U.S. semiconductor corporations Nvidia signed a nonexclusive technology license agreement with AI accelerator chip design startup Groq, Groq said on the 24th (local time).
Groq said on its blog that the deal reflects a shared goal of expanding access to high-performance, low-cost inference technology, and the license reportedly covers Groq's inference technology broadly. Financial terms of the agreement were not disclosed.
As part of the agreement, some key staff, including founder Jonathan Ross and President Sunny Madra, will join Nvidia to support the development and scaling of the licensed technology, the company said. However, Groq will continue to operate as an independent corporations, and Simon Edwards will serve as chief executive officer (CEO), it added. Groq also said it plans to continue its cloud business without disruption.
Groq, founded in 2016, has focused on designing AI accelerator chips that speed up inference for large language models (LLMs). In Sep., it raised $750 million at a valuation of about $6.9 billion (about 10 trillion won), and it set this year's annual revenue target at $500 million (about 700 billion won). Founder Jonathan Ross is known to have worked on Google's Tensor Processing Unit (TPU) development.
Meanwhile, in the afternoon, U.S. business outlet CNBC drew industry attention by reporting that Nvidia agreed to acquire Groq for $20 billion in cash (about 29 trillion won). Citing comments from Disruptive CEO Alex Davis, who led Groq's recent funding round, CNBC reported the news; if accurate, it would be the largest merger and acquisition in Nvidia's history.
However, in announcing the licensing deal, Groq did not separately confirm any acquisition or price, and both sides have provided limited official comments regarding detailed transaction terms.