Nvidia to Acquire AI Chip Startup Groq for $20B

According to a CNBC exclusive published Wednesday, Nvidia is said to have agreed to acquire U.S. AI chip startup Groq in a cash deal valued at around US$20 billion. If completed, the transaction would mark the largest acquisition in Nvidia's history and a major move in the fast-growing AI accelerator market.

People familiar with the matter said the deal was confirmed by Alex Davis, CEO of Disruptive and the lead investor in Groq's most recent funding round. Groq, founded in 2016, focuses on high-performance AI accelerator chips optimized for large language model (LLM) inference, an area where efficiency and latency are becoming increasingly critical.

Groq has been accelerating its global expansion and plans to build more than ten new data centers by 2026, surpassing the 12 facilities already completed this year. This comes despite earlier statements from Groq CEO Jonathan Ross, who said in early October that the company had no intention of being sold and that he had never sold any of his personal shares. By year-end, however, the situation reportedly shifted quickly, with Davis noting that Nvidia initiated the talks and that Groq was not actively seeking a buyer.

The startup raised US$750 million in September at an estimated valuation of US$6.9 billion, with investors including BlackRock, Neuberger Berman, Samsung, Cisco, Altimeter, and 1789 Capital. Groq is expected to brief investors on the transaction progress later the same day. Sources added that the acquisition would include Groq's core chip assets, while its early-stage cloud service, Groq Cloud, would be excluded.

If finalized, the deal would easily surpass Nvidia's previous largest acquisition—the nearly US$7 billion purchase of Mellanox in 2019. Nvidia's balance sheet supports such a move: as of the end of October, the company reported US$60.6 billion in cash and short-term investments.

Against the backdrop of surging demand for AI accelerators, Groq had reportedly set a revenue target of about US$500 million for this year. The market views its inference-focused architecture as a potential alternative to GPUs in certain workloads, especially where power efficiency and deterministic performance matter.

Groq was founded by a team of seasoned engineers, including CEO Jonathan Ross, who previously worked on Google's Tensor Processing Unit (TPU)—one of the earliest GPU alternatives for AI computing. Early regulatory filings also list Douglas Wightman, a former engineer from Google X's "moonshot factory," as a key member of the founding team.

In recent years, Nvidia has steadily expanded its investments across the AI ecosystem, backed by rising profits and cash reserves. These include stakes in AI and energy infrastructure company Crusoe, AI model developer Cohere, and cloud computing provider CoreWeave. In September, Nvidia also disclosed plans to potentially invest up to US$100 billion in OpenAI, alongside a commitment by OpenAI to deploy at least 10GW of Nvidia hardware, though no formal agreement has been announced. That same month, Nvidia revealed a US$5 billion investment in Intel as part of a strategic partnership.

Market observers say that acquiring Groq would further strengthen Nvidia's leadership in AI chips and inference computing, while potentially reshaping competition across the broader AI semiconductor landscape.

Nvidia shares traded slightly lower on Wednesday, briefly dipping to around US$186.6 before recovering to an intraday high near US$188.9. The stock later gave up gains to close at US$188.05, down about 0.3% on the day and below the US$189 level. Following reports of the Groq acquisition after hours, the shares extended their decline, slipping more than 0.3% at the time of writing.