Some communication service providers (CSPs) want to monetize their fiber networks, edge facilities, and data centers through GPU-as-a-service (GPUaaS) offerings, but AI and 6G infrastructure expert Vish Nandlall warns of the “brutal economics” of low latency and high utilization rates that hyperscalers more readily achieve. He also points out that CSPs generally lack the software stacks that bigger platform providers already offer. To Nandlall, the bigger AI opportunities lie in CSPs’ unique network and data center assets, which enable:

• AI sovereignty and data residency services for governments and enterprises whose AI infrastructure has to stay within national borders. 
• Smart landlord/infrastructure partner opportunities in which CSPs host hyperscaler infrastructure, instead of building a competing software stack. They can leverage their real estate, power, and connectivity, while partners like NVIDIA or Microsoft deploy hardware and software.
• Network integration services in 5G-enabled vertical markets, where CSPs can bundle low-latency connectivity with local compute for use cases like a smart factory and others requiring turnkey solutions with 5G network slice and on-site GPUs for real-time inference.

One way companies are trying to boost AI valuation multiples is with AI-oriented use cases. As that happens, there’s a gradual evolution from training to inference, which Digital Reality CEO Andy Power told investors on a Q3 conference will require “scalable, connected infrastructure across key metros worldwide.” While many hyperscale projects have been announced in remote areas, he says highly connected urban data centers will gain value over remote hyperscale campuses.

That’s where Power emphasized the value of Digital Realty’s urban portfolio: “Our data center capacity is situated in and around the world’s most highly connected cloud zonal markets with the highest concentration of population and GDP.”  As digital transformation, cloud, and AI continue to grow, “our ability to deliver scalable, connected infrastructure across key metros worldwide is more critical than ever.” 

One metro area trying to get more data center action is the UAE’s Abu Dhabi, which under the U.S.–UAE AI Acceleration Partnership supports collaborations between the two nations. In alignment with that partnership, Cisco and G42 will jointly develop a large-scale, sovereign AI infrastructure in the UAE, featuring AMD MI350X GPUs and Cisco’s full-stack infrastructure for high-performance AI workloads — both of which will be incorporated into G42’s “Regulated Technology Environment” framework. The framework ensures transparency, protection, and alignment with U.S. standards.The deal supports the UAE’s goal of becoming a leader in AI by 2031, despite its current dearth of GPU capacity. According to Cisco’s latest AI research, 92% of organizations in the UAE plan to deploy AI agents, and 41% expect them to work alongside employees within a year, and yet only 25% have GPU capacity. It’s a gap this and other partnerships aim to close.