Groq Raises $650M After Licensing Its Core Tech to Nvidia

Published by James Harris on June 22, 2026June 22, 2026

Groq has raised $650 million to scale its inference cloud business, roughly six months after licensing its core chip technology to Nvidia and watching its founder walk out the door with the deal. The round was led by Disruptive and Infinitum, with existing investors participating. No new valuation was disclosed; the company was last marked at $6.9 billion following a $750 million raise in September.

The Nvidia arrangement in December 2025 was a quiet but significant pivot point. Nvidia licensed Groq’s language processing unit technology on a non-exclusive basis, hired away co-founder and CEO Jonathan Ross, president Sunny Madra, and other staff, then unveiled its own LPX inference platform built on that technology at GTC in March. What Groq was left with was its cloud infrastructure: 13 data centers across North America, Europe, the Middle East, and Asia-Pacific, more than five million developers, and trillions of tokens processed weekly. The new capital will deploy Nvidia’s LPX systems, among other hardware, across those existing facilities. There is a particular irony in Groq now running the infrastructure that its licensed technology helped build.

The board chair is Alex Davis, founder and CEO of Disruptive, the lead investor, which is not a typical governance arrangement for a company at this stage.

Groq’s strategic bet is that inference, generating outputs from trained models in production, will dwarf training as the dominant compute market over time. The company cites an estimate that inference requires 15 to 20 times more compute than training across a model’s operational life. That framing is increasingly common among infrastructure players, but Groq is making the argument while simultaneously rebuilding its executive team from scratch: a new COO from xAI and Meta’s data center division, and a CTO and chief product officer joining in July who previously co-founded Nuvalence, a software engineering firm acquired by EY in 2024. Executing a capacity push to 200 megawatts by end of 2027 with a largely reconstituted leadership team is the actual test of whether the pivot holds.

The $650 million gives Groq runway to compete in a market where hyperscalers are spending at a scale that makes nine-figure rounds feel modest. The announcement frames inference demand as exponential and still early, which may be accurate, but the window for independent inference clouds to establish durable positioning before hyperscaler capacity catches up is narrowing faster than the fundraising timeline suggests.