Qualcomm has launched its AI200 and AI250 hardware offerings, targeting data center inferencing workloads.

Based on the company’s Hexagon neural processing units (NPUs) and customized for data center AI workloads, the new chip-based accelerator cards and racks have been optimized for AI inferencing.

The solutions will deliver rack-scale performance more efficiently with a lower TCO, Qualcomm said.

The AI200 rack supports 768GB of LPDDR per card of memory capacity, while the AI250 uses “innovative memory architecture” based on near-memory computing. Qualcomm said the AI250 will deliver more than ten times higher effective memory bandwidth and much lower power consumption, although it did not provide a baseline comparison for this figure.

Both the AI200 and AI250 offerings are cooled with direct liquid cooling technology, utilize PCIe interconnects for scale-up and Ethernet for scale-out, and offer 160kW rack-level power consumption. Qualcomm did not disclose information about the number of chips per rack or the compute performance the racks will offer.

Qualcomm also announced that Saudi AI venture Humain is planning to deploy 200MW of its AI200 and AI250 hardware in both the Kingdom of Saudi Arabia and other global locations.

The deployment will support the development of Humain’s AI ALLaM models alongside customer-specific solutions to address the needs of enterprises and government organizations across the Kingdom.

“With Qualcomm AI200 and AI250, we’re redefining what’s possible for rack-scale AI inference. These innovative new AI infrastructure solutions empower customers to deploy generative AI at unprecedented TCO, while maintaining the flexibility and security modern data centers demand,” said Durga Malladi, SVP & GM, technology planning, Edge solutions & data center, Qualcomm.

“Our rich software stack and open ecosystem support make it easier than ever for developers and enterprises to integrate, manage, and scale already trained AI models on our optimized AI inference solutions. With seamless compatibility for leading AI frameworks and one-click model deployment, Qualcomm AI200 and AI250 are designed for frictionless adoption and rapid innovation.”

The AI200 and AI250 solutions are expected to be commercially available in 2026 and 2027, respectively.

An unnamed, next-generation product in the Qualcomm rack-scale portfolio of products is targeted for 2028, as part of the company’s commitment to its data center roadmap with an annual cadence, Qualcomm said.