Intel Introduces Crescent Island GPU for AI Inference Workloads

Oct 20, 2025
Intel has announced a new data center GPU, code-named Crescent Island, optimized for AI inference tasks and designed for air-cooled enterprise servers. The GPU features 160GB of LPDDR5X memory and is based on the Xe3P microarchitecture, with customer sampling expected in the second half of 2026.

At the 2025 OCP Global Summit, Intel announced in a press release a new data center GPU code-named Crescent Island, designed for AI inference workloads. The GPU will offer high memory capacity and energy-efficient performance for enterprise servers.

Crescent Island is built on Intel’s Xe3P microarchitecture, a performance-optimized version of its Xe3 GPU architecture, and includes 160GB of LPDDR5X memory. The chip is optimized for power and cost efficiency in air-cooled environments and supports a wide range of data types, making it suitable for inference and “tokens-as-a-service” applications.

Intel said it is developing an open and unified software stack for heterogeneous AI systems, currently being tested on its Arc Pro B-Series GPUs, to enable early optimizations. Customer samples of Crescent Island are expected in the second half of 2026.

The announcement represents Intel’s continued expansion of its AI accelerator portfolio, following its earlier Gaudi 3 AI accelerator and Arc Pro GPU lines aimed at addressing inference performance across data centers and enterprise environments.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.

Also, consider following us on social media:

AI Chips & Datacenters AI Brief AI Brief (X)
Free newsletter

Silicon Brief

Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.