Groq Used to Make Its Own AI Chips. Now It Sells Cloud Computing — Using Nvidia's.
Groq announced on August 17, 2026 that it raised $350 million in a funding round led by Disruptive, with Nvidia planning to participate. The deal values Groq at $3.5 billion.
That valuation is a steep drop from the $6.9 billion Groq was worth in September 2025, when it raised $750 million from investors including BlackRock and Neuberger Berman. It is also a fraction of the roughly $20 billion that CNBC reported on December 24, 2025, when it described Nvidia as acquiring the AI chip startup. What apparently happened instead was a licensing agreement: Groq let Nvidia use its inference technology, and Nvidia hired Groq's founder and CEO Jonathan Ross and other top talent.
Inference is the step where an AI model actually produces answers — think of it as the difference between studying for a test (training) and taking it (inference). Groq had built its own chips, called LPUs (language processing units), to compete with Nvidia on that inference step. After losing Ross and his team to Nvidia, Groq stopped designing its own chips and started running Nvidia's systems instead. It went from being Nvidia's rival to Nvidia's customer. On August 12, 2026, Groq became an official NVIDIA Cloud Partner.
The shift has been underway since at least June 2026, when Groq raised $650 million to scale its AI cloud business. The company now runs 13 data centers across North America, Europe, the Middle East, and Asia Pacific, serving more than 6 million developers, enterprises, and AI-native companies. Groq plans to grow from 54 megawatts to more than 200 megawatts of capacity by 2027.
Alex Davis, Groq's chairman and CEO of Disruptive, led the current round. Nvidia's planned participation means the company that hired away Groq's chip team is now backing Groq's cloud business.
Groq's cloud footprint has grown steadily over the past 18 months. The company partnered with Aramco on what was described as the world's largest AI data center in September 2024, launched a European data center in Helsinki in July 2025, and announced a Sydney facility in November 2025. Saudi Arabia committed $1.5 billion to an AI-powered economy expansion with Groq in February 2025. Groq has also partnered with Meta to deliver fast AI responses for the official Llama API, with Paytm for real-time AI in Indian payments, with McLaren Racing as a Formula 1 team partner, and with the U.S. Department of Energy to advance AI infrastructure. The company powered the NAIRR Pilot and was named exclusive inference provider for the Bell AI Network.
The funding history tells the story. Groq raised $640 million in August 2024 at a $2.8 billion valuation in a round led by Cisco Investments — when it was still a chipmaker. By September 2025, the $750 million round from BlackRock and Neuberger Berman more than doubled the valuation to $6.9 billion, still based on the idea that Groq's own chips were the future. Reuters reported around that time that Groq was speaking to investors about raising between $300 million and $500 million at a $6 billion valuation.
The actual outcome was different. The Nvidia deal stripped Groq of its chip team, the $6.9 billion valuation collapsed, and the company has now raised $350 million at half that figure.
The broader context here is that Groq's journey from AI chip competitor to Nvidia-dependent cloud provider fits a pattern the industry has been watching: it is very hard to build a standalone AI chip business when the dominant player — Nvidia — controls both the chips and the software that most developers use. Groq's chips were fast and attracted big-name partners, but the company ultimately could not keep up with the money and ecosystem needed to compete with Nvidia at the chip level. The cloud pivot keeps Groq alive and gives its 6 million developers a path forward, but it turns Groq from a company with its own unique technology into yet another seller of Nvidia's GPUs in a market that already includes CoreWeave, Lambda, and others.
It is worth noting that Nvidia appears on both sides of this story. The company licensed Groq's technology, hired its founding team, and is now a planned investor in a cloud business that buys Nvidia GPUs to serve the same developers who once used Groq's own chips. Whether that level of concentration helps or hurts the market for AI inference is a question the competitive landscape will answer over the next 18 months.
The $350 million gives Groq the runway to roughly quadruple its data center capacity by 2027. Whether the economics of running Nvidia-powered cloud services work at that scale, against established competitors and given how expensive GPU clusters are, is the question this round buys time to answer.
