Groq

Groq is the AI hardware company founded by ex-Google TPU engineer Jonathan Ross, offering very high tokens-per-second inference via its LPU (Language Processing Unit) chips. Its architecture is built around clean, deterministic dataflow, which lets it run large language models at exceptionally low latency. It hosts open-weight models like Llama 3, Llama 4, Mistral, and Qwen 3 on its cloud. Alongside Cerebras it is one of the most visible faces of the TPS Wars and a credible alternative to GPU-based inference.