Heo Jae-hwan, Head of Macro Analysis at Eugene Investment & Securities, published an analysis outlining three potential scenarios for how the AI supercycle might end, following severe volatility in semiconductors and Korean stocks during 6-7월. The analysis examines whether the current AI investment boom could conclude through a 1999-style supply glut, a 2022-style interest rate and demand slowdown, or a 2008-style funding crisis. Heo's assessment draws parallels to historical tech cycles after investors experienced significant market turbulence, though he notes that long-term downward trend signals have not appeared in Korean stocks.
Three Historical Scenarios for AI Supercycle Ending
Heo identified three potential pathways based on past technology cycles. The first scenario mirrors the 1999-2000 dotcom bubble, where internet infrastructure including telecommunications equipment, internet relays, and fiber optic cables were supplied far beyond actual demand. During that period, internet portal companies and equipment makers like Cisco saw stock prices surge before collapsing. Heo noted that US technology investment as a share of GDP is currently approaching 1999 levels, with data center vacancy rates and GPU rental price declines serving as key indicators of potential oversupply, though severe oversupply signs have not yet emerged.
The second scenario resembles 2022, when the Federal Reserve's rapid interest rate increases prompted companies to cut costs and reduce IT investments including cloud services. Heo stated that the current environment differs as rate increase speeds are not expected to be as steep, with big tech operating activities remaining solid and data center investment intentions intact. However, he highlighted a growing gap between investment growth at 81% year-over-year versus revenue growth of only 21%.
The third scenario parallels the 2008 financial crisis, when subprime mortgage defaults spread throughout the financial system. Heo reported that debt levels at major big tech companies including Oracle are rising rapidly, with corporate bond issuance growth causing debt increase speeds to approach investment increase speeds. He noted that CLO and ABS issuance is expanding around private lending and neo-cloud companies.
Current Investment-Revenue Gap and Debt Expansion
Free cash flow has been declining since 2025 according to Heo's analysis. He stated that investors need to monitor whether big tech cloud revenue slows during future earnings seasons. Oracle and Coreweave, which carry significant debt burdens, have seen stock prices decline over 60% from their peaks. Heo concluded that while discussing the end of the AI cycle may be premature from a technology perspective, warning signs have increased compared to six months or one year ago.
FAQ
What three scenarios did Eugene Investment's Heo Jae-hwan identify for how the AI supercycle might end?
Heo Jae-hwan outlined three potential scenarios: a 1999-style supply glut where AI infrastructure is overbuilt beyond actual demand, a 2022-style interest rate increase and demand slowdown that reduces corporate IT spending, and a 2008-style funding crisis where debt accumulation creates financial system risks. Each scenario is based on a historical technology or financial cycle.
What investment-revenue gap did Heo identify in current big tech companies?
Heo reported that big tech investment growth stands at 81% year-over-year while revenue growth is only 21%, creating a significant gap between capital expenditure and actual sales performance. He also noted that free cash flow has been declining since 2025, and debt levels at companies including Oracle are rising rapidly with corporate bond issuance growth causing debt increase speeds to approach investment increase speeds.
