Asset management giant: The market's expectations for AI productivity are overly optimistic.

One of the world's largest insurance and asset management groups, Allianz, has its chief economist warning that market expectations for AI-driven productivity gains have already shown signs of irrational exuberance, and that the actual impact of AI on the real economy will be far more complex and uneven than what market pricing reflects.

On July 3, according to Bloomberg, Allianz Chief Economist Ludovic Subran said at an annual economic conference in France that day, "We don't really know how the adoption of AI and its impact on the real economy will play out, but the market is already very optimistic, especially regarding productivity gains—and the reality will be a much more mixed picture. For me, that's exactly where I see a certain degree of irrational exuberance." He also expressed concerns about the overall "market psychology" surrounding AI investment.

Subran's remarks echoed those of International Monetary Fund (IMF) officials earlier this week and aligned with warnings from the Bank for International Settlements (BIS) last Sunday—which has listed AI as one of the four major "stress points" threatening global economic prosperity.

AI impact will be a "mixed picture," not the comprehensive dividend the market expects

Subran acknowledged that the changes heralded by AI will be revolutionary, calling it a "Renaissance moment," and pointed out that AI will deeply transform the service economy. However, he also emphasized that this technology has given rise to some "odd phenomena" in corporate and investor behavior.

His core judgment is that the impact of AI on various economies will not be evenly distributed. The current optimistic market expectations are built on the assumption of comprehensive and rapid productivity gains, but the actual situation will be a "mixed picture"—the extent to which different industries and companies benefit will vary significantly, which is clearly at odds with the current overall market pricing logic.

Subran pointed to the sharp expansion of capital expenditures in the U.S. AI sector. Citing judgments similar to those of IMF officials, he singled out some companies for falling into a "debt expansion cycle"—massive capital investment driving up debt, while the timeline and scale of return on investment remain highly uncertain.

A Wall Street Journal article wrote that Tobias Adrian, Director of the IMF's Monetary and Capital Markets Department, said that current valuations of AI-related stocks may not necessarily form a bubble, but what financial regulators should really be wary of is that global large tech companies are using more and more medium- and long-term debt financing to invest heavily in rapidly evolving AI infrastructure. This mismatch in asset and liability maturities is the potential source of future financial stability risks.

Prior to that, in an article on June 29, the Wall Street Journal also wrote that the Bank for International Settlements warned of three major threats: AI bubble burst, inflation, and sovereign debt. The report pointed out that the AI "circular financing" structure is opaque and involves risks of multiple asset pledges, and that a downturn could trigger a credit tsunami on the scale of 2008. Combined with rising risks of secondary effects of inflation and high leverage basis trades by hedge funds that could easily trigger a deleveraging fire sale, the vulnerability of the global financial system is increasing.

Subran specifically highlighted the polarization of corporate behavior: companies like Apple and Microsoft have "not done much" in the AI space, while others have "over-invested." In his view, this divergence itself is a sign of structural imbalance in the market.

"If you are issuing debt to reward shareholders, that doesn't look like a good omen to me," Subran said. He also expressed specific concerns about the potential risks of data centers, including the risk of technological obsolescence for some data centers and the operational logic of monetizing capital expenditures.

Subran observed a clear divergence in risk pricing between the stock market and the bond market regarding AI. In the bond market, he believes investors have remained relatively rational—"When you look at corporate credit spreads in this sector, especially for hyperscalers, they are more cautious than before," he said, adding that the bond market does not show complacency, and "there are still plenty of 'debt vigilantes' in the bond market."

However, the situation in the stock market is entirely different. "On the equity side, it seems the sky is the limit, and of course that's not the case," Subran said bluntly. This equity-bond divergence, in his view, is the most direct manifestation of irrational exuberance in the current AI investment frenzy—stock market pricing reflects the most optimistic scenario, while the caution in the bond market suggests that real constraints still exist.