[Kangkook Lee's Column] Is AI Boosting Economic Growth? ... 'Still Negative'

Dr. Kangkook Lee is Professor of College of Economics at Ritsumeikan University in Japan and Affiliated Research Fellow at CUKPE

Since the launch of 'ChatGPT 3.5' in late 2022, artificial intelligence (AI) has suddenly become a reality rather than a future. As humanity asks questions to AI that has become smarter than humans and listens to their answers, we have cheered on one hand that a new era has begun, while on the other, we have started to worry about the future. Optimists, including economists, argue that AI will replace knowledge work, significantly increase labor productivity, and stimulate economic growth. However, pessimists criticize that the advancement of AI may lead to mass unemployment and deepen inequality, potentially bringing about a "dystopia" (a bleak future). Furthermore, voices calling for a slowdown in the pace of AI development have recently been growing louder, warning that if AI evolves autonomously and escapes human control, it could become a major danger. What kind of future will AI truly bring us? This is likely one of the most difficult questions our era must answer. In this article, let us first examine the impact AI is currently having on economic growth.

AI investment-led economic growth

Since the emergence of generative AI, various scenarios have been presented suggesting that AI will significantly boost productivity and economic growth in the future. Whether such a rosy future will actually materialize remains to be seen, but we must first examine the current situation. Is AI currently promoting economic growth? The answer is "yes and no."

First, a positive aspect is that the current AI boom is boosting growth rates. Looking at the United States first, the annualized real GDP growth rate for the first quarter of 2026 was 2.1%, and for the second quarter, it was 1.5%. However, a particularly interesting point regarding the recent economic growth in the U.S. is that investment related to AI has become very important.

In September 2025, Harvard University Professor Jason Furman garnered significant attention by pointing out that 92% of U.S. economic growth in the first half of that year was driven by AI-related investment, whereas excluding that component would have resulted in a very low growth rate of 0.1%. The implication was that the economy grew thanks to AI investment, even though other sectors remained stagnant. This is because the majority of the increase in U.S. real GDP at the time could be explained by the expansion of investment in AI-related information processing facilities and software. U.S. economic growth in the first half of 2026 was also led by increased corporate investment, including data centers by Big Tech companies. Calculated using a method similar to his, the increase in investment in information devices and software accounted for 69% of the GDP growth.

Of course, not all of such investments are related to AI, and it must also be considered that the trade deficit resulting from the fact that the U.S. imports of information devices exceed exports has a negative impact on growth. Therefore, there is debate regarding the current impact of AI-related investments on U.S. economic growth, but recent reports by JPMorgan and others point out that the contribution of AI-related investments to economic growth has increased to 'about one-third to half.'

The same applies to the South Korean economy, which has recently entered a semiconductor-led boom against the backdrop of the 'AI boom.' South Korea's real GDP growth rate in the second quarter was 3.7% year-on-year, while its nominal growth rate was a remarkable 26%, driven by a significant increase in semiconductor exports. In the first half of the year, semiconductor exports surged by 163% year-on-year, total exports rose by 48%, and the trade surplus increased by approximately five times. Net exports contributed about 47% to overall economic growth in the first quarter and about 43% in the second quarter. Ultimately, as seen in the United States and South Korea, the AI ​​boom can boost GDP and stimulate economic growth by expanding related investments and exports.

Of course, this acceleration of growth is limited to countries with a large share of related industries such as AI and semiconductors, and caution must be exercised that if the AI ​​bubble bursts or market conditions deteriorate in the future, it could impact growth. As pointed out in a recent report by the Bank for International Settlements (BIS), the scale of AI-related investment has grown so rapidly—even compared to past new technologies like railways—that there are opinions that a significant portion constitutes overinvestment that is difficult to generate profitability. As shown in the figure below, a recent study presented at a Brookings Institute conference reports that the scale of AI infrastructure investment in the United States will reach 3.6% of annual GDP between 2025 and 2032, exceeding the scale of any infrastructure investment in the past.

