Intel Names Its New Data Center GPU Line D-Series, First Model Is the D270

Intel announced at the 2026 OCP Global Summit that "Crescent Island", its data center inference GPU built on the Xe3P architecture, is now officially named "Intel GPU D-series". The first model is the Intel GPU D270.

The D270 packs 32 Xe cores and 256 XMX engines, delivering 1.5TB/s of FP8 compute. It supports up to 480GB of LPDDR5X memory with 1.5TB/s bandwidth. The card draws 350W, supports PCIe Gen5, and works with air cooling.

The D-series is built for large language models, long-context workloads, and concurrent inference. A single card can run a 300B-parameter model, while an 8-card system can handle a 1T-parameter model. It targets multi-tenant AI services, robotics, and other latency-sensitive tasks.

Intel also showed new rack designs with partners, based on Xeon processors. Some Xeon 6+ setups can scale to over 46,000 CPU cores, with power up to 133.45 kW.