Why the AI infrastructure bottleneck matters

Telecom companies, with their extensive real estate and existing network infrastructure, are re-evaluating their role in the AI era.

Enterprise leaders face a growing dilemma: the computational and energy demands of generative AI models are outpacing traditional data centre capacity. As organisations deploy AI beyond pilots, they encounter ballooning costs, latency issues with centralised cloud models, and increasing pressure on sustainability targets.

This infrastructure crunch is creating an opening for new players and models. Telcos see an opportunity to evolve from being simple connectivity providers to becoming foundational components of the AI value chain.

SK Telecom (SKT) is mounting a concerted effort to seize this opportunity. The South Korean telco announced it’s “evolving into a comprehensive artificial intelligence data centre (AIDC) developer” and intends to enter the global market.

Speaking at the SK AI Summit 2025, the company’s newly appointed CEO, Jung Jaihun, detailed a vision that positions SKT not just as a consumer of AI, but as a builder of its core infrastructure.

SK Telecom sets out plan for global enterprise AI infrastructure

SK Telecom’s plan involves two main thrusts: building a domestic AI infrastructure backbone and exporting its AIDC model internationally.

Domestically, the company is establishing AIDC hubs across Korea’s three major regions: the Seoul metropolitan area, the southern region (Ulsan), and the southwest region. The stated objective is to attract global capital and establish South Korea as “Asia’s biggest AI hub”. The Ulsan AIDC is a cornerstone of this plan, with intentions to expand it to a 1 GW-scale capacity through global partnerships.

For enterprise leaders outside Korea, SKT’s international expansion is the more pertinent development. The company plans to enter the Southeast Asian market using “energy-specialised AIDC solutions” developed with SK Group affiliates.

A concrete example is a planned AIDC in Vietnam. In collaboration with SK Innovation, SKT intends to build a facility that gets a stable power supply from a liquefied natural gas (LNG) power plant. In an example of industrial symbiosis, the system will also utilise “cold energy” (a byproduct of regasifying LNG) for the data centre’s cooling systems. SK Telecom’s plan directly addresses the two largest operational expenditure items for any AI-heavy infrastructure facility: power and cooling.

SKT is also serving as its own first-and-best customer. The company plans to acquire more than 2,000 NVIDIA RTX PRO 6000 Blackwell GPUs to build a ‘Manufacturing AI Cloud’. Jung confirmed this cloud will be used for “AI progression” in SK Group’s own manufacturing arms, like SK Hynix, providing the backbone for innovations such as digital twins and robotics AI.

Implementation, challenges, and ecosystem

SKT’s approach is about building data centres differently by leveraging its identity as a telco. CEO Jung Jaihun argued that the value of telco networks is being reassessed in the AI era.

“Utilising nationwide telecom infrastructure, telecom companies can uniquely bridge the gap between AIDC and on-device AI through edge AI and intelligent telecom technologies like AI-RAN,” Jung explained.

This model proposes a hybrid infrastructure where huge centralised AIDCs handle large-model training; while the telco’s existing edge network manages low-latency inference for on-device applications. Such a plan relies heavily on an ecosystem of partners, placing SKT in alignment with major cloud and hardware vendors.

On the cloud front, SKT has established an R&D partnership with Amazon Web Service (AWS) to accelerate its Edge AI implementation, building on a previous memorandum of understanding (MoU) for the Ulsan AIDC. For hardware, the company is collaborating with NVIDIA, government, and academic partners to demonstrate AI-RAN (Radio Access Network) technology, embedding AI into the network fabric itself.

To package this for customers, SKT is going beyond colocation to become what Jung called a “comprehensive AIDC developer overseeing the entire AIDC project—including design, construction, and operation.”

SKT plans to commercialise an “AIDC solution package” that includes internalised technologies for cooling and power management (out-rack and energy solutions), server interconnection (clustering solution), and server efficiency (in-rack solution).

The future of AI infrastructure sourcing

For CIOs, CTOs, and COOs, SK Telecom’s announcement provides several insights into the future of AI infrastructure sourcing.

It first signals the rise of specialised providers. While hyperscalers like AWS, Google, and Microsoft will remain dominant, telcos – with their unique assets in edge locations and power management – represent a new category of partner, especially for energy-intensive or low-latency workloads.

Second, the move frames infrastructure as a competitive driver. CEO Jung stated, “AI infrastructure is a core driver of both corporate and national competitiveness”. This reflects a wider sentiment that an organisation’s ability to access efficient, scalable, and sustainable compute will directly impact its ability to compete, elevating infrastructure decisions from a back-office IT concern to a boardroom-level discussion.

Finally, SKT’s ambition to create an AI hub in Korea highlights the growing importance of data sovereignty and “hub” strategies. As enterprises navigate complex regulations, the option to use regional, sovereign-capable AIDCs for training and inference becomes more attractive. Leaders must map their AI and data governance policies to these emerging infrastructure options.

SK Telecom’s infrastructure plan is a strong indicator that the industry is looking for new answers to the difficult economic and engineering questions posed by enterprise-scale AI.

See also: How AI-native 6G networks will boost enterprise operations

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security Expo, click here for more information.

Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.

Tags: ai, artificial intelligence, cloud, Enterprise, infrastructure, networks, Operators, sovereignty, strategy, telecoms