Intel plans to ship its new “Crescent Island” graphics processing unit by the end of this year, positioning it as a lower-cost alternative to rival offerings from Nvidia and AMD in the AI inference market.
The chip uses LPDDR5 memory and air-cooling technology, avoiding the expensive high-bandwidth memory and liquid-cooling infrastructure required by competitors like Nvidia’s Blackwell series.
“We decided to start rebuilding our muscles in AI… we are not particularly aiming for [the training market] based on past experience,” said Kevork Kechichian, who leads Intel’s data center group.
This focus on inference—when users make requests rather than training models—comes after Intel’s previous AI GPU efforts, including the Gaudi chip which saw limited commercial success, led to the cancellation of its successor.
The new chip represents Intel’s first major push into AI infrastructure under CEO Lip-Bu Tan, who took over last year following challenges with the previous leadership’s turnaround strategy.
Crescent Island will begin limited customer shipments by year-end after an 18-month development cycle.