Groq raises $650m, pivots from chips to an inference cloud
After Nvidia licensed its LPU IP and poached its founders, Groq re-staffs and bets on neocloud capacity
Summary
AI-chip company Groq confirmed a $650m funding round led by Disruptive and Infinitum, TechCrunch reported, as it re-staffs and reinvents itself as an inference cloud operator. The raise follows Nvidia's December 2025 non-exclusive licensing of Groq's LPU technology, a roughly $20bn arrangement that took founder Jonathan Ross and president Sunny Madra to Nvidia. Groq named Alan Rice, ex-xAI and Meta, as COO and added a new CTO and CPO. It now runs 13 data centers across four regions and aims to quadruple capacity to 200MW by end-2027, serving inference demand that vendors like Openai are racing to lock down.
Why it matters
Groq's story is what happens after a "non-acquisition": Nvidia took the IP and the founders, leaving Groq to compete in inference using technology its largest rival now owns. The pivot from selling chips to renting compute mirrors a wider shift where inference capacity, not silicon design, is the scarce asset.
What to watch
- Whether Groq hits the 200MW target on schedule.
- How its neocloud margins hold against hyperscaler pricing.
- Whether Nvidia's LPU license erodes Groq's technical edge.