US Markets
Home›US Markets›M&A & Deals›Nvidia ramps Groq 3 LPX rack production for Nebius dep…
Nvidia ramps Groq 3 LPX rack production for Nebius deployment
The liquid-cooled rack packages 256 Groq chips and is expected to deliver 3,400 tokens per second at neocloud Nebius later this year.
NVIDIA’s Groq 3 LPX rack is now in full production and will be deployed alongside its Vera CPUs and Rubin GPUs at neocloud Nebius later this year, according to CNBC, citing comments from Nvidia senior director Dion Harris.
The rack represents the commercialization phase of Nvidia’s $20 billion purchase of Groq’s assets in December, Nvidia’s largest deal on record. Each liquid-cooled rack packages 256 Groq chips, and Nvidia says the system can deliver 3,400 tokens per second, using a benchmark from Artificial Analysis.
The deal also involves manufacturing differences, with Groq chips produced by Samsung, while Nvidia’s own GPUs are made by TSMC. Nvidia positions Groq’s technology as extending its coverage across the AI compute stack, with LPX focused on low-latency inference while its GPUs continue to handle training and large-context processing.
CNBC also highlighted the operational speed of the integration, with Nvidia moving from announcing the Groq purchase in December to full production and a named customer in about eight months. The article further notes the 3,400 tokens per second figure compares with a 750 tokens per second promise for OpenAI’s Cerebras-powered “Ultrafast” mode.