Home / Technology / Nvidia Unveils Groq 3 LPU for AI Chatbots
Nvidia Unveils Groq 3 LPU for AI Chatbots
17 Mar
Summary
- Nvidia introduces the Groq 3 LPU chip for optimized AI chatbot performance.
- The LPU licenses technology from AI company Groq, founded in 2016.
- LPUs will work with Nvidia's Rubin GPUs for enhanced AI model computations.
Nvidia is set to release its new Groq 3 LPU processor, specifically designed to enhance the performance of large language models and AI chatbots. This development stems from Nvidia's December licensing agreement with Groq, an AI company established in 2016 that specializes in LPU technology for speed and energy efficiency.
The Groq 3 LPU will complement Nvidia's upcoming 'Vera Rubin' platform, which includes Rubin GPUs and Vera CPUs for data centers. While LPUs use faster SRAM with limited capacity, they will be deployed in large batches alongside Nvidia's GPUs. This combined approach aims to optimize AI workloads, particularly for longer prompts, with Nvidia projecting a potential 35x throughput increase for trillion-parameter models.
Nvidia's CEO Jensen Huang confirmed that the Groq chip is in production and expected to ship in Q3, with Samsung manufacturing the LPU. Analysts anticipate Nvidia shipping millions of LPUs through 2026 and 2027. The technology, likely costing tens of thousands of dollars per chip, is targeted at major AI companies such as OpenAI, Anthropic, and Meta, potentially powering future chatbot interactions and image generation.
Nvidia also highlighted its Vera Rubin platform, which has been in full production since January. The company plans to ship the Rubin-related chips, including the new LPU, in the second half of 2026. This strategic integration of LPUs and GPUs signifies Nvidia's continued push for dominance in the AI hardware market.
