Interest in data centers is intense. Development and investment are increasing rapidly as the use of artificial intelligence (AI) expands. However, the fact that an industry is growing does not make every data center a good investment.
Before asking how many will be built, we need to ask who will use them, how much they will cost to operate, and whether they can retain their value over time.
Data centers have, in fact, existed for many years in the form of corporate computer rooms and computing centers. The recent change is that the scale of facilities and their technical requirements are increasing rapidly as AI spreads. This is why it has become difficult to assess conventional data centers and very large AI data centers using the same criteria.
When discussing supply and demand for offices and logistics centers, we generally look at floor area. We compare the space being supplied with demand for its use. For data centers, electricity is the key criterion. Even with land and the ability to construct a building, it is difficult to proceed without a reliable supply of the necessary electricity. Even the criteria for assessing a location differ from those used for conventional real estate.
Here, we must distinguish between the amount of electricity requested by a project developer and demand from companies that will actually use the facility. A large number of applications to use electricity does not, by itself, establish that tenant demand will be sufficient after completion. Electricity availability, development feasibility and the securing of end users are matters that need to be verified.
The concentration of data centers in the capital region reflects the presence of companies and users, telecommunications infrastructure, and operational support. There is also demand for processing data close to users. These are reasons why demand in the capital region is unlikely to disappear easily, even as electricity supply becomes constrained and the policy emphasis on regional decentralization strengthens.
I expect a clearer pattern to emerge in which small and medium-sized data centers in the capital region grow alongside very large data centers outside it. Smaller developments and conversions of existing buildings are likely to continue in the capital region, while facilities using large amounts of electricity are likely to increase elsewhere.
The key risk for very large facilities is their tenants. If a particular company decides not to occupy an office, there is relatively greater scope to replace it with another company. By contrast, the number of companies that can use a very large data center is limited. If a major user withdraws, it is difficult to find comparable demand immediately. This is why identifying who will use the facility matters from the development and financing stages.
The role of operators is also changing as the market expands. Previously, the prevailing model involved telecommunications companies and others leasing an entire facility and then providing it to companies that needed it. A structure is now emerging in which owners secure end users directly and outsource operations to specialist companies. Alongside development capabilities, experience in reliable operation needs to be assessed.
Post-completion costs and resale distinguish asset values
From an investment perspective, what happens after development requires closer scrutiny. Current high valuations of data centers reflect expectations that demand and earnings will continue to grow. What matters is whether those expectations translate into actual operating performance. Current valuations must also be sustainable when a completed development is sold to the next investor.
Equipment replacement costs are a major variable in this process. Data centers must run power and cooling equipment continuously. Over time, maintenance and replacement can incur substantial costs. Even if rents rise, asset valuation changes if the cost of reinvesting in equipment grows faster. The rate of income growth must be considered alongside the cost of sustaining it.
It is therefore difficult to apply the same investment criteria to every data center. Small and medium-sized facilities in the capital region may offer scope for assessment from a conventional real estate perspective, supported by relatively diverse demand. By contrast, very large facilities outside the capital region may increasingly resemble infrastructure assets requiring substantial initial investment and long-term operation. The methods used to assess investment recovery periods and risk must also differ.
The experience of the logistics center market also deserves consideration. Investment flowed into facilities on expectations that they would support a growth industry, but some subsequently faced vacancies and difficulties converting to other uses. We cannot conclude that data centers will follow the same path. Nevertheless, the lesson that industry growth alone does not guarantee the profitability of an individual development project remains valid.
The outcome for the data center market remains uncertain. A firmer basis for valuation will emerge only when actual operating income and equipment replacement costs become clear and assets change hands. While acknowledging the potential for growth, investors now need to ask how much they are paying for those expectations. The more optimistic the outlook, the more specific and cautious investment decisions need to be.
Original Korean article: Edaily
English brief by Korea Wind Intelligence, koreawindintel.com
