NVIDIA has agreed to pay approximately $20 billion to acquire assets from artificial intelligence chip startup Groq, marking the company’s largest transaction on record and continuing its strategy of absorbing potential competitors before they can challenge its market dominance.
The chipmaker’s latest licensing deal mirrors a similar transaction just three months ago, reinforcing the narrative that decentralized AI infrastructure may offer the only alternative to Nvidia’s growing dominance.
The deal closed just three months after Groq raised $750 million at a $6.9 billion valuation—a round that included BlackRock, Samsung, Cisco, and 1789 Capital, where Donald Trump Jr. serves as a partner. Nvidia is acquiring all of the company’s assets substantially, except its cloud computing business, though Groq framed the transaction as a “non-exclusive licensing agreement.”
Groq CEO Jonathan Ross, a former Google engineer who helped create the search giant’s Tensor Processing Unit, will join Nvidia along with president Sunny Madra and other senior executives. The startup will continue operating independently under CFO Simon Edwards as its new chief executive.
The Groq transaction follows a pattern Nvidia established just three months earlier. In September, the company paid over $900 million to hire Enfabrica’s CEO and employees while licensing the startup’s technology. Both deals use licensing structures rather than outright acquisitions, potentially avoiding the antitrust scrutiny that blocked Nvidia’s $40 billion bid for Arm Holdings in 2022.
The Kobeissi Letter summarized Nvidia’s approach bluntly: “We will buy you before you can compete with us.”
Groq’s Language Processing Unit uses on-chip SRAM rather than external DRAM, enabling what the company claims is up to 10x better energy efficiency. This architecture excels at real-time inference but limits model size—a tradeoff Nvidia can now explore within its broader ecosystem.
The timing is notable. Google recently unveiled its seventh-generation TPU, codenamed Ironwood, and released Gemini 3, trained entirely on TPUs, to top benchmark rankings. Nvidia responded on X: “We’re delighted by Google’s success… NVIDIA is a generation ahead of the industry—it’s the only platform that runs every AI model.” When incumbents start issuing such reassurance statements, competitive pressure is clearly mounting.