Intel Xe Architecture Data Center GPU · Power- and Cost-Optimized for AI Inference · Air-Cooled Enterprise · Sampling H2 2026
Crescent Island (code name)
Intel Xe3P
160 GB LPDDR5X
AI Inference · agentic AI
Air-cooled enterprise
Broad range (FP/INT)
Tokens-as-a-service
H2 2026 (expected)
Data center GPU (PCIe)
Open unified Xe stack
Crescent Island is Intel's next-generation Xe-architecture data center GPU, purpose-built for AI inference as the industry shifts from static training to real-time, everywhere inference driven by agentic AI. Built on the Xe3P microarchitecture optimized for performance-per-watt, it pairs 160 GB of LPDDR5X memory with an air-cooled, power- and cost-optimized design for standard enterprise servers.
Announced at the 2025 OCP Global Summit, Crescent Island marks Intel's strategic pivot in data center AI — moving from the dedicated Gaudi ASIC line to a unified Xe GPU architecture that spans AI PC, edge, and data center. The design targets "tokens-as-a-service" providers with broad datatype support and high memory capacity for running large models efficiently, while Intel's open, unified software stack for heterogeneous AI systems is developed and validated on Arc Pro B-Series GPUs ahead of launch. Customer sampling is expected in the second half of 2026.
QS Compute can source Intel Crescent Island accelerators and Intel Xe data center GPU platforms for enterprise inference and agentic-AI infrastructure deployments.
Need Intel Crescent Island Accelerators?
Enterprise hardware — configured, tested, deployed. Contact us for pricing and availability.
Request Quote