[Edaily Reporter Kim Hyun-ah ] As the competition in artificial intelligence (AI) expands beyond model development to a race to secure infrastructure—such as data centers, power, and cooling—calls are growing for tax incentives for AI data centers (AIDCs) to be adjusted to reflect current realities.

The government has designated semiconductors, physical AI, and AI data centers as its “three major megaprojects” and is positioning AIDCs as core national infrastructure. The government has set a goal of expanding AIDCs to a capacity of 18 GW.

On June 29, President Lee Jae-myung personally presided over a public briefing, attended by SamsungElectronics Chairman Jay Y. Lee and SK Group Chairman Chey Tae-won, among others, who announced large-scale investment plans. This marks the emergence of AIDCs as a national-level AI infrastructure development initiative, transcending the scope of individual private-sector investments. Naver (NAVER(035420)) has also entered the AI Factory business in partnership with NVIDIA and Brookfield.

The problem is that while the government is pushing for massive AIDC investments, the current tax system fails to adequately reflect the actual business structure of AIDCs.

Last year, the government included AI in its list of national strategic technologies and allowed large and mid-sized companies to claim a 15% investment tax credit for GPUs, power, cooling, and HVAC facilities. However, critics point out that the system’s effectiveness is diminished because colocation (leasing)—the core business model of AIDCs—may be excluded from the credit eligibility.

In particular, with the construction of a 1-GW-class AIDC estimated to cost approximately 70 trillion won, and given that the government has set a target of building 18 GW of large-scale infrastructure, there are growing calls for the tax system to keep pace with the actual investment structure.

At the “Policy Forum on Fostering AI Data Centers as National Infrastructure,” hosted on the 26th by lawmakers Park Gyun-taek, Ahn Do-geol, Yang Bu-nam, Lim Moon-young, Jeon Jin-sook, Jeong Jun-ho, Jeong Jin-wook, and Jo In-cheol and organized by the Korea Internet Enterprise Association, the need to amend the Act on the Limitation of Special Taxation Measures to stimulate AIDC investment was a central topic of discussion.

Professor Jo Young-im of Gachon University, who delivered the keynote presentation, identified the following as major obstacles to domestic AIDC investment: ▲ the tax credit rate for AI facility investments is lower than that for semiconductors; ▲ power and cooling equipment are not clearly included in the scope of tax credits; ▲ the low tax credit rate for cloud commercialization facilities, which are core components of AI; and ▲ a lack of support for companies relocating from the Seoul metropolitan area to regional areas.

Professor Cho emphasized, “Global nations are making every effort to attract infrastructure investments—for example, the U.S. allows 100% immediate expense deduction in the first year, and India offers a 20-year tax exemption,” adding, “It is urgent to revise the Act to recognize cooling and power facilities as integrated infrastructure and to introduce a temporary ‘bridge tax credit’ for companies relocating from the Seoul metropolitan area to regional areas.”

“70 Trillion Won in AIDC—Who’s Going to Use It All Themselves?”

Industry representatives have pointed out that the current tax credit structure, which is based on the premise of “self-use,” does not align with the realities of the AIDC business.

Lee Young-tak SKTelecom(017670), Vice President in charge of External Cooperation at the AIDC Integrated Promotion Task Force, stated, “General data centers and AIDCs must be distinguished,” adding, “While existing data centers are simply facilities for storing information, AIDCs are facilities that produce intelligence through large-scale computation, learning, and inference, and provide it to others.”

He continued, “It costs over 70 trillion won to build a 1GW-class AIDC—how could any single company consume all of that on its own?” He emphasized, “The essence of an AIDC is a ‘factory that mass-produces intelligence and supplies it to the outside world.’”

Currently, there are 161 data centers in South Korea, with a total capacity of approximately 2.5 GW. Based on demand submitted to KoreaElectricPower (KEPCO), it is projected that an additional 150 data centers, totaling 9.3 GW, will be constructed by 2030. This means that new demand—amounting to roughly four times the current total capacity—will be concentrated within a period of less than five years.

The problem is that the current tax credit system is designed around “direct use” by companies. The colocation model—in which a specialized operator builds a large-scale AIDC and provides space and facilities to multiple companies—may not qualify for the tax credit.

