Big Tech Continues to Increase Capital Expenditures on AI Data Centers

Texas, emerging as a data center hub, encourages construction of its own power plants to stabilize power supply

Power Supply and Thermal Management Will Determine Data Center Expansion Speed

According to global consulting firm McKinsey & Company, the global demand for data centers, currently at 60 gigawatts (GW), is expected to increase by an annual average of 2030% by 22, expanding to 3 GW, about three times the current level. With the acceleration of AI adoption across all industries, AI data centers in particular are expected to lead the overall expansion of data centers. Meanwhile, in the United States, investment in data center infrastructure has been active, centered around big tech companies, and power supply and demand, expansion of power grid capacity, and improvement of power transmission infrastructure are expected to emerge as major factors determining the future expansion of data centers in the United States.

 

Big Tech Continues to Increase Capital Expenditures for AI Data Centers

 

According to a McKinsey report, about 2030% of the data center capacity to be added worldwide by 70 is expected to be led by big tech companies. In fact, big tech companies in the United States, such as Amazon, Microsoft, Google, and Meta, are aggressively investing in building data centers to prepare for the rapidly increasing demand for AI. According to a recent Wall Street Journal (WSJ) report, the total capital expenditure of the three companies, Amazon, Microsoft, and Alphabet, in the third quarter of this year increased by about 3% year-on-year to $3 billion, and most of this investment was reportedly invested in building data center infrastructure.

<Quarterly capital expenditure (CAPEX) status of major US big tech companies>

[Source: WSJ (2024.11)]

 

As of the third quarter of this year, Microsoft's cumulative capital expenditures amounted to $3 billion, which is about 530% of the company's total sales during the same period. In addition, Meta's capital expenditures to sales ratio during the same period also came to about 28%, which is a significant increase compared to Microsoft and Meta's capital expenditures to sales averages of 24% and 10%, respectively, over the past 12 years.

 

<'24, Comparison of Capital Expenditures to Sales Ratios of Major Big Tech Companies in the U.S. from '14 to '23>

[Source: WSJ (2024.10)]

 

Data center infrastructure still lacking compared to supply

Despite the big tech companies' aggressive investment in data center infrastructure, data centers in the United States are still in short supply. According to the 2024 First Half North American Data Center Trends Report by global commercial real estate company CBRE, the data center vacancy rate in North America in the first half of this year was recorded at an all-time low of 2.8%, and in particular, the data center vacancy rate in Northern Virginia, known as the "data center capital of the world," was only around 1%. In addition, about 80% of the data center capacity currently under construction has already been leased, so companies hoping to lease data centers need to start leasing 2-4 years in advance of completion in order to prepare for future data center demand.

 

This data center shortage comes amid a worsening power shortage in the United States. On November 1, the Federal Energy Regulatory Commission (FERC) denied a request from power supplier Talen Energy to increase the amount of power it supplies from its Pennsylvania nuclear power plant to a nearby Amazon data center from 300 megawatts (MW) to 480 MW, citing concerns about grid stability and rising electricity prices.

 

Accordingly, it is expected that new data centers to be built in the U.S. in the future will expand to outlying areas such as Indiana, Iowa, and Wyoming, where available power supply is relatively abundant and the degree of imbalance between power demand and supply is not high. However, the power grid infrastructure connecting outlying areas to major cities is still insufficient, and this shortage of power and transmission infrastructure in the U.S. is expected to act as a major factor limiting the expansion of new data center supply. Accordingly, McKinsey predicted that a data center capacity shortage of about 2030 GW could occur in the U.S. alone by 15.

 

<Data Center Distribution by Region in the US>

[Data: McKinsey & Company (2024.10)]

Note: Dark blue indicates an area where data center operations are active, gray indicates an area under development, and light blue indicates an area where development is expected in the future based on abundant power supply.

 

As stable power supply has emerged as a key task for data center operation, some data center operation companies are turning to on-site power generation linked to renewable energy ESS, fuel cells, and, in the long term, small modular reactors (SMRs) instead of connecting to the existing power grid of the power company.

