Intel launches D270 GPU for inference-optimized AI serving

Intel has announced the D270, a GPU designed specifically for AI inference in data centers. The accelerator features up to 480 GB of LPDDR5X memory and is optimized for agentic AI applications, long-context processing, and high-concurrency workloads that require serving many requests simultaneously.

The D270 positions itself on practical power consumption and cost, aiming to compete with other inference-focused accelerators in the growing data center AI market. The announcement underscores Intel's strategy to capture demand from companies deploying large language models and AI agents in production environments.