Growing role of telecom in 'sovereign AI'

In preparation for DTW Ignite 2026, TM Forum published research showing that telecoms are getting serious about sovereign AI, with 77% identifying it as a commercial opportunity. In TMF's discussion paper, Making Sovereign AI Real, TM President Nik Willetts and Chairman Steffen Roehn explain, "Sovereign AI will not be made real by owning GPUs or building data centers. It will be made real by controlling the enforcement layer that governs how AI behaves in motion – across every interaction, in real time. Regulators around the world are moving fast to make this a legal requirement."

To that last point, the EU AI Act exerts pressure on telecom and tech companies to manage AI according to regional laws, and the EU's recently proposed Cloud and AI Development Act (CADA) is the body's first attempt to make sovereign AI an auditable legal requirement. RCRWireless' James Blackman recently questioned the EU's role approach, stating that "Europe is trying to govern the new AI era with sledgehammer policy tools designed for a different age."

In the United States, the approach is more fragmented, with a patchwork of state-level laws, federal executive orders, and sector-specific agency enforcement. The most recent would be EO 14409, "Promoting Advanced Artificial Intelligence Innovation and Security," which creates a voluntary pre-release review framework iin which AI developers share advanced "covered frontier models" with federal agencies for up to 30 days before public release, allowing time for national security and cyber risk assessment.

According to RCRWireless guest columnist Vish Nandlall, sovereignty strategies fall into 5 distinct models, which he calls "dealmaker, procurement, subsidy, governance, dependency." He sees a major bottleneck, with countries subsidizing around assets they do not control, and purchasing access to "someone else’s silicon, someone else’s models, and someone else’s operating layer."

Regardless of the model or the blurring of the lines among all of them, sovereign AI will elevate the role of telcos – increasingly strategic partners to AI data center owners because of their influence over domestic AI infrastructure. Not only are they investing in connectivity, as with high-capacity fiber optics, but the are also facilitating secure data residency, and localized edge networks that allow countries to process and control AI workloads within geographic borders. Leading telcos, like AT&T, Deutsche Telekom, SK Telecom, and TELUS have not only the netowrks, but also thousands of pre-existing, grid-connected sites, central offices, and base stations – all crucial to data center owners that want to deploy and build at pre-powered telecom sites.

The opportunity for telcos is substantial, and according to TM Forum, 72% of CSPs are already investing in or exploring sovereign AI, with 98% adapting or exploring changes to network architecture, and 96% aligning more closely with national or regional policy.

To get a copy of the TM Forum's position paper, contact newsroom@tmforum.org

Predictive memory technology: MEXT claims it delivers 2 - 4x increase in usable memory capacity. Its acquisition by AMD means DC operators can defer DRAM investments while still offering customers large-memory instances. 

China's $2T yuan AIDC plan: A grid of Chinese AI data centers will be built over 5 years as part of a $295 billion (2 trillion yuan) plan. At least 80% of the hardware and software, incl. AI chips, must be sourced Huawei and other domestic suppliers.