A trader works on the floor of the New York Stock Exchange in New York on Oct. 12. Markets anticipate a busy week ahead of earnings results and inflation data.UPI/YONHAP
Choy Jung-hyuck
The author is a professor of economics, finance and asset management at Hanyang Cyber University.
Despite sharp increases in interest rates and oil prices, the U.S. S&P 500 index has remained resilient, near record highs. Beneath the surface, however, the picture is less encouraging.
According to Goldman Sachs, the median S&P 500 constituent was trading 16 percent below its 52-week high at the end of September. Although the overall index remained close to its peak, many individual stocks had already undergone substantial corrections. Strong performances by a handful of large companies were masking broader market weakness.
Artificial intelligence is the principal force supporting the index. JPMorgan reported that the 10 largest companies accounted for 40.6 percent of the S&P 500's total market capitalization at the end of September. Including businesses across AI-related industries, such as semiconductors, cloud computing, software and electricity, raised that proportion to 53.2 percent.
Another important development concerns valuations. The index's forward 12-month price-to-earnings ratio, or PER, declined 14 percent, from 22 at the end of last year to 19 in September.
The ratio measures stock prices against expected corporate earnings over the following year. Its decline suggests that earnings forecasts improved faster than share prices. Rising interest rates, higher oil prices and uncertainty over returns on AI investment have restrained further market gains.
Concentration risk is even more pronounced in Korea's stock market.
Unlike the United States, where AI-related businesses span semiconductor design, cloud services and other activities, Korea relies heavily on memory chips. This leaves the domestic market particularly exposed to fluctuations in semiconductor demand and prices.
According to the Korea Financial Investment Association, Samsung Electronics and SK hynix together accounted for an average 52.8 percent of the Kospi's market capitalization in September. A correction in either company's shares could therefore have substantial consequences for the entire market.
Addressing this risk requires a more sophisticated understanding of diversification.
Investors traditionally diversify by holding numerous stocks across different industries. But the number of holdings matters less than the economic forces driving their earnings. A portfolio containing dozens of companies may still be heavily concentrated if most depend on the same AI investment cycle.
Past market movements offer useful lessons. During the DeepSeek shock in January last year and the technology stock correction in July this year, low-volatility and high-dividend stocks performed relatively well.
These companies generally experience smaller price fluctuations or distribute a larger proportion of profits to shareholders, making them potential alternatives when technology shares come under pressure.
In 2022, when interest rates surged, the S&P 500 fell 18.1 percent, while the Invesco S&P 500 High Dividend Low Volatility ETF, known as SPHD, gained 0.6 percent. Exchange-traded funds, or ETFs, allow investors to hold baskets of securities through a single investment product.
Such strategies are not without drawbacks. When technology stocks rebounded strongly in 2023, SPHD significantly underperformed the broader market.
Goldman Sachs introduced an S&P 500 index excluding AI-related companies, known as SPXXAI, in February. Rather than betting against AI, the index offers exposure to businesses whose earnings depend on different economic drivers.
The objective is not to abandon AI's growth potential but to reduce excessive reliance on a single investment theme.
Investors confident in AI's long-term prospects should therefore prepare for short-term corrections without sacrificing opportunities for growth. By combining AI holdings with companies supported by independent sources of earnings, they can reduce concentration risk while remaining positioned to benefit from technological progress over time.
This approach also requires recognizing that diversification cannot eliminate market losses. Different assets may decline together during periods of financial stress. Nevertheless, spreading investments across companies with distinct earnings drivers can help limit the damage caused by a downturn in any one sector. The central question is therefore not simply how many stocks an investor owns, but whether those holdings provide genuinely different sources of risk and return over time and across market cycles.
AI 강세장, 분산투자를 다시 생각할 때
최정혁 한양사이버대학교 경제금융자산관리학과 교수
금리와 유가 급등에도 불구하고 미국 S&P500 지수는 사상 최고치 부근에서 견조한 흐름을 이어가고 있다. 그러나 수면 아래의 상황은 다르다. 지난 9월 말 골드만삭스의 분석에 따르면 S&P500 구성 종목의 52주 최고가 대비 주가 변동률 중간값은 -16%였다. 지수는 사상 최고치 부근에 머물러 있지만, 상당수 종목은 이미 큰 폭의 조정을 겪은 것이다. 소수 대형주의 강세가 시장 내부의 부진을 가리고 있는 셈이다.
지수를 지지하는 것은 인공지능(AI)이다. JP모건에 따르면 지난 9월 말 기준 S&P500 시가총액에서 상위 10개 기업의 비중은 40.6%였다. 반도체·클라우드·소프트웨어·전력 등 AI 관련 산업을 폭넓게 포함하면 53.2%에 달했다.
주목할 점은 지수는 견조하지만 12개월 선행 주가수익비율(PER)은 지난해 말 22배에서 지난 9월 말 19배로 14% 하락했다는 사실이다. 예상 이익이 빠르게 개선됐지만, 금리·유가 상승에 따른 부담과 AI 투자 수익성의 불확실성이 주가 상승을 제한한 결과다.
AI 집중 위험은 국내 증시에서 더욱 뚜렷하게 드러난다. 미국은 반도체 설계, 클라우드를 포함해 AI 관련 기업의 사업과 수익원이 비교적 다양하다. 반면 한국은 메모리 반도체 의존도가 높다. 금융투자협회에 따르면 9월 삼성전자와 SK하이닉스의 코스피 시가총액 비중은 월평균 52.8%였다.
이러한 위험에 분산투자로 대응하려면 종목이나 업종의 수보다 기업이익을 움직이는 요인에 따라 다양화해야 한다. 많은 종목과 업종에 나눠 투자하더라도 대다수가 AI 관련 기업이라면 분산 효과를 기대하기 어렵다.
과거 시장 움직임에서도 아이디어를 찾을 수 있다. 지난해 1월 딥시크 충격과 올해 7월 기술주 조정에서 저변동성·고배당주가 상대적으로 양호한 성과를 보였다. 금리가 급등했던 2022년에도 S&P500이 18.1% 하락했지만, 저변동성·고배당 ETF(SPHD)는 0.6% 상승했다. 물론 기술주가 강하게 반등한 2023년에는 SPHD가 시장 수익률에 크게 뒤처졌다.
지난 2월 골드만삭스가 S&P500에서 AI 관련 기업을 제외한 지수(SPXXAI)를 선보인 것도 같은 맥락이다. 이는 AI 하락에 베팅하기보다 AI와 다른 이익 원천을 가진 기업에 투자하도록 설계된 지수다. AI 성장 기회를 유지하면서 특정 산업에 대한 의존도를 낮추려는 접근이다.
AI의 성장 가능성을 믿는 투자자라면 단기 조정에 대비하면서 투자 기회를 유지하는 전략이 필요하다. AI와 다른 수익원을 가진 기업으로 분산 투자해 집중 위험을 낮추고, 장기적인 성장에 참여할 수 있는 투자 구조를 만들어야 한다.
This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
