Intel Crescent Island GPU Launch: Next-Gen Xe3P Graphics Architecture and 160 GB LPDDR5X for AI Inference

Intel has unveiled an innovative AI inference GPU solution for data centers, designated as Crescent Island. This cutting-edge product is based on the new Xe3P architecture, marking a significant step forward for the company in the realm of artificial intelligence.

Xe3P Architecture Enhances AI Inference Capabilities: Crescent Island Set to Release in 2H 2026

In response to previous challenges in meeting AI objectives within data centers, Intel is reinventing its approach. The company aims to enhance efficiency and expand memory capacity with this novel offering. Crescent Island was initially hinted at during the Intel Tech Tour 2025, generating anticipation within the industry.

This latest GPU harnesses the newly developed Xe3P architecture, which was showcased at the recent deep dives involving Panther Lake and Xe3. Crescent Island promises to be a notable advancement over the existing Xe3 architecture and will be integrated within the next-gen Arc family of products, particularly the Arc C-Series. This architecture is designed to be flexible, accommodating everything from integrated graphics processors (iGPUs) in consumer electronics to robust AI GPUs for data centers.

Intel Crescent Island GPU Unveiled: Features Next-Gen Xe3P Graphics Architecture, 160 GB LPDDR5X For AI Inference

The new data center GPU code-named Crescent Island is being designed to be power and cost-optimized for air-cooled enterprise servers and to incorporate large amounts of memory capacity and bandwidth, optimized for inference workflows. Key features include:

Xe3P microarchitecture with optimized performance-per-watt

160GB of LPDDR5X memory

Support for a broad range of data types, ideal for “tokens-as-a-service” providers and inference use cases

Intel’s open and unified software stack for heterogeneous AI systems is currently being developed and tested on Arc Pro B-Series GPUs to enable early optimizations and iterations. Customer sampling of the new data center GPU code-named Crescent Island is expected in the second half of 2026.

via Intel

Designed for both power efficiency and cost-effectiveness, the Crescent Island GPU is aimed specifically at air-cooled data center applications focused on AI inference workloads. According to Intel, the Xe3P architecture prioritizes performance per watt, which is crucial for maximizing operational efficiency.

The GPU boasts an impressive 160GB of LPDDR5X memory, differentiating it from competitors like NVIDIA and AMD, which are increasingly utilizing high-bandwidth memory (HBM) solutions such as HBM3E and are already discussing future HBM4 technologies for upcoming generations like Rubin and MI400. Rising demand for HBM has strained supply chains, leading to increased costs, making Intel’s LPDDR5X strategy a potentially advantageous position in terms of cost-performance balance.

Moreover, the Xe3P architecture will accommodate a diverse array of data types, positioning it well for “Tokens-as-a-Service” applications and various inference scenarios. This flexibility opens up new avenues for service providers and enterprise users alike.

Intel is concurrently refining its open and unified software stack tailored for heterogeneous AI systems, leveraging insights gained from its existing Arc Pro B-Series lineup. Such groundwork will enable forthcoming iterations to capitalize on early optimizations. Anticipation is building for the customer sampling of the Crescent Island GPU, slated for the second half of 2026, promising further revelations on its capabilities in the coming months.

Source & Images