The Inference Coup: NVIDIA’s $20B Groq Deal Swallows the TPU’s Creator

Groq has announced that it has signed a non-exclusive licensing agreement with NVIDIA covering inference technologies. The deal is framed as a licensing arrangement rather than a conventional acquisition, accompanied by a substantial personnel shift as Groq’s founder Jonathan Ross, a former core developer of Google’s TPU, will lead part of Groq’s senior technical team directly into NVIDIA.

According to the announcement, the agreement is explicitly non-exclusive, meaning Groq will continue to operate as an independent company, while key figures including CEO Jonathan Ross and COO and President Sunny Madra move to NVIDIA to oversee the licensed technologies. Former CFO Simon Edwards will step in as Groq’s new CEO, ensuring continuity of operations, with GroqCloud remaining unaffected.

Founded in 2016 and based in Mountain View, California, Groq’s LPUs are designed for the inference phase of large language models, a contrast to NVIDIA’s GPUs; the move is described as an acquihire in effect, allowing NVIDIA to gain technology and expertise while preserving Groq’s nominal independence. 26 December 2025.