Nvidia pushes back on The Information's report about its Groq 3 LPU

Nvidia pushes back on The Information's report about its Groq 3 LPU

Nvidia integrates Groq 3 LPU (150 TB/s, 500 MB on-chip SRAM) into Vera Rubin platform to handle low-latency inference for agentic AI alongside GPUs. This consolidates fast-inference access within Nvidia's stack, reducing differentiation for alternative chip startups and strengthening Nvidia's end-to-end AI control.