Why Groq matters
In December 2025 Nvidia agreed to pay about $20 billion for a non-exclusive license to Groq's LPU inference architecture and hired founder Jonathan Ross and core engineers, while Groq kept its IP and continues to operate GroqCloud independently. At GTC in March 2026 Nvidia introduced the Groq 3 LPU, due to ship in the third quarter of 2026. Senators asked the FTC in 2026 whether the structure sidestepped merger review.
The numberNvidia paid about $20B in December 2025 to license Groq's LPU architecture and hire its leadership
On The All Things AI 100
Listed under Chips, compute, and AI clouds. The silicon and the data centers everything else runs on.
Quick answers
- What does Groq do?
- Runs GroqCloud, an inference service built on its LPU chips designed for fast token generation. In December 2025 Nvidia agreed to pay about $20 billion for a non-exclusive license to Groq's LPU inference architecture and hired founder Jonathan Ross and core engineers, while Groq kept its IP and continues to operate GroqCloud independently.
- What AI capabilities does Groq offer?
- Groq offers GroqCloud, an inference service that runs large language models on its language processing unit chips. Business users call an API to generate text with low latency for applications such as chat, summarization, and content generation. Groq supplies the hardware and hosted inference rather than end-user productivity software.
- Why is Groq on The All Things AI 100?
- Groq is listed under Chips, compute, and AI clouds. The silicon and the data centers everything else runs on. Nvidia paid about $20B in December 2025 to license Groq's LPU architecture and hire its leadership.
- Where is Groq based, and who owns it?
- Groq is based in Mountain View, CA, was founded in 2016 and is private; LPU architecture licensed to Nvidia (Dec 2025).