Nvidia has reached a $20 billion deal with AI chip startup Groq to license its AI inference hardware, while also securing key talent, including the company's founder. The unexpected move highlights the escalating battle for dominance in the rapidly expanding AI chip market.
Groq on Wednesday, December 24, said that it has entered into a non-exclusive licensing agreement with Nvidia in order to provide expanded access to its high-performance, low-cost inference chips. To help advance and scale the licenced technology, Groq CEO Jonathan Ross, president Sunny Madra, and other members of the AI chip startup will be joining Nvidia. Groq’s new CEO will be Simon Edwards, and it will continue to operate as an independent company even after the Nvidia partnership.
Although the complete financial specifics have not been revealed, Nvidia's $20 billion licensing agreement with Groq is said to be its largest technology acquisition to date, nearly three times Groq’s $6.9 billion valuation after a $750 million funding round earlier this year. This non-exclusive deal signals important shifts within the AI sector, with two key developments emerging:
Nvidia’s changing dominance: The deal comes at a time when Nvidia’s control over the AI chip market is being challenged. It highlights Nvidia's strategy to license and absorb inference technology from emerging competitors, as the focus shifts from training-centric GPUs to custom silicon for AI inferencing.
Acqui-hires in AI: The agreement underscores the increasing importance of in-house talent and specialized expertise in the AI industry, where acquiring both technology and skilled personnel is becoming as critical as access to hardware itself.
Since Nvidia's AI chip business skyrocketed in mid-2023, its quarterly revenue has surged from $7 billion to $57 billion, driven by demand from hyperscalers like Google, Microsoft, Amazon, Meta, and Oracle. However, Nvidia’s success has spurred competition from rivals like AMD, Broadcom, and AI startups such as Groq and Cerebras. Major tech firms like Google and Apple are also developing in-house chips to reduce reliance on Nvidia’s GPUs.
Despite competition, Nvidia maintains an edge due to the quality of its chips. However, Google's Gemini 3 AI models, built on its custom Tensor Processing Units (TPUs), are emerging as a potential challenge. In response, Nvidia has ramped up investments in chip startups, amassing $60.6 billion in cash and short-term investments by October 2025.
Nvidia has made strategic acquisitions, such as buying Mellanox for $7 billion in 2019, and investing in companies like Enfabrica, Crusoe, Cohere, and CoreWeave, while also striking a $100 billion deal with OpenAI for chip deployment.
Groq, a US-based startup founded in 2016, specializes in AI inference chips called LPUs (Language Processing Units), designed to optimize pre-trained models for real-time AI tasks. Founded by ex-Google engineer Jonathan Ross, Groq's products are used by major companies like Meta, Google, and OpenAI to improve AI performance at lower costs.