In what analysts see as a move to compete more directly with major chipmakers, semiconductor and telecommunications equipment company Qualcomm Technologies Inc. introduced its new AI inference solutions for data centers: Qualcomm AI200 and AI250 chip-based accelerator cards and rack systems.
According to Qualcomm, the solutions are designed to deliver high performance, energy efficiency, and scalability for running generative AI applications.
“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 and GM for Technology Planning, Edge Solutions and Data Center at Qualcomm Technologies Inc.
Both rack systems include direct liquid cooling for better heat management, PCIe for scaling up, Ethernet for scaling out, confidential computing for secure AI workloads, and a rack-level power capacity of 160 kW.
The Qualcomm AI200 features a rack-level inference setup built for low total cost of ownership (TCO) and optimized performance for large language and multimodal model inference. It supports 768 GB of LPDDR per card, providing high memory capacity and flexibility for AI workloads.
The AI250 will use a new memory architecture based on near-memory computing, which Qualcomm said can offer over 10 times higher effective memory bandwidth and lower power use. The design supports disaggregated AI inference, allowing more efficient hardware utilization while meeting performance and cost targets.
“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,” Malladi said.
The software stack supports popular machine learning and generative AI frameworks, along with optimization tools such as disaggregated serving. Developers can deploy models directly from Hugging Face using Qualcomm’s Efficient Transformers Library and AI Inference Suite.
Qualcomm said the AI200 is expected to be available commercially in 2026, followed by the AI250 in 2027.