Intel Unveils Crescent Island GPU for AI Inference Workloads
Intel has launched its next-generation Data Center GPU, codenamed Crescent Island, targeting inference workloads with 160GB of LPDDR5X memory and Xe3P Celestial architecture. The chipmaker confirmed a 2026 release for the power-efficient accelerator, positioning it as a contender in the AI infrastructure race.
The Xe3P microarchitecture—a refined variant of Intel's Panther Lake processor design—will also power future Arc C-Series GPUs. While performance metrics remain undisclosed, the memory configuration suggests enterprise-scale capabilities for cloud providers handling diverse AI data types.