Particle.news
Download on the App Store

Nvidia Licenses Groq Tech, Absorbs Core Team in Reported $20 Billion Inference Play

Jensen Huang plans to fold Groq’s low-latency processors into the company’s AI factory architecture, signaling a push to control real-time AI workloads.

Overview

  • Groq confirmed a non-exclusive license with Nvidia for its inference technology as founder Jonathan Ross and president Sunny Madra depart to join Nvidia, while Simon Edwards becomes Groq CEO and GroqCloud continues operating.
  • Reports indicate roughly 90% of Groq employees will move to Nvidia in a transaction widely described as an acqui-hire designed to capture IP and talent while limiting merger scrutiny.
  • In an email obtained by CNBC, Huang said Nvidia will integrate Groq’s low-latency processors into its AI factory platform to serve a broader range of inference and real-time workloads.
  • Multiple outlets peg the deal’s economic value near $20 billion, with payouts to Groq shareholders and employees reportedly structured as about 85% upfront, 10% in mid-2026, and the remainder at end-2026.
  • Analysts say the move bolsters Nvidia’s positioning in inference by pairing its GPU ecosystem with Groq’s LPU architecture optimized for ultra-low-latency, single-stream workloads.