Vice President Lee pointed out, “An AIDC is not something you build solely for your own needs,” adding, “The fundamental nature of an AIDC is to provide the intelligence it generates to external parties.”

He added, “Because these are large-scale facilities and have the fundamental attribute of colocation—that is, leasing—which involves delivering the intelligence they produce to external parties, it is difficult to receive the 15% tax credit for national strategic technology commercialization facilities under the current system,” expressing concern that “it will likely remain difficult for companies in Korea to qualify for the 15% credit in the future.”

Ministry of Finance and Economy: “Principle of Restricting Leasing”… Ministry of Science and ICT: “Both Owned and Leased Facilities Qualify as AIDC”

The government is taking a cautious stance on the issue of recognizing colocation as eligible for tax credits.

Cho Moon-kyun, Director of the Tax Incentives Division at the Ministry of Finance and Economy, explained, “The inclusion of AI in national strategic technologies was based on its industrial significance,” but noted, “There is a fundamental principle of limiting tax credits to research and development and investments for a company’s own business, and restricting them to leased facilities.”

However, he noted that it is necessary to consider the fact that AIDC serves as the infrastructure underpinning the domestic AI industry.

Director Cho said, “We must examine the rationale for supporting AIDCs,” adding, “Since data centers play a role in supporting the domestic AI industry itself, we need to carefully review the justification for supporting investments beyond their own internal demand.”

The Ministry of Science and ICT maintains that the concept of AIDC itself encompasses both in-house use and leased facilities.

Jang Ki-cheol, Director of the Artificial Intelligence and Data Promotion Division at the Ministry of Science and ICT, explained, “We view data centers used for training and inference as encompassing both self-use and leased models,” adding, “We took this into consideration during the process of drafting the Special Act on AIDC.”

However, he noted that the decision on whether to expand tax credits must be based on policy objectives. While there is a need to support AIDC investments that contribute to domestic companies and the industrial ecosystem, the question of whether to provide the same tax benefits for facilities intended for foreign companies requires separate consideration.

Kakao: “AI Tax Credit at 20%”… Tax Gap in Relocation to Regional Areas Also an Issue

Calls have also emerged from companies that build and operate their own AIDC facilities to raise the tax credit rate and address the tax gap that arises during the relocation process to regional areas.

Kim Se-woong, Vice President ( Kakao(035720) ), stated, “Kakao is not so much an AIDC operator as it is building facilities internally for its own business,” arguing that the investment tax credit rate for the AI sector—currently at 15%—needs to be raised to 20%, the same level as for semiconductors.

He also pointed out tax loopholes arising from the process of relocating to regional areas in accordance with regulations in the Seoul metropolitan area and policies promoting the decentralization of power generation. Specifically, he noted that there is a lack of support mechanisms to back transitional investments when companies are considering or proceeding with the relocation of data centers from the Seoul metropolitan area to regional areas.

“The Golden Time Is Now”… Tax Revenue to Increase by Over 20% in 5–7 Years

Specific recommendations were also made regarding the timing of tax incentives and their long-term fiscal effects.

Professor Kim Jin-ki of Korea Aerospace University emphasized, “If we miss this golden opportunity, it will be difficult to catch up, and catching up will require even more resources,” adding, “Government-level tax policies are a powerful tool, but they must be used now. Once we fall behind, even tax policies won’t work.”

He continued, “This isn’t about a long-term plan spanning 10 or 20 years—the timing is what matters.” He explained, “While providing tax incentives may appear to reduce tax revenue in the short term, if we boldly support infrastructure investment now to revitalize the ecosystem, a virtuous cycle will emerge in 5 to 7 years: industrial growth and job creation will lead to an increase in tax revenue of approximately 20% or more.”

AIDCs are core infrastructure that goes beyond simple server hosting facilities to perform AI model training and inference. As AI competition expands beyond model performance to include the race to secure infrastructure—such as GPUs, power, cooling, and networks—there are concerns that if tax policies fail to keep pace with actual investment trends, domestic AIDC investment itself could be delayed.

Given that the government has set the massive goal of building 18 GW of AIDC capacity, the key challenge now is how to bridge the gap between national-level investment plans and the actual investment environment for companies. Industry and academic circles unanimously view the period around 2030 as the “golden age” for the domestic AIDC industry and agree that now is the time to revise the Special Taxation Act and its related enforcement regulations to better reflect current realities.