Data center construction costs $49 million per MW as AI infrastructure investment extends beyond chips to power and construction
Global data center spending is expected to exceed $1 trillion, approximately 1,320 trillion won, in 2026. It is expected to more than double to $2.1 trillion, approximately 2,780 trillion won, in 2028. The increase reflects sharply rising costs for building the computing infrastructure needed to train and run artificial intelligence (AI) models.
On the 23rd, local time, Reuters cited a Bank of America (BofA) Global Research report in reporting on the large amounts of funding being invested in servers, network equipment, power infrastructure and cooling systems. The analysis suggests that competition to secure AI chips is expanding into construction and infrastructure investment.
According to BofA, data center construction costs about $49 million, approximately 65 billion won, per megawatt (MW) of capacity. Information technology (IT) equipment accounts for 78% of that cost. Servers represent the largest share at about $31.4 million, approximately 41.6 billion won, per MW. Network equipment costs $4.8 million, approximately 6.4 billion won, per MW. AI-dedicated facilities known as neoclouds are more expensive at about $57 million, approximately 75.5 billion won, per MW because they require specialized hardware to support intensive AI workloads.
Spending extends beyond semiconductors and servers. Costs for individual infrastructure components, including electrical switchgear, generators, emergency power and cooling equipment, were estimated at $100,000 to $1.1 million, approximately 100 million to 1.5 billion won, per MW. New business opportunities are emerging for industrial manufacturers and engineering and construction companies that previously had little connection to the technology sector.
Citing the BofA report, Reuters reported that the global data center market would grow from $275 billion, approximately 364 trillion won, in 2022 to $2.93 trillion, approximately 3,880 trillion won, in 2030. BofA expected AI demand to drive average annual growth of 37% from 2025 through 2030.
There are also risks. Developers face difficulties securing power, building permits and grid connections. Rising costs and opposition from local residents could delay projects. Long-term profitability depends on whether revenue from AI products and services can justify the investment. BofA said actual spending could vary depending on the pace of AI adoption, financing conditions and the ability to bring new facilities into operation.
Original Korean article: Digital Daily
English brief by Korea Wind Intelligence, koreawindintel.com
