Intel: Our upcoming AI chip will be cheaper, faster than Nvidia, AMD solutions

Intel plans to ship an AI chip later this year that uses cheaper technology and cooler technology than offerings from Nvidia and AMD, as the US chip maker looks to capitalize on a major shift in its economy.

Kevork Kechichian, who heads Intel’s data group, told the FT that the company is “starting with the basics” as it tries to challenge its competitors in the growing market for AI-powered semiconductors.

Its new “Crescent Island” unit is designed to speed up the “inference” tasks, the stage when the user requests, instead of training the models, the places where Nvidia processors are big.

An early attempt to create a GPU for training AI models called “Gaudi” saw poor sales, and its successor was canceled last year.

“We decided to start flexing our muscles in AI … (but) we don’t really want (the training market) to rely on past experiences,” said Kechichian, who joined Intel last year from hardware maker Arm.

He added that the new chip will begin shipping to customers later this year, following an 18-month development plan.

Intel also wants to take advantage of two challenges that Nvidia and AMD face: the need to combine high-bandwidth memory and water-cooling architecture.

Crescent Island is an air-cooled chip that uses LPDDR5 memory, a much cheaper memory than the HBM used in chips like Nvidia’s Blackwell.

The effort is Intel’s first push into the lucrative AI market under CEO Lip-Bu Tan, who took over last year after Pat Gelsinger was ousted over concerns that his turnaround strategy was failing.