The inference throughput per watt is up to 10 times higher than Blackwell, and per-token cost is reduced to one-tenth.

The five rack systems form the complete AI factory infrastructure: Vera Rubin NVL72 GPU rack; Vera CPU rack (256 Vera CPUs, twice as efficient and 50% faster than traditional CPUs); Groq 3 LPX inference acceleration rack; BlueField-4 STX storage rack (designed for AI agent key-value caching, with inference throughput up to 5 times higher); Spectrum-6 SPX Ethernet rack.

In energy management, NVIDIA simultaneously announced the DSX platform: DSX Max-Q can deploy 30% more AI infrastructure within a fixed power limit, and DSX Flex can activate previously unused 100 gigawatt grid capacity.

Cloud providers including AWS, Google Cloud, Microsoft Azure, Oracle Cloud, as well as CoreWeave, Lambda, Nebius, and system vendors like Cisco, Dell Technologies, HPE, Lenovo, and Supermicro, all announced plans to launch Vera Rubin products in the second half of this year. Anthropic, Meta, Mistral AI, and OpenAI have explicitly stated they will use this platform to train larger-scale models.

Huang predicted that the combined orders for Blackwell and Vera Rubin systems will reach at least $1 trillion from 2025 to 2027, doubling the $500 billion forecast given at last year’s GTC.