At a glance
• According to Stanford HAI, corporate AI investment reached $252.3 billion in 2024, and the share of respondents saying their organizations use AI rose from 55% in 2023 to 78% in 2024.
• According to the IEA, data centers consumed 415 TWh in 2024, about 1.5% of global electricity consumption, and their electricity use has grown by an average of about 12% a year since 2017.
• A KDI analysis found that, as of 2023, 38.8% of jobs in Korea could technically have 70% or more of their tasks automated, yet a survey of 800 companies at the end of September 2023 found an AI adoption rate of 2.3%.

1. Why AI and the Economy Now Go Hand in Hand

Numbers are the reason artificial intelligence can no longer be read as a story about technology alone. The International Energy Agency (IEA) notes that the market capitalization of AI-related companies in the S&P 500 has grown by about $12 trillion since 2022. The Stanford Institute for Human-Centered AI (HAI) put corporate AI investment at $252.3 billion in 2024. According to the same source, private investment rose 44.5% from the previous year and mergers and acquisitions rose 12.1%. When this much money flows into a single field, it is natural to ask what traces it leaves on jobs and productivity, on electricity and on wages.

These figures, however, come from different institutions studying different populations at different times. This article first looks at how exposed workers are to generative AI and how much they actually use it. It then turns in order to investment and changes in working hours, adoption and costs inside companies, electricity and supply chains, and how gains and burdens are distributed. Next it examines the Bank of Korea's simulations of the Korean economy and recent institutional changes, before closing with three takeaways. Throughout, we have tried to separate what the sources say from our own interpretation.

2. The ILO's 2025 Update and a Bank of Korea Survey: One in Four Workers Exposed, 63.5% of Korean Workers Using It

In its 2025 update, the International Labour Organization (ILO) reported that one in four workers worldwide is in an occupation with some degree of exposure to generative AI. The analysis is described as examining about 30,000 tasks at the six-digit occupational classification level, combining human and AI judgment. The average automation score was 0.29, slightly lower than 0.30 in 2023. At the same time, as capabilities for generating voice, images and video have grown, automation scores rose for many tasks in media- and web-related occupations. This suggests that even when the average barely moves, trends can differ from one occupation to another.

A Bank of Korea household survey found that 63.5% of workers were using generative AI. Counting only work-related use, the rate was 51.8%. The Bank of Korea compared this with the United States, saying Korea's usage rate is roughly twice as high, and said the technology is spreading eight times faster than the internet did. The two surveys are worth reading together because they set a measure of potential, exposure, alongside a measure of actual behavior, use. Still, the ILO studied occupations worldwide while the Bank of Korea surveyed workers in Korea, so the two figures cannot be compared directly.

AI-generated illustration of a server rack aisle in a data center at dawn
Recreated illustration · Not an actual photograph — concept image of a server rack aisle in a data center at dawn

3. Stanford HAI, the IEA and the Bank of Korea: $33.9 Billion in Generative AI Investment and 3.8% Less Working Time

According to Stanford HAI, private investment in generative AI reached $33.9 billion in 2024. That was up 18.7% from 2023 and more than 8.5 times the 2022 level. In the same year, US private AI investment was $109.1 billion, about 12 times China's ($9.3 billion) and 24 times the UK's ($4.5 billion). The IEA calculated that global data center investment nearly doubled after 2022, reaching $500 billion in 2024. How much of the IEA's data center investment figure is for AI, however, cannot be determined from this source alone.

A Bank of Korea survey offers a partial glimpse of how this investment is changing working time. According to the survey, using generative AI cut working hours by an average of 3.8%, or 1.5 hours in a 40-hour week. The resulting potential productivity gain was estimated at 1.0%. The reduction in working time was larger for less experienced workers, which the Bank of Korea interpreted as a leveling effect that narrows skill gaps. Because the word "potential" is attached, it is more prudent to treat how much the time saved has actually boosted output as still at the estimation stage.

4. Stanford HAI, the Bank of Korea and KDI: 78% Organizational AI Use, 11% Exposure to Physical AI, 38.8% of Jobs Automatable

In Stanford HAI's data, the share of respondents saying their organizations use AI jumped from 55% in 2023 to 78% in 2024. Among organizations using AI in service operations, 49% of respondents reported cost savings, compared with 43% in supply chain management and 41% in software engineering. Most respondents who reported savings, however, said they amounted to less than 10%. The survey suggests that while the number of organizations using AI has grown quickly, the cost savings are not yet large.

There are also areas where people work alongside machines. According to Stanford HAI, China installed 276,300 industrial robots in 2023, six times as many as Japan and 7.3 times as many as the United States. The Bank of Korea estimated that the share of workers who work with autonomous robots, that is, who are exposed to physical AI, is currently 11% and will rise to 27%. KDI's analysis found that, as of 2023, 38.8% of jobs in Korea could technically have 70% or more of their tasks automated. Because of the qualifier "technically," caution is needed before reading this figure as the number of jobs that will actually be replaced.

AI-generated illustration of an office worker at a laptop in the afternoon, seen from behind
Recreated illustration · Not an actual photograph — concept image of an office worker at a laptop in the afternoon, seen from behind

5. The IEA's 415 TWh of Data Center Electricity in 2024 and KDI's 2.3% AI Adoption Rate Among Korean Firms

According to the IEA, data centers used 415 TWh in 2024, about 1.5% of global electricity consumption. Since 2017, data center electricity consumption has grown by an average of about 12% a year, more than four times faster than total electricity consumption. Stanford HAI's data also notes that Microsoft announced a $1.6 billion deal to restart a reactor at Three Mile Island to secure power for AI. The IEA noted that AI-focused data centers are 10 times more capital-intensive than aluminum smelters, and that China currently accounts for about 99% of the world's refined gallium supply. These figures show that the AI economy is also bound up with physical constraints such as electricity and materials supply.

