Insider Brief
- Multiverse Computing and Qualcomm Technologies are collaborating to optimize AI models for Qualcomm Dragonfly AI200 and AI250 accelerators, aiming to increase data-center performance while reducing memory use and power consumption.
- Earlier demonstrations using Qualcomm’s Cloud AI100 Ultra accelerator showed compressed models delivering up to 93% faster responses, 54% higher throughput, 45% lower memory use and 21% lower power consumption without reduced accuracy.
- The companies said the approach could let data-center operators handle more inference requests and run more models on existing hardware without adding accelerators or expanding infrastructure.
PRESS RELEASE — Multiverse Computing today announced a collaboration with Qualcomm Technologies, Inc. to bring highly efficient AI models to data centers worldwide. Multiverse Computing’s AI models will be specialized for Qualcomm Dragonfly™ AI200 and AI250 accelerators. The collaboration focuses on Qualcomm Technologies’ AI acceleration hardware and Multiverse Computing’s model optimization technology, to enable data center operators with a way to run large-scale AI workloads with higher performance and lower power consumption.
For data center operators, the direct benefit is more capacity on the same hardware. By optimizing AI models before they run on Qualcomm Dragonfly AI200 and AI250 accelerators, Multiverse Computing reduces the compute and memory footprint each model requires. That frees up headroom on existing deployments, letting operators serve more inference requests, run more models concurrently, or scale their AI services without adding new accelerators or expanding data center infrastructure.
The efficiency gains were demonstrated live at Mobile World Congress in March 2026, where Multiverse Computing and Qualcomm Technologies showcased a compressed open-source large language model running on Qualcomm® Cloud AI100 Ultra accelerator. In a real-time emergency medical reporting use case, the compressed model demo delivered up to 93% faster response times and up to 44% higher throughput than the uncompressed base model, while reducing memory usage by up to 45% and power consumption by up to 21%, with no loss in accuracy.
In a second demonstration, an on-premises Retrieval-Augmented Generation (RAG) chatbot for querying confidential financial documents ran up to 35% faster and delivered up to 54% higher throughput, while cutting memory usage by up to 45% and power consumption by up to 14%, again with no loss in accuracy.
Applications running on the new Qualcomm Dragonfly AI200 and AI250 accelerators will be poised to deliver even better benchmarks results.
“Qualcomm Technologies has built its leadership in segments such as mobile and IoT, where maximizing performance within strict power and efficiency constraints has always been essential. Those same principles are now becoming critical in AI data centers. By combining Qualcomm Technologies’ highly efficient and performant AI accelerators with Multiverse Computing’s model compression and optimization technology, we can help customers achieve breakthrough improvements in performance, cost, and energy efficiency at scale.”
Victor Gaspar, Chief Sales Officer at Multiverse Computing
“As AI adoption accelerates across enterprise and cloud infrastructure, customers need solutions that deliver high performance with greater efficiency. Qualcomm Technologies brings industry-leading AI acceleration, while Multiverse Computing works closely with customers, their AI models, and their specific use cases to optimize real-world deployments. Together, we can deliver highly efficient AI solutions tailored to the needs of modern data centers.”
Dino Flore, Vice President, Technology at Qualcomm Europe Inc.
The collaboration reflects a broader shift in the data center industry toward efficiency-first AI infrastructure, as operators look to scale AI deployments without a proportional increase in hardware, energy, and cost. By combining Qualcomm Technologies’ AI accelerators with Multiverse Computing’s optimized AI models, the two companies enable data center operators and enterprises with a practical path to scalable AI.
