Nvidia: Contract concluded with inference chip startup

Nvidia has agreed to a licensing deal with chip startup Groq, and Groq employees will also move to Nvidia as part of the $20 billion deal.

listen Print view
Sign of Nvidia at new headquarters in Silicon Valley.

(Image: JHVEPhoto/Shutterstock.com)

2 min. read

Nvidia apparently concluded a contract worth around $20 billion with a startup called Groq on Wednesday, as part of which Groq CEO and founder Jonathan Ross and other employees of the startup will move to Nvidia. Both companies describe the deal as non-exclusive, and Groq is expected to remain independent as a brand. The contract is described as the largest deal in Nvidia's company history.

Groq was founded in 2016 by the creators of Google's Tensor Processing Unit. The startup manufactures chips designed for inference (Language Processing Architecture). This refers to the operation of AI models, i.e. what happens "under the hood" when using trained AI models, for example, asking something, requesting predictions, or drawing conclusions from data.

Groq chips are reportedly up to ten times faster in inference than conventional GPUs, where Nvidia is considered the market leader. Experts expect the greatest demand for chips in the future for operating AI models, which was likely a driving reason for the contract conclusion. Nvidia's own chip technology has so far shone more in the area of training AI models.

Videos by heise

In an internal Nvidia email, CEO Jensen Huang wrote, according to the website of the US business and financial news channel CNBC, that they want to integrate Groq's chips into Nvidia's factory architecture to serve "an even broader range of AI inference and real-time workloads."

Nvidia has experienced enormous growth in recent times – at the end of October, according to CNBC, the company had $60.6 billion available in "cash and short-term investments," i.e. money that Nvidia can simply spend, compared to $13.3 billion at the beginning of 2023. (kst)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.