On March 16, analyst Ming-Chi Kuo's latest supply chain survey indicated that following Nvidia's investment in Groq, LPU shipment forecasts have been significantly revised upward, with total shipments expected to reach 4 to 5 million units from 2026 to 2027. To maintain its ultra-low latency advantage during the inference decoding phase, Nvidia plans to increase the number of LPUs per rack from 64 to 256. Mass production of the new architecture racks is expected to commence in the fourth quarter of this year, with shipments reaching approximately 300 to 500 units in 2026 and increasing to 15,000 to 20,000 units in 2027.
