Nvidia debuts Groq 3 LPX inference system with up to 35x speedup

Nvidia CEO Jensen Huang unveiled a new inference system at the company’s annual GTC keynote. The Nvidia Groq 3 LPX integrates technology from startup Groq and pairs it with Nvidia’s Vera Rubin architecture. Huang said the chip can speed up inference workloads by up to 35 times.

Samsung will manufacture the Groq chip, and Nvidia expects the system to ship in the second half of this year. The launch builds on a roughly $20 billion deal struck with Groq in December, under which Nvidia licensed Groq’s technology and hired its top engineers.

Huang had previously hinted at the collaboration during an earnings call, and the Wall Street Journal had reported that Nvidia was preparing a Groq-based inference system. Nvidia projects at least $1 trillion in demand for its Blackwell and Rubin AI systems through 2027, up from about $500 billion projected through 2026.

nvidia, groq 3, lpx, groq, vera rubin, jensen huang, samsung, inference speedup, blackwell, groq deal