SEOUL, Dec. 18 (Korea Bizwire) — As governments around the world race to build artificial intelligence systems under national control, a new analysis suggests that telecom operators may emerge as the biggest beneficiaries of the shift toward so-called “sovereign AI.”
In a report published Nov. 26, credit rating agency Morningstar DBRS said telecom companies are structurally well positioned to play a central role as countries seek to secure their own data, computing power and networks — the three pillars of sovereign AI strategies.
Telecom operators that move early to build or support sovereign AI infrastructure are likely to gain significant ground in enterprise and government markets over the next five years, the report said.
Their advantages include deep experience operating large-scale networks, superior access to fiber infrastructure and data centers, and longstanding expertise in power management and facility operations.
The logos of South Korea’s major telecom operators — KT Corp., SK Telecom Co. and LG Uplus Corp. (Image courtesy of Yonhap)
While sovereign AI is an advanced, capital-intensive field, DBRS emphasized that its success will hinge as much on governance and regulatory compliance as on technology.
Telecom operators, which already manage sensitive national-scale infrastructure within strict legal frameworks, are well placed to meet the demanding oversight requirements of sovereign AI systems.
Citing McKinsey data, the agency projected that GPU-as-a-Service platforms built by telecom companies could generate several billions of dollars in market value by 2030.
Telecom carriers worldwide are already accelerating investments. Canada’s Telus opened the country’s first sovereign AI factory in Quebec in September.
In Europe, Italy’s Fastweb is building a supercomputer to run Italian-language large language models, while Telenor in Norway and Iliad in France have launched similar initiatives.
In South Korea, SK Telecom is participating in the government’s national AI foundation model project alongside semiconductor, data and service companies.
The consortium has begun developing a 500-billion-parameter foundation model, with plans to scale toward the multi-trillion-parameter range in the coming years.
Kevin Lee (kevinlee@koreabizwire.com)
