Qualcomm unveils AI200 and AI250 AI inference accelerators — Hexagon takes on AMD and Nvidia in the booming data center realm

Qualcomm has unveiled the AI200 and AI250 AI inference accelerators, set to launch in 2026 and 2027 respectively, to compete with AMD and Nvidia in the data center market. These accelerators are based on Qualcomm Hexagon NPUs and offer improved efficiency and lower operational costs for generative AI workloads. The AI200 will feature 768 GB of LPDDR memory, PCIe interconnects for scale-up, and Ethernet for scale-out scalability, with a power envelope of 160 kW per rack. The AI250, launching a year later, will include a near-memory compute architecture to boost memory bandwidth and support disaggregated inference capability. Qualcomm is also developing an end-to-end software platform optimized for large-scale inference, supporting major ML and generative AI toolsets for seamless model deployment.