Key Takeaways:
Allianz's chief economist says markets are pricing AI productivity gains that may take years to materialize, if they arrive at all.
Allianz Chief Economist Ludovic Subran warned that market expectations for AI-driven productivity gains have entered irrational exuberance territory, as the gap between stock market optimism and bond market caution widens. Speaking at an annual economic conference in France on July 3, Subran said the actual impact of AI on the real economy will be far more complex and uneven than current pricing reflects, calling the divergence between equity and credit markets the most visible sign of overheating.
"We do not really know the degree of AI adoption and what the impact on the real economy will be, but the market is already very optimistic, especially on productivity — and the reality will be a more mixed picture," Subran said. "For me, that is where I see a degree of irrational exuberance."
AI-related capital expenditures across major hyperscalers exceeded $360 billion in 2025, funded through a combination of operating cash flow and long-duration debt issuance, according to BlackRock Investment Institute's midyear outlook. Subran singled out the debt expansion cycle as a key risk, echoing warnings from the International Monetary Fund and the Bank for International Settlements. IMF Monetary and Capital Markets Director Tobias Adrian said last week that large technology companies are increasingly using medium- and long-term debt to finance rapidly depreciating AI infrastructure, creating an asset-liability mismatch that could become a source of financial stability risk. The BIS went further, warning that opaque "circular financing" structures in AI investment could trigger a credit shock comparable to 2008 if the cycle reverses.
The stock-bond divergence on AI risk
Subran pointed to a structural disconnect between how equity and credit markets are pricing AI exposure. In bond markets, investors have remained relatively disciplined — credit spreads on hyperscaler debt have tightened but not to levels that suggest complacency. "When you look at credit spreads on hyperscalers, they are more cautious than before," Subran said. "There are still plenty of bond vigilantes."
Stock markets tell a different story. "On the equity side, the sky seems to be the limit, and that is certainly not the case," Subran said. The divergence is the most concrete evidence of irrational exuberance in the AI trade: equities pricing the most optimistic scenario while debt markets price in real constraints on returns.
Subran also flagged uneven corporate behavior as a structural warning sign. Apple Inc. and Microsoft Corp. have been measured in their AI spending, he said, while other companies are "over-investing." He expressed specific concern about data center risks, including technological obsolescence and the difficulty of monetizing capital expenditure. "If you are issuing debt to reward shareholders, that is not a good sign in my view," Subran said.
What it means for investors
The warning comes as S&P 500 consensus earnings growth of 24% for 2026 reflects broad expectations for AI-driven productivity gains, according to J.P. Morgan research. But with valuations already elevated and significant hype priced into the most obvious AI beneficiaries, the risk of a retracement is real even if the fundamental capital expenditure trajectory remains intact. For investors seeking AI exposure without paying peak multiples, the infrastructure layer — power, cooling, connectivity, and data center real estate — trades at more modest valuations while still capturing exposure to the same spending cycle, though Subran's warning suggests even those bets carry execution risk if the productivity payoff proves slower than expected.
This article is for informational purposes only and does not constitute investment advice.
