AI & Tech

Qualcomm offers test drive of its rack-scale Nvidia AI challenger

Qualcomm has opened cloud-based trials of its Dragonfly AI200 rack-scale inference system to selected customers through Cirrascale Cloud Services.

✦ Extractive summary — built from the sources' own reporting•From 1 source•1 min read•Updated November 19, 2026, 2:03 PM UTC
Key points
  • Qualcomm says the system supports models ranging from 7 billion to 10 trillion parameters, including long-context, retrieval-augmented generation (RAG), and agentic workloads.
  • Its Blackwell-based NVL72 systems use fast – and costly – HBM3e stacks, while Qualcomm has opted for LPDDR, more commonly found in mobile devices.
  • Dragonfly AI200 is Qualcomm's pitch for running large language, multimodal, and other AI models without Nvidia-sized power and operating bills.

He claimed that the AI200 brings the company's experience with high-performance, low-power computing to a rack-scale infrastructure, in an architecture designed around the performance, memory, and cost requirements of inference workloads.

Whether it delivers those savings is something customers can now put to the test on Cirrascale's AI Innovation Cloud, with a route into production deployments, Qualcomm says.

Additional details
  • Each AI200 accelerator card carries 768 GB of LPDDR memory, with 56 cards providing about 43 TB in a liquid-cooled rack.
  • "AI inference is becoming one of the largest and most important computing workloads in the datacenter, and the industry needs a more efficient approach," said Tony Pialis, Qualcomm's executive vice president and general manager for datacenter.
  • "Our customers need efficient, scalable infrastructure that can support increasingly sophisticated models without allowing power and operating costs to become the limiting factor," said Cirrascale CEO Dave Driggers.
  • The AI200 is part of Qualcomm's Dragonfly datacenter lineup, which also includes the AI250 and AI300 accelerators and the C1000 server CPU.
  • As we previously explored, packing in memory doesn't necessarily mean matching Nvidia's performance.

How this summary was made: extractive summary — built from the sources' own reporting. It is drawn only from the reporting of the sources listed below — read the originals for full detail. How we summarize →

Advertisement

Every article behind this summary — 1 article from 1 outlet