Qualcomm Technologies and Multiverse Computing have entered into a collaboration focused on optimising AI models for deployment in data centres.
The partnership centres on adapting Multiverse Computing’s AI model optimisation technology to work with Qualcomm’s Dragonfly AI200 and AI250 accelerators.
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
This initiative is expected to help data centre operators run large-scale AI workloads with improved performance and reduced power usage.
The collaboration aims to specialise Multiverse Computing’s AI models for Qualcomm’s acceleration hardware, enabling operators to increase capacity while using existing infrastructure.
By compressing and refining AI models before they are deployed on Qualcomm Dragonfly AI200 and AI250 accelerators, Multiverse Computing states it can lower the compute and memory requirements of those models.
This approach may enable operators to handle more inference requests, run several AI models at the same time, or scale up their services without needing to invest in additional accelerators or expand physical infrastructure.
The companies demonstrated these efficiency gains in March 2026 during the Mobile World Congress event.
In one demonstration, they showcased a compressed open-source large language model running on the Qualcomm Cloud AI100 Ultra accelerator in a real-time emergency medical reporting scenario.
This model achieved response times up to 93% faster and throughput up to 44% higher compared with the uncompressed variant, while memory usage dropped by up to 45% and power consumption fell by up to 21%.
The tests were reported to show no loss of accuracy.
In a separate demonstration, an on-premises Retrieval-Augmented Generation chatbot used for querying sensitive financial documents delivered up to 35% faster response times and up to 54% higher throughput.
Memory usage was reduced by up to 45% and power consumption by up to 14%, also without any reported accuracy loss.
Multiverse Computing chief sales officer Victor Gaspar said: “Qualcomm Technologies has built its leadership in segments such as mobile and IoT, where maximising performance within strict power and efficiency constraints has always been essential. Those same principles are now becoming critical in AI data centres.”
In June 2026, Qualcomm Technologies announced a multi-generation agreement to supply data centre central processing units (CPUs) supporting the growing compute footprint of Meta.
