A Year, Two Opposite Stories
TREE NEWS reports: In roughly twelve months, the dominant story Wall Street tells about the US economy has flipped almost exactly 180 degrees. A year ago, the consensus rested on two mutually reinforcing beliefs: that artificial intelligence would raise productivity and push inflation lower, and that the US Treasury could keep the bond market under control, with long-end yields staying anchored. Today, that framework has been abandoned. The prevailing view is that heavy AI-related corporate borrowing is itself a major force pushing Treasury yields higher, and that the Treasury — whatever tools it reaches for — has lost its grip on long-term rates.
The shift was documented by Deutsche Bank’s global head of FX research, George Saravelos, after a dense round of client meetings along the US East Coast. He also flagged a telling rumor circulating in those meetings: talk of pausing 20-year Treasury issuance and sharply shortening the overall maturity of supply. The mere existence of that discussion, he noted, reflects deep anxiety about long-end supply-demand imbalance.
Why This Matters for Markets
If the market now believes AI-driven issuance is a structural driver of higher yields, the implications ripple across every asset class.
- Bonds: A structurally steeper curve and higher term premium become the base case. Long-dated Treasuries lose their traditional role as a safe-haven ballast, which is why many investors are reportedly extremely cautious on fixed income.
- Equities: Higher long-end yields raise the discount rate applied to AI’s distant future cash flows. The same companies driving the AI capex boom become more sensitive to every uptick in yields.
- Currencies: A loss of confidence in US fiscal management tends to weigh on the dollar over time, even as higher yields attract capital in the short run — a tension that makes FX positioning unusually difficult.
- Commodities and crypto: Energy demand from data-center buildouts is a genuine new variable for power and fuel markets, while crypto assets remain a high-beta proxy on liquidity conditions and risk appetite.
The Pendulum Has Swung Too Far
Saravelos disagrees with the new consensus. In his view, the narrative pendulum has overshot, and the market’s pessimism on fixed income may be an overcorrection. But he is far more concerned about a different tail risk that he says is almost entirely unpriced: a breakdown in the AI ecosystem itself.
He points to a discussion with data-center industry specialists who raised three concerns — deteriorating underwriting standards, uncertainty over whether energy supply can meet construction demand, and doubt that the projected explosion in inference compute will find enough paying demand. One research paper cited in the report estimates that AI labs would need to generate close to $4 trillion in revenue to earn a return on data-center investment, a striking benchmark for the sustainability of the current boom.
The risk Saravelos identifies as most underpriced is not France’s fiscal troubles — a fresh source of euro pressure that clients overwhelmingly view with pessimism, drawing comparisons to the 2010-2015 euro crisis — but rather an AI accident, a failed IPO, or a disappointing revenue print. Given extreme concentration in AI names, such an event would hit the dollar hard and provide powerful support to bonds. Markets, he argues, are barely reflecting that possibility.
Key Takeaways for Investors
- The consensus has moved from ‘AI is disinflationary’ to ‘AI issuance is inflationary for yields’ — a full reversal in one year.
- Extreme caution on fixed income may be overdone, but the market is ignoring a genuine two-sided risk in AI.
- The most underpriced scenario is an AI ecosystem shock — safety incident, IPO failure, or revenue miss — which would weaken the dollar and strengthen bonds.
- Watch French fiscal developments as a new, unanticipated drag on the euro.
- Positioning should account for concentration risk in AI-linked assets, not just the direction of rates.
