Qualcomm Partners with Cirrascale to Launch Customer Evaluation of Its AI200 Inference Accelerator
IT之家 November 20 news: Qualcomm announced on the 17th of this month local time that, together with AI private cloud company Cirrascale, it has opened evaluation of the Dragonfly AI200 inference accelerator to select customers. Customers can explore how their AI workloads perform on the chip through a cloud environment.

Dragonfly AI200 is Qualcomm's first rack-scale AI solution, with 768GB LPDDR5X memory per card, supporting PCIe Gen6 scale-up and RoCE scale-out.
Each 140kW liquid-cooled ORv3 rack consists of 56 cards, with total memory reaching 43TB and 0.414 PB/s of memory bandwidth, capable of running inference for 10T-class ultra-large models in a single rack while delivering a lower total cost of ownership.
Related Articles
- AMD launches Instinct MI450X GPU accelerator: same core specs as MI455X, memory reduced to 288GB HBM4
- Rockchip RK1860 specs leak: 64TOPS dense INT8 compute, up to 10GB DRAM integrated
- AMD completes acquisition of Taalas, plans to integrate the latter's chips into its accelerator roadmap
- Axelera AI unveils Titania architecture for data-center-class AIPU chip, production in 2027Q3
- South Korea unveils edge AI semiconductor roadmap, plans to complete development and verification of 10 chips by 2030
- UK AI chip startup Fractile receives government funding, valuation reaches $7 billion