The AI Bubble Debate Was Looking at the Wrong Market

The people who called the AI bubble talked about America. In May, Michael Burry and Paul Tudor Jones compared the AI rally to the dot-com bubble, pointing at semiconductor and AI-infrastructure valuations in U.S. markets ^[1]. In July, the Treasury went further — a draft report warning of systemic risk to the U.S. economy, with AI firms entangled in stock markets, chip manufacturing, utilities, and private credit, all framed as domestic exposure ^[2]. Neither the investors nor the government agency mentioned Korea, or any Asian market, at all. They were watching the wrong screen. On July 8, South Korea's KOSPI closed at 7,246.79, down 20 percent from its June 25 peak above 9,000 — the first major index to enter bear territory in the AI correction cycle ^[3]. The crash was attributed explicitly to both AI-trade reassessment and geopolitical tensions from U.S. airstrikes on Iran and the revocation of an oil-sale waiver ^[3]. Two shocks, compounding. Neither one appeared in any American bubble diagnosis. The first shock was the AI supply chain, and it hit Korea harder because Korea sits closer to the demand side. Samsung and SK Hynix anchor the KOSPI's tech-heavy weighting; when SK Hynix slowed expansion of its next-generation HBM4 memory chips because production forecasts for Nvidia's upcoming Rubin processor were trending downward, the KOSPI crashed 10 percent on June 22, triggering a circuit breaker ^[4]. The same global tech rout that day clipped the Nasdaq 2.2 percent and the S&P 500 1.4 percent ^[4]. Neither American index came close to bear territory. The AI correction alone was not enough to break a U.S. market. But it was enough to cut Korea's index in half the magnitude, because Korea's largest companies live inside the Nvidia supply chain in a way America's broad indexes do not.

what actually broke the market that the bubble callers were watching

The American bubble thesis: Risk to U.S. stock markets, chip manufacturers, utilities, and private credit — the Treasury's framing, echoed by Burry and Jones. All channels run through domestic U.S. exposure. ^[2][1]

What actually broke: Korea's KOSPI, down 20 percent in two weeks — driven by AI supply-chain concentration (SK Hynix HBM4 slowdown on softening Nvidia Rubin demand) compounded by oil-import vulnerability (24.7 percent oil-price spike, 30-month-high inflation, 70 trillion won in capital outflows). ^[3][4][5]

The second shock was oil, and it hit Korea because Korea is heavily dependent on imported energy. South Korea's government had warned back in April that Middle East tensions created increasing downside risks to growth, citing rising oil prices and supply disruptions ^[6]. By June, the warning had become reality. The U.S.-Iran escalation — airstrikes on Iran and the revocation of an oil-sale waiver — sent oil prices surging. South Korea's inflation hit a 30-month high of 3.2 percent in June, driven primarily by a 24.7 percent spike in oil products ^[5]. Foreign investors pulled 70 trillion won out of Korean markets over 20 consecutive sessions, weakening the won to near its lowest level since March 2009 ^[7]. The KOSPI fell in two waves. The first came June 22-23, the pure AI-trade repricing: SK Hynix slowing HBM4, the circuit breaker, a 10 percent drop ^[4]. Then the index briefly rebounded when an Iran peace deal eased geopolitical tensions and Samsung surged on buyback reports ^[8]. That respite ended in July, when U.S. airstrikes on Iran and the oil-waiver revocation delivered the second wave — the oil-import vulnerability shock that pushed the KOSPI through the 20 percent threshold into bear territory ^[3]. The first wave was an AI story. The second was a geopolitical story. Korea broke because it caught both. The decoupling was visible and stark. On July 6-7, the Dow hit a record and the Nasdaq rose 1.1 percent ^[9]. The same day, the KOSPI dropped 6 percent and Samsung plunged 7 percent despite reporting a 19-fold profit jump ^[9]. Wall Street rallied. Seoul cratered. The institutions calling the bubble were staring at the market that was going up while the market that was breaking down was one they had never mentioned. Not every Asian market followed Korea down. Hong Kong's Hang Seng rose nearly 2 percent on July 8, lifted by reports that China's DeepSeek is developing its own AI chip ^[10]. India's Sensex dipped only half a percent, partly cushioned by Saudi Arabia cutting August crude prices for Asian buyers ^[10]. The damage concentrated where a market was simultaneously exposed to the U.S.-led AI supply chain and dependent on Middle East oil — a combination that describes Korea more precisely than any other major economy, and one that no American bubble diagnosis had any reason to consider. The Treasury report and the hedge-fund warnings were not wrong about the risks they identified. The AI buildout has concentrated valuations, stretched credit, and loaded utility demand in ways that do echo the late-1990s. But they diagnosed a disease by studying only the patient they knew. The first body to show symptoms was one they had never examined — and the symptom that proved fatal, the compounding of AI exposure with oil-import fragility, was a vulnerability they had no framework to see.