Groq has raised $650 million to rebuild itself following Nvidia's dominance.

      Groq has announced a $650 million funding round aimed at transforming itself into an AI inference cloud, six months after Nvidia compensated its investors and recruited its founder. The $650 million round is a wager on the superiority of specialized chips over GPUs.

      For years, Groq was one of Nvidia's most vocal competitors, but in December, Nvidia significantly weakened the company. Now, Groq has revealed its strategy for revival, raising $650 million in a round led by Disruptive and Infinitum, with existing investors also reinvesting.

      The funding was initially reported in May, and the company has now finalized it and identified the team responsible for utilizing the capital.

      What Nvidia left behind

      In December, Nvidia secured a non-exclusive licensing agreement for Groq’s chip technology, reportedly valued at around $20 billion. It recruited founder and CEO Jonathan Ross, president Sunny Madra, and several engineers. Nvidia has since integrated this technology into its own product lineup, launching the Groq 3 LPX inference system at its GTC event in March, which is based on the licensed designs. This left Groq with its data centers, software, and a challenging question: what is its new identity?

      A pivot to the cloud

      The response is focused on inference—the operation of trained AI models rather than their creation. Groq is firmly investing in what it terms its “neocloud” division. The company currently operates 13 data centers across North America, Europe, the Middle East, and Asia-Pacific, claiming to serve over five million developers and process trillions of tokens weekly at the sites hosting those models.

      The new funds will enhance its infrastructure with the latest hardware, including Nvidia’s LPX system. Groq aims to increase total capacity to 200 megawatts by the end of 2027.

      New leadership

      Adam Winter has been appointed as CEO, with Matt Eng as the chief financial officer and Disruptive founder Alex Davis as chairman. Alan Rice joins as COO, having previously worked on xAI's Colossus project, at Meta’s data centers, and in the US Navy submarines. From July, Sinclair Schuller will become CTO, and Rakesh Malhotra will serve as chief product officer. Schuller and Malhotra previously co-founded and sold enterprise software companies Apprenda and Nuvalence, with Malhotra having spent about ten years working on Microsoft's cloud products.

      The same investors, again

      An interesting aspect of this funding round is that many of the investors involved are the same ones who benefited when Nvidia cashed out in December. Disruptive and Infinitum, both of which hold board seats, led this round. Earlier investors have included Samsung, Cisco, and BlackRock. Groq has not revealed a new valuation, but its last valuation stood at $6.9 billion, following a $750 million round in September.

      A crowded, expensive market

      Groq is entering a highly competitive market. Demand for inference is rapidly increasing, and significant investments are flowing into this space. AI infrastructure startup Baseten recently secured $1.5 billion at a valuation of up to $13 billion. Leading labs like OpenAI and Anthropic continue to drive up computing costs, heightening the need for cost-effective methods of model operation.

      The bottom line

      Groq's proposition is straightforward: as AI transitions from model training to operation, purpose-built inference chips should outperform general-purpose GPUs in both speed and cost. According to Groq, the inference process will ultimately require 15 to 20 times more computing power than training. The challenge is equally clear—Groq now shares its core chip intellectual property with Nvidia, the very company it aimed to compete against. Whether a streamlined Groq can still lead in the inference market is the pivotal question at the heart of the $650 million investment.