The agile telco can benefit from the rush to the edge by organisations wanting to get the most out of their AI data. Antony Savvas looks at recent technology moves in the market.
Qualcomm has completed its acquisition of Modular, a software platform that gives developers a unified way to optimise and deploy generative and agentic AI workloads across computing systems.
The acquisition will accelerate the expansion of Qualcomm’s AI platforms across devices, data centres, edge infrastructure, and personal and industrial AI. Modular’s commitment to an open, heterogeneous ecosystem will continue, says Qualcomm, while delivering performance across CPUs, GPUs, neural processing units (NPUs) and custom silicon.
“Combining Modular’s AI-native software platform with our solutions, industry scale and ecosystem partnerships accelerates our ability to deliver high-performance, energy-efficient AI solutions from edge to cloud,” says Cristiano Amon, the president and CEO of Qualcomm Technologies. “With Modular’s world-class engineering team, we’re enabling a new and open approach to AI software development, enabling AI to run efficiently across any hardware while maximising performance.”
In another edge AI telecoms development, Spain’s edge AI specialist Multiverse Computing has sealed a €500m Series C funding round at a €1.5 billion pre-money valuation. The company addresses two specific industry needs: AI being run on devices, not just through data centres, and the demand for sovereign AI.
Device manufacturers and infrastructure operators believe the next generation of AI has the potential to run directly on devices at the edge, and Multiverse Computing’s CompactifAI makes that possible. And, with efficient and sovereign AI at scale, the same tech promises to do “twice the work on half the energy”.
The money raised will be used for product development and to extend the firm’s go-to-market, including cloud and telecoms service providers, and global system integrators.
Customers and partners already span manufacturing, telecoms, finance, energy, aerospace, cybersecurity, defence, and health and life sciences, including Allianz, Bank of Canada, Bosch, PwC and Telefónica.
“We have proven that AI can run at full performance on a smartphone, inside a sovereign data centre, or on a factory floor with no cloud connection. This round is the moment we scale that proof across every industry that needs it,” says Enrique Lizaso, co-founder and CEO of Multiverse Computing.
CompactifAI is a compression technology that applies tensor networks, a mathematical framework from quantum physics, to the problem of AI model compression. It reduces the size of large language models “by up to 80-95%” with “immaterial accuracy loss”. The result is AI that runs faster, consumes less energy, and can be deployed in environments where sending data to a hyperscaler is too expensive, too slow, or not permitted.
Multiverse’s platform spans the full deployment spectrum for efficient AI: compressed models optimised for cloud, on-premises, and on-device environments, orchestrated by the CompactifAI Router that decides in real time whether a workload should run locally or be sent to the cloud. The company is also applying an end-to-end software layer for next-generation AI factories and data centres, that brings model compression, GPU orchestration, AI services, and sovereign-grade controls into a unified platform.
For enterprises, this means that rather than having to route every workload through a hyperscaler and replacing existing infrastructure, Multiverse adds an efficiency layer above it that enables them to match each workload with the right model and backend – “potentially saving hundreds of millions in the process”, says the firm.
Multiverse models are already being deployed across millions of devices and systems, including drones, cameras, satellites, vehicles, and telecom infrastructure, and are expected to be embedded into AI PCs.
Fancy that
And, fancy that… Multiverse Computing has recently announced a collaboration with Qualcomm, to bring “highly efficient” AI models to data centres worldwide. Multiverse’s tech will be specialised for Qualcomm Dragonfly AI200 and AI250 accelerators, to enable data centre operators to run large-scale AI workloads “with higher performance and lower power consumption”, said the new partners.
For data centre operators, the direct benefit is more capacity on the same hardware. By optimising AI models before they run on Qualcomm Dragonfly AI200 and AI250 accelerators, Multiverse reduces the compute and memory footprint each model requires. This 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 centre infrastructure.
“Qualcomm 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. By combining Qualcomm’s highly efficient and performant AI accelerators with our model compression and optimisation technology, we can help customers achieve breakthrough improvements in performance, cost, and energy efficiency at scale,” says Victor Gaspar, chief sales officer at Multiverse Computing.
The two companies say they are providing edge data centre operators and enterprises a practical path to “scalable AI”, and, no doubt, similar alliances will be springing up elsewhere across the communications industry.
Antony Savvas