Low-latency inference chips and a cloud, partly licensed to Nvidia.
60OVRBenchDraft · rated on 3 of 6
Ratings
Groq licensed its inference technology to Nvidia in a deal reported at about $20bn, which faces a reported Justice Department review and a Delaware stockholder suit.
Dominance55draft
A practice-squad inference chip maker with no stated share, whose technology Nvidia licensed in a deal reported at about $20bn.
From the Atlas key figure: Non-exclusive licence of its inference technology to Nvidia (Dec 2025), reported at about $20bn. Its founder and team joined Nvidia. Groq continues as a separate company. Source: groq.com.
Low-latency inference chips and a cloud, partly licensed to Nvidia.
Silicon · AI accelerators · Practice squad
Key figure Non-exclusive licence of its inference technology to Nvidia (Dec 2025), reported at about $20bn. Its founder and team joined Nvidia. Groq continues as a separate company.
Why it matters Nvidia paid for low-latency inference technology it did not have.
What could change it A reported Justice Department review and a Delaware stockholder suit over how the deal was structured.