Nvidia has struck a major deal to license technology from AI chip startup Groq and hire its CEO, Jonathan Ross. The move signals Nvidia’s aggressive push to control the “inference” market—the phase where AI models actually generate answers for users—rather than just the training phase where it already dominates.
Ross, a veteran who helped launch Google’s AI chip program, will join Nvidia along with Groq President Sunny Madra and a team of engineers. Despite losing its top leadership, Groq stated in a blog post that it will continue as an independent company. Simon Edwards will step in as the new CEO to run Groq’s remaining cloud business.
This arrangement follows a growing trend in Silicon Valley known as “acqui-hiring.” Tech giants like Microsoft, Amazon, and Meta have recently paid massive sums to license technology and hire founders from startups such as Inflection and Adept AI without formally acquiring the companies. This strategy allows them to absorb top talent and intellectual property while dodging the intense antitrust scrutiny that comes with full mergers. Analysts suggest this “non-exclusive license” structure helps maintain the appearance of competition for regulators, even as Nvidia absorbs Groq’s driving force.
The deal addresses a specific technical challenge for Nvidia. While Nvidia’s chips are the gold standard for teaching AI models, startups like Groq and Cerebras have challenged them on speed when the AI interacts with users. Groq uses a unique architecture with on-chip memory (SRAM) that skips the need for external high-bandwidth memory. This makes their chips exceptionally fast for chatbots, though it limits the size of the models they can run.
By licensing this tech, Nvidia aims to shut out rivals like AMD. Although rumors circulated about a potential $20 billion acquisition, the companies did not confirm financial figures. However, with Groq recently valued at $6.9 billion, the move cements Nvidia’s determination to lead the next phase of the AI boom.
