Maryanne Baines is a distinguished authority in cloud technology and infrastructure, possessing years of experience evaluating global tech stacks and the industrial application of high-scale digital solutions. Her expertise lies at the intersection of hardware capabilities and cloud deployment, where she frequently advises on how massive infrastructure projects translate into economic power. Today, she provides her perspective on the monumental shifts occurring in East Asia as national strategies pivot toward total AI dominance.
The discussion centers on the massive “mega projects” designed to catapult South Korea into the top tier of global AI powers alongside the United States and China. Key themes include the staggering financial commitments from industry giants, the technical hurdles of building gigawatt-scale data centers, and the strategic integration of semiconductor manufacturing with localized energy solutions. We also explore the global supply-demand gap in digital infrastructure and how the development of a specialized workforce in robotics and “physical AI” completes the vision of a fully integrated technological ecosystem.
South Korea’s ambition to join the ranks of the US and China as a global AI power is a massive undertaking. How do the “mega projects” recently announced lay the groundwork for this kind of national transformation?
This is a bold, high-stakes move that represents a complete alignment between the public sector and the nation’s largest industrial titans like Samsung and SK Group. By formalizing the AI G3 strategy, the government is essentially saying they aren’t content with just being a participant in the digital economy; they want to be one of the three pillars holding it up. These “mega projects” provide the physical and financial skeletal structure for that dream, starting with an initial investment phase of approximately 550 trillion won. It’s a comprehensive approach that doesn’t just look at software but focuses on the heavy-duty semiconductors, physical AI, and the massive data centers required to run them. The sheer gravity of the investment, which could exceed 1,000 trillion won by 2035, shows a level of national commitment that is rare to see, aiming to create a self-sustaining loop of innovation and production.
SK Telecom has set a staggering 15GW target for its AI data center capacity. What are the logistical and technical hurdles of building out infrastructure at this unprecedented scale?
Building to a 15GW capacity is a Herculean task that changes the very definition of what we consider a data center project. To put that in perspective, the initial phase alone aims for 8.4GW, with construction expected to kick off by the first half of 2028. Logistically, you are looking at a massive puzzle involving site selection, securing stable power from nuclear or liquefied natural gas sources, and managing a global supply chain for cooling and compute. SK Telecom is already breaking ground on a 1GW-scale cluster in Ulsan and planning another 2GW in the Gyeongsang region, which requires a literal reshaping of regional infrastructure. You can feel the heat and intensity of this buildout when you realize that even a single 1GW facility can cost around KRW 70 trillion because of the high-performance computing hardware and rising memory costs.
With the financial stakes reaching into the hundreds of trillions of won, how does the industry plan to navigate the immense costs associated with high-performance computing and memory?
The financial architecture behind these projects is just as complex as the engineering, relying on a mix of internal corporate investment, strategic global partnerships, and long-term customer contracts. Samsung alone is looking at a domestic investment of 2,450 trillion won between 2026 and 2040, with a massive 2,100 trillion won of that earmarked for semiconductor clusters in Pyeongtaek and Yongin. There is an emotional weight to these numbers; they represent a “bet-the-company” strategy on the future of intelligence. Companies are also looking toward project financing and overseas capital to bridge the gap, especially as high-bandwidth memory prices continue to climb. It’s a race where the entry fee is astronomical, but the cost of not participating is even higher, leading SK Group to plan for an average annual investment of over 100 trillion won over the next decade.
The plan focuses heavily on regional development, specifically in areas like Ulsan and the Gyeongsang region. Why is the geographic distribution of these data centers so critical to the success of the national AI strategy?
Distributing these facilities isn’t just about finding space; it’s about creating an “AI hub” that breathes life into regional economies and ensures the grid can handle the load. By spreading 2GW across Gyeongsang and 1GW in the Jeolla region, the project avoids putting too much strain on a single metropolitan area while supporting the government’s regional development goals. This approach allows the “AI Factory” concept to integrate directly with local industries, potentially revitalizing manufacturing through real-time data processing. It’s a sensory experience of seeing traditional industrial zones transformed into high-tech corridors where data flows as freely as physical goods once did. This decentralized model also helps in managing the massive power requirements, which the International Energy Agency predicts will double global data center consumption to 945 TWh by 2030.
Beyond the physical data centers, there is a significant focus on semiconductors and high-bandwidth memory. How does domestic production of these chips give a unique advantage in the global race?
South Korea holds a distinct “home-field advantage” because it is the world’s kitchen for the very ingredients AI needs to survive. Samsung and SK Hynix are pouring hundreds of trillions of won into new fabs—including 400 trillion won for a new chip production base and 56 trillion won specifically for advanced high-bandwidth memory in Cheonan and Onyang. When your data center is sitting just miles away from the fab producing the HBM chips it uses, the synergy is unparalleled in terms of supply chain reliability and technical integration. This close proximity allows for faster iteration on the “next-generation” AI data centers, or AI Factories, which are scheduled to begin operations as early as 2027. It creates a closed-loop ecosystem where the hardware and the infrastructure are developed in lockstep, something few other nations can replicate.
The project also mentions “physical AI” and the training of 10,000 robotics specialists. What role does this human and robotic element play in the broader vision for an AI-driven economy?
You can’t run a 15GW empire with just servers; you need the intellectual and mechanical “hands” to maintain and evolve it. The government’s goal to train 10,000 specialists over five years is a recognition that the bottleneck for AI isn’t just silicon, but human talent and the integration of AI into the physical world. This “physical AI” foundation model approach means the data centers will eventually power industry-specific robots that can operate in data factories or manufacturing plants. It’s a vision of a future where the line between the digital and the physical blurs, with robots managing the very high-density cabinets that 29% of operators are already struggling to cool. By investing in the people and the robotics today, they are ensuring that the massive 1,100 trillion won semiconductor projects have a skilled workforce ready to operate them.
Looking at the global landscape, McKinsey forecasts a massive data center supply shortage by 2030. How does the aggressive South Korean timeline position it to fill these gaps in the international market?
With McKinsey forecasting global demand reaching up to 219GW by 2030, and the US potentially facing a 15GW shortfall, South Korea is positioning itself as the global “overflow” and core hub for Asia. Their aggressive timeline, including moving up fab completion dates by 12 years in some cases, is a direct response to a market where supply simply cannot keep pace with 19% to 22% annual growth. By building out 15GW of capacity, they aren’t just serving domestic needs; they are building a surplus that global technology companies can tap into. It is a strategic play to become the “core AI infrastructure hub” for the entire continent, leveraging their expertise in nuclear power and large-scale facility operations to provide a stability that other regions might lack.
What is your forecast for the South Korean technological infrastructure over the next decade?
I expect that by 2035, we will see South Korea successfully bridge the gap between being a leading hardware exporter and becoming the world’s most sophisticated AI operations hub. The integration of 15GW of capacity with a domestic supply of high-bandwidth memory will likely create a new standard for “AI Factories” that other nations will scramble to emulate. We will see a landscape where regional clusters in Ulsan and Yongin are no longer just industrial sites, but the beating heart of a global digital nervous system, powered by a workforce that is as comfortable with robotics as they are with code. The massive financial commitments today, totaling over 1,000 trillion won, will solidify their position as an indispensable titan in the AI G3, making them a primary destination for global technology partnerships and the definitive hub for Asian data infrastructure.
