Nvidia is making a strategic leap in the critical AI inference arena through a monumental deal valued at approximately $20 billion to license the technology and acquire the core engineering team of pioneering chip startup Groq.

This landmark agreement, exclusively reported by CNBC, stands as Nvidia’s largest on record and underscores the intensifying battle for supremacy in powering the next generation of artificial intelligence applications.

The move follows an emerging industry pattern where tech giants pursue massive “acqui-hire” styled transactions to rapidly integrate specialized talent and innovative technologies without formal corporate acquisitions.

While Nvidia’s hardware dominates the market for training massive AI models, the sector for inference, where these trained models deliver real-time responses to user queries is fiercely competitive.

Groq has distinguished itself with cutting-edge technology optimized specifically for high-speed, efficient AI inference, positioning it as a key challenger to Nvidia alongside established rivals like AMD and other startups. By bringing Groq’s expertise in-house, Nvidia directly bolsters its inference capabilities, aiming to solidify an end-to-end dominance across the entire AI development and deployment lifecycle.

Under the terms of the deal, Nvidia gains a non-exclusive license to Groq’s intellectual property while welcoming Groq founder Jonathan Ross, President Sunny Madra, and key engineers to its ranks. Notably, Groq will continue to operate independently under CEO Simon Edwards, maintaining its cloud business.

This structure mirrors recent high-value talent-and-technology transfers in Silicon Valley, such as Microsoft’s $650 million licensing pact with an AI startup and Meta’s significant investment to secure Scale AI’s leadership, all designed to navigate regulatory scrutiny while securing vital innovation.