Intel unveils GPU D-series inference card and Xeon rack designs for agentic AI at OCP 2026

Intel

At the Open Compute Project Global Summit, Intel showed its GPU D-series, an air-cooled PCIe card built for memory-heavy AI inference, alongside partner Xeon rack designs. Intel says the card fits existing data centers without special racks or cooling, and that customer sampling is underway.

• GPU D-series (formerly Crescent Island), on the Xe3P architecture: up to 480 GB of LPDDR5X memory, 1.5 TB/s bandwidth and 1.3 petaflops of FP8 compute within 350 watts.
• Intel says one card runs models up to 300 billion parameters, and an eight-GPU system handles more than one trillion.
• The 480 GB figure applies to ODM-branded cards; the Intel-branded PCIe card has 160 GB of LPDDR5X.
• Select Xeon 6+ rack configurations scale beyond 46,000 cores per rack at 133.45 kW total rack power, many built on the Open Rack v3 standard.
• Performance varies by configuration and test date, and some features need enabled hardware, software or service activation.