Groq confirms $650M raise to rebuild as an AI-inference neocloud after Nvidia gutted it
Six months after Nvidia's $20B licence-and-hire deal stripped out its founders, Groq bets its second act on cloud, not chips
加入列表
还没有列表。
Summary
Groq confirmed a $650M growth round on 22 June 2026, led by Disruptive and Infinitum, to rebuild itself as an AI-inference neocloud rather than a chip vendor. The pivot follows a December 2025 deal in which Nvidia paid a reported $20B to non-exclusively licence Groq's LPU technology and hire away founder-CEO Jonathan Ross and other senior staff, a "not-acqui-hire" that hollowed out the hardware business. Groq says its cloud now spans 13 data centers across North America, Europe, the Middle East and APAC, serves over five million developers and processes trillions of tokens weekly. The company did not disclose a new valuation; it was last marked at $6.9B after a $750M round in September 2025.
The split
US venture coverage (TechCrunch, Axios) treats this as a survival-and-reinvention story: Nvidia took the silicon and the founders, Groq keeps the cloud and the brand. The infrastructure press (DatacenterDynamics) foregrounds capacity, the Middle East and APAC sites built on sovereign-AI demand. What goes mostly unsaid in the upbeat framing is how unusual it is to raise $650M months after your founding CEO and core IP walked to your largest rival.
By the numbers
- $650M, growth round size.
- $20B, reported value of Nvidia's December 2025 licence-and-hire deal.
- $6.9B, last disclosed valuation (September 2025).
- 13, data centers across four regions.
- 5M+, developers on the platform.
- Trillions of tokens, processed weekly, per Groq.
Why it matters
Inference, not training, is where the next compute margin sits, and Groq is a test of whether a startup can hold an inference-cloud niche against Nvidia, which now owns its old hardware edge. The Nvidia deal also shows the chip giant buying capability through licences and hires rather than acquisitions, sidestepping antitrust scrutiny.
What to watch
- A disclosed post-money valuation, the tell on whether this was a markup or a flat raise.
- Whether Nvidia's own "Groq 3 LPX" inference hardware undercuts Groq's cloud on price.
- Middle East and APAC capacity additions tied to sovereign-AI contracts.