JPMorgan Strategy Chairman: U.S. Dollar Hegemony Remains on Solid Ground, but AI Concentration Risk Warrants Caution
Claire Weston
JPMorgan's Michael Cembalest published a 45-page report concluding that US dominance over global capital markets remains structurally intact, but classic bubble signals have emerged in AI, compounded by erosion in rule-of-law predictability and science policy.
Has the dollar's reserve status actually been shaken?
The report checked the dollar's share across cross-border lending, international debt, FX trading, central-bank reserves, trade invoicing, and SWIFT payments — virtually unchanged, some even ticking up.
China lacks the conditions to replace the dollar: its capital account is closed and its money-supply structure does not match. This means → no credible substitute is waiting in line.
Gold's rising share in reserves mostly reflects higher gold prices, not active dollar dumping by central banks.
Are foreign investors actually pulling out of US assets?
Foreign holdings of US Treasuries, corporate bonds, and equities are still growing in dollar terms — just slower than US debt itself is expanding.
In plain terms = foreign capital hasn't left; America is simply borrowing faster, so the "share" looks like it shrank.
The early-2025 "Sell America" trade — triggered by some Trump executive orders, with US stocks, the dollar, and Treasuries all falling together — has already faded. The dollar stabilized and US equities resumed outperforming global peers.
Why do US companies keep outperforming over the long run?
The report found US firms earn higher return on assets and return on equity than European, Japanese, and Chinese peers across nearly every major sector — not just tech.
This means → the long-term valuation discount on non-US equities is not sentiment bias; it is backed by a fundamental gap.
Over almost any long-cycle window since the 1980s, US stocks have delivered significant excess returns relative to the rest of the world.
How hot does AI have to get before it counts as a bubble?
The report lists classic bubble signals: semiconductor technicals extremely overbought, hedge-fund exposure surging, leveraged ETFs amplifying swings, and market breadth narrowing.
The top ten S&P 500 constituents now account for roughly 40% of total market cap, but the report argues the concentration risk *within* AI deserves more concern.
Specifics: Nvidia still dominates AI accelerator chips; frontier labs like OpenAI and Anthropic are burning cash at extreme speed; hyperscaler capex runs into the hundreds of billions while free-cash-flow margins are thinning; low-cost Chinese open-source models are gaining share; and labs are raising token prices rapidly to cover costs.
Why is TSMC dependence a structural weak point?
The US ranks first on virtually every global AI-readiness and activity index. China is close behind and closing the gap in both chips and models.
Yet the entire global supply chain — including US tech giants — still relies heavily on TSMC in Taiwan for the most advanced-node chips (advanced node refers to manufacturing processes at 7 nm and below).
In plain terms = America designs the AI "brain," but the factory that builds it sits in one place — Taiwan. That is the single most fragile link in the chain.
What is the unresolved question the report leaves behind?
The report confirms that AI-related productivity gains have partially offset the drag other policies put on growth. Productivity in information and data processing rose notably after ChatGPT launched in late 2022.
But it also flags two policy risks: declining rule-of-law predictability, and government cuts to science funding paired with sidelining of scientific expertise.
This reflects a concern that sits deeper than economic data — it is institutional. Can America's other structural advantages keep offsetting the erosion in legal predictability and science policy over time? That is the open question the report leaves for investors.
Content is for reference only, not financial advice.