NVIDIA unveils 5 AI infrastructure racks at San Jose GTC
NVIDIA just announced a big move to shake up how AI data centers work, showing off its latest tech at the GTC conference in San Jose.
CEO Jensen Huang says using NVIDIA's full lineup can make these centers run smoother and more profitably.
It even displayed five different racks covering all aspects of AI infrastructure, while Huang also talked about robotics and AI in space, basically, a one-stop shop for all things AI.
LPX rack speeds token processing
Launching later this year, the LPX rack mixes powerful Rubin GPUs with Groq 3 LPUs and uses on-chip SRAM for much faster data handling.
Ian Buck, who leads hyper-scale and high-performance computing at NVIDIA, shared that this setup seriously speeds up token processing, making it a big step forward for anyone building or running AI systems.