Nvidia unveiled the Nvidia Groq 3 LPX inference server rack at its GTC 2026 conference. The system integrates specialized technology from AI chip startup Groq into Nvidia's hardware ecosystem.
The server rack features 256 Nvidia Groq 3 Language Processing Unit (LPU) inference accelerators. These components provide a collective 128GB of on-chip SRAM memory. This architecture prioritizes high-speed data access to accelerate complex AI workloads.
The launch supports Nvidia's strategy to meet rising demand for low-latency generative AI applications. The Groq 3 LPX is a component of the broader Vera Rubin platform. Nvidia expects the hardware to be available in the second half of 2026.