Consequently, questions are being raised as to whether this massive investment in AI is immediately sustainable. The Economist recently pointed out that while the revenue and profits of AI service companies like OpenAI are growing rapidly, the pace is not fast enough to justify the surge in AI-related investments. Furthermore, according to the aforementioned study, the top five hyperscalers are actively raising funds through debt and external investments, as capital expenditures are projected to significantly exceed operating cash flow by 2026. However, the problem is that the multilayered and complex financial structures of data centers, including leases, do not appear on tech companies' balance sheets and entail significant risks. Historically, while innovative new technologies have led to long-term growth, they have always been accompanied by overinvestment or bubbles; when these bubbles burst, the economy has often faced a crisis. We must closely monitor how severe such economic fluctuations might manifest in the current situation, which is driven by AI investment.

Where does the effect of increased productivity lie?

 However, we must not forget that the economic growth resulting from the recent increase in AI-related investment differs from the path envisioned by optimists who praise AI. The commonly discussed rosy future regarding economic growth is a scenario in which labor productivity (GDP/working hours) is expected to rise significantly as AI replaces human mental labor to achieve greater production with the same amount of work. In particular, AI is expected to drive an increase in 'total factor productivity growth' (TFP), which represents pure technical and organizational productivity—calculated by subtracting increases in labor and capital inputs from GDP growth. Indeed, several experimental studies conducted at the workplace level, such as in customer service or software development , report that the use of generative AI like ChatGPT has increased worker productivity at a micro level.

However, is the recent advancement of AI technology actually driving productivity growth across the entire economy? Unfortunately, the answer to this question is still negative. First, as shown in the graph below, an analysis of long-term total factor productivity data for the entire U.S. economy does not appear to show a higher growth rate before and after 2023, when generative AI was introduced. In particular, after the pandemic, the growth in total factor productivity adjusted for capacity utilization was relatively low. This suggests that productivity increased as a result of expanded demand and increased sales driven by economic recovery, regardless of technological innovation.

A report by the British investment bank Barclays last July pointed out that the adoption of AI technology in businesses and households is gradual, and that evidence of a resulting structural increase in productivity is scarce. The study emphasized that the temporary rise in total factor productivity following the pandemic is closely linked to increases in capacity utilization, and that various empirical analyses have shown that productivity is not rising faster in industries more exposed to AI technology. While this may be a disappointing result for optimists, other studies present similar conclusions.

According to a study by the U.S. Federal Reserve released last July, while AI-related investments have increased since 2024, they have not stimulated overall labor productivity growth. The study reports that there has been no significant change in labor productivity trends over time, particularly in industries with high AI exposure such as information technology and finance, as well as those with low exposure such as utilities and construction. Of course, the authors point out that the gap between industries has recently widened slightly, so attention should be paid to future changes.

Ultimately, even if AI is generating productivity effects at the level of individual jobs, it is not yet leading to an increase in productivity for the entire economy. The Federal Reserve points out that the reason for this is that while micro-level studies primarily measure productivity at the job level, in reality, factors such as coordination costs or bottlenecks in the production process may prevent these effects from translating into improvements in the productivity of the entire firm. In the past, there were also high expectations that information technologies like computers would significantly boost productivity, but their effects in reality were limited. Professor Solow, a leading expert on economic growth theory, even wrote in 1987, "We can see the computer age everywhere, but not in productivity statistics."

Then, will AI be similar? Of course, it is still too early to be disappointed. This is because various studies predict that AI could bring about significant growth effects in the future, and active debates are underway within the economics community. In the following article, let's forecast the economic growth effects of AI, focusing on future scenarios rather than the present.

(This article was originally published as a column in Hankyoreh in Korean and translated into English with the help of Google Translate. The views expressed in this article are those of the author(s) and do not necessarily represent the official stance of the center.)