HeadlinesBriefing favicon HeadlinesBriefing.com

Nvidia's Groq Deal Aims to Secure A.I. Inference Lead

New York Times Top Stories •
×

Nvidia unveiled a new A. I. product at its GTC conference, combining its chips with Groq's specialized inference hardware to combat rising competition. The move follows a $20 billion licensing agreement struck last December, aiming to make inference faster and cheaper.

This shift addresses a critical market transition: A. I. workloads are increasingly focused on inference – generating responses – rather than training models, where Nvidia has historically dominated. Analysts see this as Nvidia fighting to maintain its 90% share of the A.

I. chip market, which has faced pressure from rivals like Google and Cerebras. The new product, developed rapidly over three months, is a direct response to customers like OpenAI and Meta seeking alternatives due to Nvidia's inference lag. Nvidia's ability to win over these customers will determine its share of the inference market, though its core revenue from training chips remains strong. The Groq partnership also offers a strategic supply chain advantage, as Groq's chips are manufactured by Samsung, bypassing TSMC's capacity constraints.