Intel launches new GPU line for AI workloads

On October 13, 2026, Intel announced the official naming of its new line of data center inference-optimized GPUs, based on the Xe3P architecture, as the Intel GPU D-series. The first product in this series is the Intel GPU D270, which features 32 Xe cores, 256 XMX engines, and a FP8 performance of 1.5TB/s. It is equipped with up to 480GB of LPDDR5X memory, offering a bandwidth of 1.5TB/s. The card consumes 350W of power and uses the standard PCIe AIC form factor, supporting PCIe Gen5. It is designed for large language models, long-context workloads, and concurrent inference services, with single-card support for 300B models and an 8-card system capable of handling 1T models. Intel also showcased a new rack design with Xeon 6+ processors, capable of scaling to over 46,000 CPU cores with a power consumption of up to 133.45 kW. The Intel GPU D-series is intended for applications such as multi-tenant AI services, robotics, and other latency-sensitive workloads. This launch marks a significant step in Intel's expansion into the AI hardware market, emphasizing its commitment to providing high-performance solutions for data center environments.