TLDR: Qualcomm Technologies has unveiled its new AI200 and AI250 chip-based accelerator solutions, designed to deliver unprecedented rack-scale performance and energy efficiency for generative AI inference in data centers. These solutions promise industry-leading total cost of ownership (TCO) and feature an innovative memory architecture, comprehensive software stack, and seamless compatibility with major AI frameworks, with commercial availability expected in 2026 and 2027.
SAN DIEGO – Qualcomm Technologies, Inc. today announced a significant advancement in artificial intelligence infrastructure with the introduction of its next-generation AI inference-optimized solutions for data centers: the Qualcomm® AI200 and AI250 chip-based accelerator cards and racks. These new offerings are poised to redefine rack-scale AI inference performance, emphasizing high performance per dollar per watt, and are designed to enable scalable, efficient, and flexible generative AI across various industries.
Key highlights of these solutions include their ability to deliver rack-scale performance and superior memory capacity, crucial for rapid data center generative AI inference while ensuring an industry-leading total cost of ownership (TCO). The Qualcomm AI250, in particular, stands out with an innovative memory architecture that leverages near-memory computing. This design is projected to provide a generational leap in efficiency and performance for AI workloads, offering more than 10 times higher effective memory bandwidth and significantly reduced power consumption. This innovation facilitates disaggregated AI inferencing, optimizing hardware utilization to meet diverse customer performance and cost requirements.
Both the AI200 and AI250 solutions are built upon Qualcomm’s established Neural Processing Unit (NPU) technology. The AI200 is specifically engineered as a rack-level AI inference solution to provide low TCO and optimized performance for large language models (LLMs), multimodal models (LMMs), and other demanding AI workloads. It supports an impressive 768 GB of LPDDR per card, ensuring high memory capacity and cost-effectiveness, which translates to exceptional scalability and flexibility for AI inference tasks.
Durga Malladi, SVP & GM, Technology Planning, Edge Solutions & Data Center, Qualcomm Technologies, Inc., commented on the launch, stating, “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.” Malladi further highlighted the ease of adoption, adding, “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.”
From a technical standpoint, the rack solutions incorporate direct liquid cooling for enhanced thermal efficiency, PCIe for scale-up capabilities, and Ethernet for scale-out deployments. They also feature confidential computing to ensure the security of AI workloads and are designed for a rack-level power consumption of 160 kW.
The accompanying hyperscaler-grade AI software stack is optimized end-to-end for AI inference, supporting leading machine learning (ML) frameworks, inference engines, and generative AI frameworks. It also includes LLM/LMM inference optimization techniques such as disaggregated serving. Developers will benefit from seamless model onboarding and the convenience of one-click deployment for Hugging Face models, facilitated by Qualcomm Technologies’ Efficient Transformers Library and Qualcomm AI Inference Suite. This comprehensive software ecosystem provides ready-to-use AI applications, agents, tools, libraries, APIs, and services essential for operationalizing AI.
Also Read:
- Qualcomm Unveils Next-Gen Snapdragon Processors, Accelerating On-Device Generative AI Across Mobile and PC
- Reliance to Inject $12-15 Billion into AI Infrastructure, Spearheading India’s AI Revolution
Qualcomm AI200 is anticipated to be commercially available in 2026, followed by the AI250 in 2027. Qualcomm Technologies has committed to an annual cadence for its data center roadmap, focusing on continuous innovation in AI inference performance, energy efficiency, and TCO.
