Noah Smith cuts through the noise of daily market fluctuations to argue that South Korea's recent stock collapse isn't just a local anomaly, but a potential canary in the coal mine for the global artificial intelligence investment boom. He challenges the comforting narrative that this was merely a case of "irrational exuberance" by regular investors, suggesting instead that the crash exposes a fundamental disconnect between the trillions spent on AI infrastructure and the actual revenue those systems are generating. For busy observers trying to parse whether the AI revolution is a new industrial age or a speculative bubble, this analysis offers a crucial reality check on the math behind the hype.
The Mechanics of a Leverage-Driven Crash
Smith begins by dismantling the idea that stock market movements are purely rational responses to earnings. He notes that "explaining stock market movements is always a little bit of a fool's errand," yet he proceeds to dissect the specific alchemy that turned South Korea's market into a volatile asset class. The fundamental story was undeniably strong: the AI boom created a "mother of all memory booms," with companies like SK Hynix seeing operating profits surge from under $10 billion to over $35 billion in a single year. This wasn't just a stock rally; it was a reflection of a genuine supply shortage, a dynamic familiar to those studying the 2025–2026 global memory crisis where the concentration of production in the hands of a few Korean firms gave them unprecedented pricing power.
However, Smith argues that the sheer scale of the price increase was driven less by these fundamentals and more by the introduction of dangerous financial tools. He points out that "the most frenzied buyers were regular Korean people — the proverbial taxi drivers and teenagers" who utilized newly available leveraged single-stock ETFs to borrow heavily against their portfolios. This mechanism created a self-reinforcing loop where "as more and more buy in and the stock goes up more and more, the perception of a structural upward trend is only reinforced."
"When prices fall, leveraged ETFs have to sell some of what they hold... All of this created extra selling pressure, and so increased the rate at which Korean stock prices fell since late June."
This financial feedback loop is the engine of the crash. Smith's analysis is particularly sharp here because it highlights how regulatory gaps allowed retail investors to borrow against volatile assets without understanding the risks. The administration of financial markets in South Korea is now scrambling to restrict these leveraged ETFs, a belated recognition that the system had become fragile. Critics might note that foreign institutional investors also played a role by taking profits early, but Smith rightly emphasizes that the domestic leverage was the accelerant that turned a correction into a freefall. The parallel to the 1873 railroad bust is apt: when the debt-fueled speculation meets a reality check, the collapse is violent.
The AI Revenue Gap
While the financial mechanics explain the speed of the crash, Smith pivots to a more unsettling question: was the valuation ever justified? He suggests that fears of an AI bubble are "quietly creeping back in," driven by a stark mismatch between capital expenditure and actual income. The narrative shifted in 2026 when tools like Claude Code seemed to prove AI had found "product-market fit," leading to a belief that data center construction was a safe bet. Yet, Smith warns that "writing software is different from selling it."
The core of Smith's argument rests on a terrifying calculation regarding the return on investment for Big Tech. Citing The Economist, he notes that "covering AI capex through identifiable AI income requires revenue on the order of $2.5trn per year, more than tech's entire combined revenue today." Current AI revenues, estimated at roughly $150 billion annually, would need to grow by a factor of 17 to justify the current spending levels. This is not a minor discrepancy; it is a chasm that suggests the market is pricing in a future that may not materialize.
"No one knows what the return on investment is."
This quote, attributed to Ken Mahoney of Mahoney Asset Management, encapsulates the growing anxiety on Wall Street. Smith uses this to argue that the "moat" protecting companies like Anthropic and OpenAI is less secure than previously thought. He introduces a new variable: the rise of Chinese competitors like Moonshot AI, whose models are nearly equal to American counterparts but likely backed by state subsidies. If American firms cannot maintain a pricing premium, their ability to service the debt and capital costs of their data centers evaporates.
"If Anthropic and OpenAI lose the overall b2b market to cheap Chinese competition... there's not much chance that they'll be able to pay for the data center boom."
This geopolitical angle adds a layer of complexity often missing from bubble theories. It suggests the crash isn't just about over-optimism, but about a potential structural failure in the global AI value chain. Smith's framing is effective because it moves beyond the "AI is a bubble" cliché to ask why the bubble might burst: the math simply doesn't add up unless revenue growth accelerates exponentially, which is far from guaranteed.
Global Contagion and the Bottom Line
The implications of the South Korean crash extend far beyond the KOSPI index. Smith points out that the selloff has spread to the "Magnificent 7," erasing $2 trillion in market value as investors grow increasingly cautious about the "hundreds of billions of dollars Big Tech is spending." The crash in South Korea, driven by domestic leverage, may have simply been the first domino to fall in a broader reassessment of AI valuations.
"So although South Korea's epic stock crash was probably related to Korea-specific financial factors, it could also herald the return of the 'AI bubble' story."
This is the piece's most significant takeaway: the local financial excesses in Korea acted as a stress test for a global narrative that is already showing cracks. The volatility in memory stocks, which are critical to the AI supply chain, serves as a warning signal for the entire sector. Smith's analysis is compelling because it refuses to treat the crash as an isolated event, instead weaving it into a larger tapestry of global capital flows and technological uncertainty.
Critics might argue that Smith is too pessimistic about the timeline for AI monetization, noting that early infrastructure investments often precede revenue by years, as seen in the internet boom of the late 1990s. However, the sheer scale of current spending relative to current revenue makes this a unique and precarious moment. The "tokenmaxxing" phenomenon, where companies use AI tools without generating real value, further complicates the picture, suggesting that demand may be artificial rather than organic.
Bottom Line
Noah Smith's strongest contribution is his dissection of the "revenue gap," forcing readers to confront the possibility that the AI boom is being funded by a level of speculation that cannot be sustained by current earnings. The argument's biggest vulnerability is its reliance on the assumption that Chinese competition will successfully erode Western pricing power, a dynamic that remains uncertain. For the investor, the lesson is clear: the era of blind faith in exponential AI growth is over, and the market is now demanding hard proof of profitability before the next leg up.