Qualcomm Opens Dragonfly AI200 Cloud Trials

Selected customers can now evaluate Qualcomm's rack-scale AI inference system through Cirrascale Cloud Services.

Updated on Nov. 19, 2026 in Semiconductors

Isometric editorial illustration of a tall server-rack enclosure with blue status lights, beside a cable bundle and cooling unit, representing cloud trials of AI inference hardware.
Qualcomm is letting selected customers test its Dragonfly AI200 rack-scale AI inference system through Cirrascale Cloud Services before committing to production purchases. AI Illustration. Upload story photo >

Live Poll

Do you expect the cost of AI-powered online services to rise next year?

Qualcomm opened cloud-based trials of its Dragonfly AI200 rack-scale inference system to selected customers through Cirrascale Cloud Services. Cirrascale had demonstrated the system at SC26 in Chicago on November 17-18.

Why it matters

The trials give customers a route to test large AI model inference on Qualcomm hardware before committing to production deployments. Qualcomm positions the AI200 as a way to run large models without Nvidia-sized power and operating costs.

Each AI200 accelerator card carries 768 GB of LPDDR memory, and 56 cards provide about 43 TB in one liquid-cooled rack, supporting models from 7 billion to 10 trillion parameters. Qualcomm uses LPDDR, which is more common in mobile devices, rather than the HBM3e stacks found in Nvidia's Blackwell-based NVL72 systems.

The players

Qualcomm

Qualcomm is a San Diego-based semiconductor company best known for its Snapdragon mobile chips and wireless technology licensing.

Cirrascale Cloud Services

Cirrascale Cloud Services is a cloud provider that offers GPU and AI accelerator infrastructure to enterprise and research customers.

Nvidia

Nvidia is the dominant maker of GPUs used for AI training and inference, and the main competitor in the rack-scale AI systems market.

The details

Selected customers can evaluate the Dragonfly AI200 through Cirrascale's AI Innovation Cloud. The Dragonfly AI Infrastructure Management Suite handles provisioning, monitoring, workload orchestration, and fault handling across rack-scale deployments.

Timeline

  1. June 2026: Qualcomm set out broader ambitions at its Investor Day.

  2. November 17-18, 2026: Cirrascale demonstrated the AI200 at SC26 in Chicago.

  3. November 19, 2026: Qualcomm opened cloud-based AI200 trials to selected customers.

Market Landscape

Qualcomm's AI200 is positioned as an alternative to Nvidia's Blackwell-based NVL72 systems, which use HBM3e memory stacks. The cloud trials mark an early step in Qualcomm's push into datacenter AI accelerators against Nvidia's dominant position.

Organizations evaluating AI inference infrastructure can now request access to the Dragonfly AI200 through Cirrascale's cloud. The trial path offers a way to test workloads before committing to on-premises rack deployments.

The takeaway

For organizations weighing AI inference hardware, cloud trials offer a low-commitment way to compare memory architectures before buying rack-scale systems. The results will show whether LPDDR-based designs can compete with HBM-based alternatives on real workloads.

Further reading

Explore more Semiconductors coverage on Wisevoter.

Source note: This article includes information reported by TheRegister.

Live Poll

Do you expect the cost of AI-powered online services to rise next year?