[NVIDIA's $20 Billion Was Well Spent: Groq Chips Officially Mass-Produced] Beating AI Newsflash: NVIDIA's Groq 3 LPX has entered full-scale mass production. In December last year, NVIDIA spent approximately $20 billion to acquire the licensing rights to Groq's technology and brought on board key members of the team, including founder Jonathan Ross. Eight months later, the first hardware product from this deal has officially launched. The Groq 3 LPX is a system specifically designed to accelerate large model inference, featuring 256 Groq 3 LPU chips. In Artificial Analysis testing, after inputting 100,000 tokens into the Gemma 4 31B model, the Groq 3 LPX achieved an output rate of 3,431 tokens per second. In the same test, the fastest publicly available API at the time could only handle about 870 tokens per second, making the Groq 3 LPX nearly four times faster. In practical deployment, Rubin GPUs handle the heavier computations, while the Groq 3 LPX focuses on speeding up token generation. For Coding Agents, this significantly reduces the waiting time for the model to "speak" during consecutive calls, whether dozens or hundreds of times. Nebius will become the first AI cloud provider to adopt the Groq 3 LPX. Interestingly, Groq itself has also become one of the first adopters and will collaborate with Dell for deployment. After licensing its technology to NVIDIA last year, Groq is now using the "Groq" hardware produced by NVIDIA this year. [Original Link]