Societe Generale's top short seller: The AI boom is replicating the Asian financial crisis; the fatal flaw is the debt "time bomb."

Societe Generale's chief strategist, Albert Edwards, warned that the current AI investment boom bears a striking resemblance to the 1997 Asian financial crisis— not because the technology itself is useless, but because capital inflows are far outpacing productivity improvements, and historically, it has always been creditors, not engineers, who have filled this gap.

In his latest Global Strategy Weekly, Edwards points out that total factor productivity (TFP) data, which measures the actual benefits of technology, has shown no improvement so far, but global AI capital expenditure has skyrocketed. Goldman Sachs predicts that global AI investment will exceed $1 trillion in 2026 alone. Meanwhile, OpenAI's annualized revenue was disclosed to be about $20 billion lower than previously expected . Upon this news, the Nasdaq plunged more than 1% that day, directly impacting the narrative of "strong demand."

In Edwards' view, the real trigger for this crisis is whether the cheap money supporting AI expansion can continue to be supplied; disappointing productivity data is not the main reason. The bond market's "vigilante" is testing the vulnerabilities of markets around the world—from Japan to France, bond sell-offs are spreading rapidly. As AI tech giants raise massive amounts of debt, they precisely need the same group of duration buyers to absorb hundreds of billions of dollars in new bond issuances, a logic strikingly similar to that of Asia's reliance on short-term dollar borrowing to maintain prosperity in the 1990s.

TFP data "malfunctions": AI boom lacks productivity support

This round of warnings began with a rather dry chart. Apollo's chief economist, Torsten Slok, released a report titled "No Signs of AI in Productivity Data." The data shows that TFP, adjusted for capacity utilization, is currently slightly below zero, showing no signs of acceleration since the start of the AI capital expenditure cycle, while hourly output growth has remained stable at around 2.5%.

Slok points out that the strong hourly output growth coupled with stagnant TFP is a typical characteristic of "capital deepening," rather than a signal of technological shock—giving every employee a new monitor (or a $40,000 GPU) will increase output per hour, but it won't make the company truly more efficient. The AI boom is "clearly visible in investment data and stock valuations, but not yet reflected in productivity statistics," meaning that the so-called returns "are still predictions, not facts."

Bank of America strategist Michael Hartnett also noted this pattern, adding a noteworthy detail: TFP and the consumer confidence index have been highly correlated over the past half-century, and both are currently declining. Chicago Fed President Austan Goolsbee has also previously warned that persistent productivity stagnation will challenge the current narrative.

Asian Mirror: Edwards' Classic Reverse Bet and Its Historical Echoes

This isn't the first time Edwards has expressed skepticism about the "miracle narrative." He recalls that his "first major, highly risky, non-consensus judgment" in his career was that the "East Asian economic miracle" was essentially a massive economic and financial bubble. His theoretical basis comes from economist Paul Krugman's 1994 article in *Foreign Affairs*, "The Myth of the Asian Miracle"—Krugman argued that the high growth of Asian economies stemmed from the massive accumulation of capital and labor, rather than genuine improvements in efficiency, and that weak TFP growth fundamentally undermined the bullish narrative.

At the time, the mainstream view dismissed this idea. The World Bank published "The East Asian Miracle" in 1993, and the "peak of arrogance and optimism" occurred in August 1996—another World Bank report praised the "Thailand macroeconomic miracle," less than a year before the baht collapsed.

Edwards jokingly referred to his bearish framework from back then as "Noddynomics," which was met with widespread ridicule from the market. During a roadshow in the United States, he compared the US tech bubble of the late 1990s to the Thai economic bubble, even requiring his then-boss to intervene to protect him from angry clients. Both bubbles eventually burst.

Depreciation Black Hole: Net investment stagnates, while capital keeps piling up.

Edwards' second accusation comes from research by his former colleague, Rob Parenteau. Parenteau points out that while total business investment driven by AI is growing rapidly, net business investment is almost flat after deducting depreciation : in nominal terms, total investment accounts for about 14% of GDP, while net investment has stagnated at around 3% for about a decade; in real terms, total investment reached a record high of about 15.5% of GDP, but the real net value is only about 3.5%, roughly the same as in 2015 and 2019.

Edwards also points out that companies are extending the depreciation period for assets such as GPUs, which makes net investment data look better from an accounting perspective, but masks weak real economic returns. This echoes Michael Burry's previous criticism of the stretched assumptions about the lifespan of GPUs.

Goldman Sachs' calculations further quantify this risk. According to its latest research, if the six largest U.S. hyperscale cloud service providers have a zero return on investment (ROIC) on AI capital expenditures, they would need approximately $920 billion in revenue annually to cover depreciation and operating costs alone.

Goldman Sachs estimates AI capital expenditure in three phases: approximately $633 billion from 2023 to 2025, approximately $1.73 trillion from 2026 to 2027, and a staggering $4.14 trillion from 2028 to 2030. To achieve a 15% ROIC in just the second phase of capital expenditure, the six largest hyperscale vendors would need to generate approximately $1.42 trillion in revenue between 2028 and 2030.

GDP data doesn't match: The AI boom is more reflected in prices than output.

GDP data also failed to provide much support for the bulls.

Edwards points out that U.S. corporate fixed investment contributes less than one percentage point to GDP growth, a fraction of the peak level in the late 1990s (which contributed more than two percentage points).

Structurally, while equipment investment saw a nominal growth rate of nearly 14%, the real growth rate "barely reached double digits"; non-residential building expenditures contracted, with a nominal decline of about 3% and a real decline of nearly 6%—even including data center construction.

Edwards points out that the large amount of "AI boom" in GDP figures reflects price increases rather than production expansion. The root cause is that all participants are simultaneously scrambling to buy the same batch of chips, memory, and transformers, driving up prices. This also explains, to some extent, why the Federal Reserve is concerned about the inflationary risks brought about by AI, rather than the deflationary effects.

Debt "time bomb": A crisis will be triggered once cheap capital withdraws.

In his report, Edwards points out that the real root cause of the Asian crisis is not disappointing productivity data, but the sudden halt in the inflow of cheap external capital that underpinned resource misallocation. He warns that the bond market's "vigilante" is operating in the same vein: recently, the repatriation of Japanese funds first drove a sustained sell-off of US Treasuries, which then spread to France, where French government bonds are now heading for their worst decade of performance since 1803.

Against this backdrop, AI technology companies are still issuing bonds on a large scale—Broadcom, Oracle, SpaceX and others have joined the AI chip financing wave, further increasing the supply pressure in the bond market, which directly competes with the demand of ultra-large-scale manufacturers for long-term capital buyers.

Edwards' conclusion hits the nail on the head: Thailand's miracle didn't die from disappointing productivity, but rather came to an abrupt end the moment creditors realized it. The biggest risk facing the current AI boom is a repeat of the same script—only the financing tools have changed from short-term dollar loans to investment-grade bonds, private credit, and special purpose vehicles.

Edwards cautiously stated that he was not declaring the AI craze dead, but rather questioning: why is the market consensus so certain when the only statistical evidence proving this is not a bubble refuses to cooperate? The answer to this question may only become truly clear after the capital tide recedes.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.