Adoption figures from Korean companies tell a rather different story. In the Survey on Informatization Statistics cited by KDI, the AI adoption rate among private companies with 10 or more employees was 2.7% as of the end of 2021. Among large companies with 250 or more employees alone, the rate reached about 20%. A KDI survey of 800 Korean companies at the end of September 2023 likewise found a low adoption rate of 2.3%. This points to a gap between global investment trends and adoption across Korean businesses as a whole, with large differences by company size.

6. The JRC and KDI: EU's 7% Share of Generative AI Activity, Teachers More Exposed Than 90% of Occupations, and Negative Effects Concentrated Among Younger Workers

The European Commission's Joint Research Centre (JRC) reported that the EU ranks third with 7% of global generative AI activity, compared with 60% for China and 12% for the United States. Europe produced 21% of generative AI research papers, ranking second in the world, with more than 3,000 papers in 2023. The EU's patent filings, by contrast, accounted for only 2% of the global total. These figures suggest that research output and patent filings do not necessarily move in the same direction.

The JRC's material also covers education and fairness. In JRC research, teachers had higher AI exposure than 90% of other occupations. In addition, an experiment with AI models used for financial decision-making revealed a gender bias favoring men by about 4%. These cases suggest that AI's economic impact also needs to be examined in areas where people deal directly with others, such as classrooms and bank counters.

In KDI's analysis, a higher regional AI impact rate had no effect on whether people were in wage employment overall, but it was estimated to reduce workers' wages, particularly women's average wages. By occupation, employment in professional jobs rose, led by information and communications specialists and technical occupations, while employment in elementary service and labor jobs fell. The negative effects on employment and wages were concentrated among younger workers, such as men aged 30–44 and women aged 15–29, and among those with a junior college degree or higher. For middle-aged and older workers and those with a high school education or less, there was no change or even a positive effect. This shows that even when the overall totals look similar, the direction of the impact can diverge across groups.

AI-generated illustration of industrial robot arms on a factory assembly line in the morning
Recreated illustration · Not an actual photograph — concept image of industrial robot arms on a factory assembly line in the morning

7. Bank of Korea Simulations and the AI Framework Act: A Possible 4.2–12.6% Boost to GDP, Taking Effect in January 2026

In a Bank of Korea model simulation, AI adoption could raise Korea's productivity by 1.1–3.2% and its GDP by 4.2–12.6%. The productivity gains, however, were not evenly spread across all firms; they were most pronounced at large companies and long-established firms. In a 2023 IBM survey cited by the Bank of Korea, 48% of large Korean companies said they had already adopted AI. The Bank of Korea also pointed out that Korea accounted for about 23% of global semiconductor exports as of the first half of 2024. Because the IBM survey and the KDI survey differ in their targets and methods, the figures of 48% and 2.3% cannot be compared directly.

Figures on the worker side also need to be considered. The Bank of Korea estimated that 24% of Korean workers could benefit from AI adoption, while 27% are likely to be negatively affected. In a Bank of Korea survey, 48.1% of workers said AI would have a positive effect on society, while 17.5% said it would be negative. Some 32.3% of workers expressed willingness to contribute to an AI technology development fund, and factoring in their willingness to pay, it was calculated that about 38 trillion won could be raised over five years.

On the institutional side, legislation is now in place. According to the Korea Ministry of Government Legislation's National Law Information Center, the Framework Act on the Development of Artificial Intelligence and the Establishment of a Foundation for Trust was enacted on January 21, 2025, as Act No. 20676. It is set to take effect on January 22, 2026. How the Act's specific provisions will affect investment or employment cannot be determined from the sources used in this article alone.

8. Three Takeaways from Reading the Numbers Together

First, there appears to be a gap between the scale of investment and the effects visible on the ground. A roughly $12 trillion rise in market capitalization and $252.3 billion in corporate AI investment are big numbers. Yet most organizations that reported cost savings said they were under 10%, and the potential productivity gain estimated by the Bank of Korea was 1.0%. In the KDI survey, the adoption rate among Korean companies was 2.3%. It can therefore be argued that it is too early to gauge economy-wide change from investment figures alone.

Second, the impact does not appear to be evenly distributed. In the Bank of Korea's analysis, productivity gains were most pronounced at large and long-established companies. KDI's analysis estimated negative effects among younger workers and those with a junior college degree or higher, as well as a decline in women's average wages. At the same time, the Bank of Korea interpreted the larger reduction in working hours among less experienced workers as a leveling effect. In short, the same technology can appear to both widen and narrow gaps depending on which indicator one looks at.

Third, Korea's circumstances seem to hold both opportunities and challenges. The Bank of Korea noted that Korea ranked 15th out of 165 countries on an AI preparedness index. It also estimated that, amid population aging, Korea's GDP would shrink by 16.5% over 2023–2050 without AI adoption. Because these figures are estimates that rely on models and assumptions, they are better read as conditional calculations than as settled forecasts. Looking at investment, adoption, distribution and institutions together can be a starting point for understanding AI and the economy.

Related reading: AI and Quantum Computing: What Quantum Machine Learning Can and Cannot Do Yet · Personal Names Around the World: Beyond the Western Two-Field Format

Sources