< Back to all clusters
[TECHNOLOGY] · South Korea, United States · 13 sources

started · updated

NVIDIA enters full production of Groq 3 LPX AI inference chips

NVIDIA has announced that its Groq 3 LPX inference accelerator is now in full production. This technology, integrated into the Vera Rubin platform, is designed to provide ultra-low latency for agentic AI systems, such as those used for real-time coding and complex reasoning.

Following a $20 billion asset acquisition from the chip designer Groq, NVIDIA is deploying the Groq 3 LPX racks, which consist of 256 language processing units (LPUs) and utilize 500 MB of on-chip SRAM to minimize memory bottlenecks. In benchmarks using the Gemma 4 31B model, the system achieved a record 3,400 output tokens per second.

Key customers and partners include Nebius, which will be the first AI cloud to adopt the technology, and SpaceXAI, which plans to deploy NVIDIA Vera CPUs to support its AI infrastructure, including potential expansion into space-based data centers via the Starmind satellite. The Groq 3 LPX chips are being manufactured by Samsung Electronics' foundry division.

Entities

Groq · Jensen Huang · Microsoft · Nebius · Nvidia · Samsung Electronics · SpaceXAI · Vera Rubin

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

about 7 hours ago