As generative artificial intelligence (AI) becomes mainstream, global information technology companies are rapidly expanding their investments in AI infrastructure. South Korea's economy is highly exposed to the global AI investment cycle, which analysts say can serve as both a growth driver and a risk factor.
According to a report by the National Assembly Budget Office titled 'Expansion of AI Investment and Domestic Equipment Investment Risks,' global capital expenditures related to AI are expanding at a pace faster than previous major investment booms. The increase in AI infrastructure investment resembles the infrastructure investment cycles seen during historical transitions to general-purpose technologies.
According to Perez, who presented this as a technological revolution process, all five technological revolutions since the Industrial Revolution have gone through a 'installation phase,' where technological infrastructure is built, and a 'deployment phase,' where technology spreads throughout the economy. A turning point accompanied by financial crises appears between these two phases.
The installation phase is led by financial capital, transitioning from an 'emergence phase' where initial technological innovations appear to a 'frenzy phase' where speculative capital concentrates on new technology infrastructure. During this process, infrastructure can be overbuilt, and asset prices may soar.
The report analyzes that the current AI industry is showing signs similar to Perez's frenzy phase, with a simultaneous surge in infrastructure investment by hyperscalers and rising valuations of AI-related companies. The influx of financial capital is outpacing the actual spread of technology, and future expectations may be reflected in asset prices, while improvements in real productivity have yet to be clearly confirmed.
Indeed, the combined capital investments of five hyperscalers—Amazon, Microsoft, Google, Meta, and Oracle—have increased from $150 billion in 2023 to an estimated $735 billion in 2024, and $395 billion in 2025. This marks an approximate fivefold increase over three years since 2023.
The pace of investment expansion is also accelerating. The capital investment growth rate is projected to rise to 50.8% in 2024, 74.7% in 2025, and 86.0% this year. The proportion of capital investment relative to operating cash flow for the four major companies has increased from 39.0% in the first quarter of 2024 to 69.7% in the first quarter of this year, indicating that about 70% of generated cash is being reinvested in equipment.
However, the report suggests that the expansion of AI investment cannot be viewed merely as excessive investment. If demand for AI computing grows faster than expected, the costs of customer attrition due to equipment shortages may outweigh the costs of overbuilding. The competition to secure infrastructure ahead of rivals is also cited as a factor driving investment.
Future investment plans are expected to remain substantial. The combined capital investments of the five hyperscalers are projected to increase from $735 billion in 2026 to $1.495 trillion by 2030. However, the growth rate is expected to decline from 86.0% this year to 6.4% by 2030, indicating a gradual slowdown in the pace of investment expansion.
The concern lies in how adjustments in the investment cycle will impact domestic equipment investment. As of 2024, AI-related companies account for 26% of equipment investment among South Korean listed companies, higher than the 23% in the United States, while other countries show less than 5%. Notably, domestic AI-related equipment investment is concentrated in a few semiconductor companies like Samsung Electronics and SK Hynix, meaning individual companies' investment decisions significantly influence overall equipment investment and economic trends.
Domestic equipment investment increased by 5.4% in the first half of this year compared to the previous year, marking a turnaround, and the government is also pursuing large-scale investment plans centered on semiconductors. While continued global AI investment could positively impact domestic equipment investment and exports, a downturn in the global investment cycle could lead to concentrated downward shocks.
The report also identifies risk factors in the financial structures of hyperscalers. While operating profitability is maintained, the expansion of AI investment is reducing free cash flow, and the sources of investment funding, which were previously centered on internal cash, are shifting to external financing such as corporate bonds. Physical constraints on data center construction, including power grid and transmission capacity shortages, cooling facilities, permits, and electricity costs, are also cited as factors that could limit investment expansion.
Economic analyst Hwang So-jeong stated, "Given that the global expansion of AI investment has a positive effect on domestic equipment investment, it is necessary to examine both the sustainability of the investment cycle and its potential for adjustment." She emphasized the need to continuously monitor global investment trends and changes in major companies' investment plans, especially considering the concentration of AI investment in a few domestic companies.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.