Groq

Groq operates an AI inference cloud for developers and enterprises running AI models. Its platform combines three layers: GroqMetal supplies dedicated bare-metal infrastructure, GroqCore adds a production inference stack, and GroqAssured adds enterprise governance and auditability. Customers can control dedicated computing capacity or use the inference stack without managing the underlying infrastructure themselves.

Founded by Jonathan Ross and Douglas Wightman, Groq grew from custom chip development into cloud infrastructure. Adam Winter became CEO in 2026. Its Language Processing Unit, or LPU, uses a compiler to schedule computation and data movement across chips, with deterministic execution and on-chip SRAM. This software-controlled architecture makes execution timing predictable and places memory alongside compute, rather than relying on runtime scheduling to coordinate each operation.

In December 2025, NVIDIA entered a non-exclusive licensing agreement for Groq’s LPU technology and hired certain employees. The transaction involved $17 billion in consideration, including imputed interest; NVIDIA purchased no equity interests, existing products or customer contracts. Groq continues operating its cloud business. In August 2026, the company reported more than six million developers, thousands of AI-native companies and Fortune 500 enterprise customers, with 13 data centers across North America, Europe, the Middle East and Asia-Pacific.