Jonathan Ross

Groq’s December 24, 2025 announcement names Ross among staff joining NVIDIA to advance licensed inference technology. The agreement is non-exclusive: Groq says it will keep operating independently, with Simon Edwards becoming CEO and GroqCloud continuing to serve customers. The distinction matters because the technology license and staff transfer do not mean Groq itself became a NVIDIA subsidiary.[1]

Ross’s work connects custom AI processors at Google with Groq’s Language Processing Unit. Inference means running an already trained model to answer a request. Groq built its processor and cloud around this execution stage, where response time affects whether an application can feel interactive.[2][3]

In his Masters of Scale conversations, Ross frames speed as part of the product experience rather than a hardware statistic alone. A model that returns answers quickly can support applications that wait for several model calls or respond to a live conversation. His earlier TPU work and Groq’s LPU both use specialized hardware for neural-network computation, but the company’s inference focus addresses serving trained models rather than only the cost of training them.[2][3]