Nvidia has entered a non-exclusive licensing agreement with Groq, an AI chip designer known for its language processing units (LPUs), to accelerate inferencing technology. As part of the deal, Nvidia has also hired several senior Groq executives, including founder Jonathan Ross and former president Sunny Madra. “We’ve taken a non-exclusive license to Groq’s IP and have hired engineering talent from Groq’s team to join us in our mission to provide world-leading accelerated computing technology,” an Nvidia spokesman said.
Groq’s LPUs are designed for AI inferencing and are lower-powered and lower-priced compared to Nvidia’s GPUs, which are primarily used for training AI models. Groq also operates an inferencing-as-a-service business called GroqCloud and continues to run independently under new leadership, with Simon Edwards as CEO.
The licensing agreement allows Nvidia to diversify its memory sourcing and reduce dependence on high-bandwidth memory, which is in short supply. Groq integrates static RAM (SRAM) into its chips, offering improved performance and lower power use than dynamic RAM, while avoiding current memory shortages. By structuring the deal as IP licensing and hiring engineers, Nvidia avoids acquiring GroqCloud or facing potential antitrust scrutiny. Nvidia did not provide details on all the roles of the hired Groq executives, though Ross is now chief software architect, and Madra serves as VP of hardware.