Nvidia’s Groq 3 LPX racks go into full production after $20B Groq deal

Nvidia says its Groq 3 LPX rack is in full production following its $20 billion acquisition of Groq, to be deployed with Vera CPUs and Rubin GPUs at neocloud Nebius and go online later this year, underscoring the push for ultra-low-latency AI inference; the LPX rack uses 256 Groq 3 chips with 500 MB SRAM per chip and can deliver about 3,400 tokens per second, with Samsung manufacturing Groq chips and TSMC handling Nvidia GPUs as Nvidia ramps Vera Rubin shipments toward multi-year sales.