NVIDIA Senior Director Dion Harris stated that Groq racks will be deployed alongside Vera CPUs and Rubin GPUs at new cloud provider Nebius, going live later this year. NVIDIA acquired the assets of chip startup Groq for $20 billion last December, marking its largest acquisition ever. Groq chips are manufactured by Samsung, while NVIDIA's GPUs are manufactured by TSMC. NVIDIA packages 256 individual Groq 3 chips into its LPX rack, claiming its Groq 3 LPX rack can deliver 3,400 tokens per second. Low-latency inference is increasingly important in AI, especially in coding, to ensure the responsiveness of AI agents.