Intel Will Launch Crescent Island GPU for AI

The company debuted its new inference-focused hardware at the 2025 OCP Global Summit.

Updated on Oct. 12, 2026 in Semiconductors

Intel Will Launch Crescent Island GPU for AI

Live Poll

Do you believe new hardware innovations will make AI technology more affordable for your business?

Intel has introduced the Crescent Island GPU, an inference-focused processor designed to support real-time artificial intelligence workloads. The company expects to begin customer sampling for the new device in the second half of 2026.

Why it matters

The GPU aims to support agentic AI applications by delivering specialized inference processing capabilities. By integrating with existing Xeon processors and open software, the platform seeks to fit into standard air-cooled server environments.

The Crescent Island GPU features 32 Xe cores and 256 XMX engines, supported by 160 GB of LPDDR5X memory in the reference design. The card consumes 350 watts and supports a wide range of data types from FP4 to FP64.

The players

Intel

Intel is a multinational technology company and semiconductor manufacturer that designs and produces microprocessors and AI hardware solutions for global data centers.

OCP Global Summit

The OCP Global Summit is an annual industry event hosted by the Open Compute Project that focuses on collaborative innovation in data center hardware and infrastructure technologies.

The details

Utilizing the new Xe3P architecture, the hardware is built specifically for deployment in conventional, air-cooled server racks. The broader system platform can support up to 480 GB of memory to facilitate demanding real-time inference tasks.

Intel introduced the GPU at the 2025 OCP Global Summit.
Customer sampling is expected to begin in the second half of 2026.

The Tech Race

This development marks a shift toward specialized hardware designs that offload inference tasks from general-purpose CPUs. By building this GPU to complement the Xeon processor platform, Intel aims to solidify its position in the competitive AI data center hardware market.

Data centers deploying this hardware may see improved efficiency for AI inference tasks, potentially lowering the costs of real-time intelligent services. The air-cooled design allows for easier integration into existing server infrastructure compared to liquid-cooled alternatives.

The takeaway

Specialized inference chips like the Crescent Island GPU are becoming essential as demand for real-time AI agents grows in professional environments. Businesses should evaluate their current cooling and power infrastructure to determine if they can support these new high-density hardware designs.