 

For example, in June, ECL, a US data center startup, unveiled the world's first hydrogen fuel cell-based off-grid modular data center in Mountain View, California. Then in September, it announced plans to build 'ECL TerraSite-TX6', a 9GW AI factory data center with a total investment of 80 billion dollars near Houston, Texas. This data center will be powered solely by hydrogen supplied through a separately constructed pipeline without an external power supply, and as a first step, it plans to provide 1MW of capacity to Lambda, a US GPU cloud service company, by next summer.

 

Texas, emerging as a data center hub, encourages construction of its own power plants to stabilize power supply

 

Texas is considered one of the major bases for the data center industry in the United States due to its abundant natural gas resources, vast land, and business-friendly policies. Currently, there are a total of 291 data centers located throughout Texas, and in particular, the northern Texas region, known as Dallas-Fort Worth (DFW), became the second largest data center hub in the United States after northern Virginia with a total capacity of 591 MW as of the first half of this year. In addition, according to The Information, an American IT media outlet, in October, OpenAI announced that it would build the largest AI data center ever in Abilene, Texas by mid-2026 in cooperation with Oracle and AI data center startup Crusoe. In January, Google announced that it would invest $1 billion in two data centers and expand cloud infrastructure in Texas, and the trend of large-scale investment in data centers centered on major companies in the Texas region is expected to continue for the time being.

 

<Data Center Distribution by Region in Texas>

[Data: Data Center Map (2024.11)]

 

<Major data center investment projects under construction in Texas>

(Unit: U$)

[Source: Compilation of data from Crusoe, Stream Data Centers, The Information, Google, Data Center Dynamics, and KOTRA Dallas Trade Center]

 

Meanwhile, in preparation for the expected increase in power grid load due to the increase in such large-scale data center projects, Texas is reportedly preparing a policy to encourage data center operating companies to meet a portion of their own power demand. According to a recent Bloomberg report, the Public Utility Commission of Texas, an agency responsible for regulating, managing, and establishing policies for public services such as electricity, telecommunications, and water in Texas, announced that companies building new AI data centers will have to supply some of their own power if they want to be connected to the Texas power grid within 12 to 15 months. This is interpreted as a response to alleviate the burden of rising power prices that may affect general consumers due to the surge in data center demand and to maintain the stability of the power grid. Meanwhile, it is also evaluated as a measure that can simultaneously achieve the two goals of attracting data centers and ensuring a stable power supply, as it mentions that companies will be allowed to sell their remaining power to the existing power grid.

 

implication

 

According to McKinsey, demand for data centers in the US is expected to grow rapidly from the current 2024 GW in 25 to 2030 GW by 80. In his re-election pledge, Agenda 47, President-elect Trump actively expressed his pledge to encourage fossil fuel mining and support the energy industry by easing environmental regulations to achieve American energy independence, and expressed his ambition to achieve economic revitalization through affordable electricity and energy supply. President-elect Trump has also expressed a strong will to achieve innovation in the AI ​​industry by minimizing AI-related regulations, such as abolishing the 'AI Executive Order' implemented by the previous administration. Therefore, it is expected that the next administration's policies to ensure a stable power supply and promote the development of the AI ​​industry will have a more positive effect on the growth of the US data center industry.

 

Meanwhile, the rapid growth of the U.S. data center market is expected to create new opportunities for our companies in various fields. In particular, as heat management and energy efficiency improvement are emerging as key tasks for data center operation, our companies are expected to do well in the air conditioning and cooling solution sector based on their technological prowess and experience. In addition, the movement to improve the power grid infrastructure for stable power supply to data centers is expected to provide positive opportunities for companies supplying power equipment such as transformers and transmission cables, and in the long term, demand for SMR-related facilities and technologies, which are attracting attention as a stable carbon-free power source, is expected to increase.

 

However, there is also a possibility that protectionist sentiment, such as strengthened tariff policies, will be reinforced after the new government takes office. Therefore, Korean companies seeking to advance into the U.S. data center industry will need to respond flexibly to future policy changes through strategic responses such as strengthening local supply chains through expansion of production facilities in the U.S. and developing high value-added products that emphasize energy efficiency and the environment.

 

Source: WSJ, McKinsey & Company, Data Center Map, Crusoe, Stream Data Centers, The Information, Google, Data Center Dynamics, KOTRA Dallas Trade Center data

 

Source: KOTRA