The market has effectively become the economy it once merely served. Barry Gill argues that a supposedly diversified portfolio may now simply be a single leveraged AI bet.
Circular financing has become the most fashionable worry in the AI trade. Chip makers invest in the model builders that buy their chips, vendors book as revenue what they have in effect lent, and diagrams of the interconnected web have been doing the rounds for well over a year.1 However, I believe this lens is far too narrow.
The circularity that should concern a long-term investor is not a handful of deals inside the technology complex. It is, instead, a loop that runs through the whole economy and, quite probably, through most portfolios. Let me first lay down some facts.
Start with US households. In aggregate they hold equities worth more than three times their annual disposable income, against 1.9 times at the top of the dot-com boom and a long-run average of about one. Equities make up almost half of household financial assets, and about 70 percent of that is held directly rather than through stable intermediaries like pension plans, mutual funds or insurers.2
Wealth distribution amplifies everything. By Moody’s estimate, the top 10 percent of households by income now account for close to half of all US consumer spending and the top 20 percent for about 60 percent,3 meaning that, in effect, the confidence of the many has come to rest on the portfolios of the few.
Moreover, the earnings underpinning those portfolios have themselves been flattered. The way reported profit is measured has changed four times in a generation, and the changes have tended to lift earnings per share without adding a dollar of cash. S&P 500 earnings grew 52 percent in the second quarter of 2026, the fastest since 2021. More than a third of that growth came from two companies, Alphabet and Amazon, which booked about USD 150 billion of gains on equity investments between them.4
Set against all of this, compensation for owning this risk is thin. Equity market capitalization stands at about 2.5 times GDP, against 1.6 times at the dot-com peak5, the equity risk premium is close to a 23-year low6 and profit margins, which are currently at record highs, are what Jeremy Grantham famously called probably the most mean-reverting series in finance.7
Fragility and reflexivity
Fragility and reflexivity
All economic roads currently lead back to the market. Fiscal receipts are a good example. Capital gains have supplied anywhere from 5 percent to more than 14 percent of individual income tax revenue since 2010, and they rise and fall with the market far more sharply than the rest of the tax base. Realized gains swung from almost 9 percent of GDP in 2021 to 3.4 percent in 2023.8 All this comes at a moment when federal net interest already exceeds USD 1 trillion a year and USD 10 trillion of debt is being refinanced in 2026.9 Monetary policy is equally constrained by the market’s reaction to it.
The AI build-out is now visible in the national accounts. Over the past year, spending on computers, software and R&D has added about one percentage point to annualized GDP growth, more than the same categories contributed at the height of the dot-com boom. Much of the hardware is imported, so the lift to domestic output is smaller.9
Increasingly, it is funded with debt. Hyperscalers issued more than USD 125 billion in 2025 and almost twice that by early August 2026, and they account for only about 40 percent of the nearly USD 500 billion issued across the wider AI complex this year.10 Or put another way, the expansion is drawing strength from a single capital cycle.
So, you can begin to see how tightly all this connects. Stock market increases lift household wealth, household wealth funds consumption, consumption sustains earnings, earnings justify the market, with the state and the central bank both depending on all this holding together. Capital spending works much the same way in that a hyperscaler’s outlay is a chip maker’s profit today and only counts as a cost over the years it chooses to depreciate it.
Finance has a name for such loops. The investor George Soros called it reflexivity, by which he meant the tendency of prices to feed back into the fundamentals they are meant to reflect. He set the idea out in 1987, although he traced it to his study under the philosopher Karl Popper decades earlier.11
A loop like this only holds while the crowd engaged in it still disagrees. A market made up of investors weighing the evidence differently produces a price that carries useful information. However, a market in which everyone has converged on the same story produces a price that is close to useless. Narrative dominance drives out disagreement, spreading conviction from investor to investor like a virus.13
Nowhere to hide
Nowhere to hide
A major lesson of 2026 is surely the limits of diversification in the face of an all-consuming narrative, one that forces us to re-examine traditional sector, factor and asset-class labels.
Consider a conventional portfolio. US large-cap equity, where the build-out has to all intents and purposes become the index: Global equity outside the US, whose largest names (TSMC, Samsung, SK Hynix, ASML14) are the semiconductor supply chain the build-out depends upon investment-grade credit, carrying record hyperscaler issuance; private credit, lending against data centers and the chips inside them; infrastructure, meaning power generation and grid connection; and Real estate, in which the data center is now the growth asset.
At the same time, the old defenses have wobbled. The once-reliable negative correlation between stocks and bonds that underwrote the balanced portfolio is gone. In last year’s “Sell America” episode, the dollar broke with equities and bonds together, withdrawing the hedge just when it was most needed. Equally, index concentration has turned the passive allocation into an active one.
Borrowing from ecology, Andrew Haldane called this a robust yet fragile system, sturdy against small shocks and apt to fail all at once, exactly as a monoculture shrugs off ordinary weather until the one blight it cannot survive takes hold. It is also what Andrew Lo meant when he said a market’s efficiency is ecological – present when the species trading it are varied and absent when they collapse onto a single strategy.15
What to do?
What to do?
I am not advocating stuffing cash under the mattress and locking the door. Far from it. What investors need now is more second-order thinking, and diversification by causal pathway rather than by asset-class label, deliberately owning things that need different stories to be true.
Lower correlations, where they can be found, improve risk-adjusted returns rather than merely dampening volatility. Europe could be a genuine structural diversifier and not merely a value trade. Alternatives can contribute too, although clear specialist knowledge and experience are required. And the typical safe havens – government bonds and gold – arguably still deserve more attention than investors tend to give them.
I cannot foresee which weak spot cracks open, or where the loop breaks down. The achievable aim should be insurance – a portfolio not wholly reliant on any single narrative.
Which returns us to the original worry. The circle I see is one that joins the market, consumer, earnings, receipts and policy response into a single reflexive system. Investors cannot step outside a loop the whole economy is standing in. We can, however, decline to own the same bet six times over, and ask what is left standing when the story finally reverses.
1 “The Bull Market’s Seventh Inning,” Morgan Stanley Wealth Management (Lisa Shalett), 1 October 2025.
2 Financial Accounts of the United States (Z.1), Federal Reserve, 11 September 2026; disposable personal income, Bureau of Economic Analysis, 30 September 2026.
3 “Top 10% of Earners Drive a Growing Share of US Consumer Spending,” Bloomberg (Jonnelle Marte), 16 September 2025; “The richest 20% are the only ones powering the U.S. economy, says top economist, but their prospects are entirely reliant on teetering stock prices,” Fortune (Eleanor Pringle), 26 June 2026.
4 “Earnings Insight,” FactSet, 28 August 2026.
5 Financial Accounts of the United States (Z.1), Federal Reserve, 11 September 2026; gross domestic product, Bureau of Economic Analysis, 30 September 2026.
6 “Historically, stocks have offered a big premium over bonds. Suddenly, the difference has almost vanished,” Fortune (Shawn Tully), 29 September 2026.
7 “Earnings Insight,” FactSet, 28 August 2026; “Why Jeremy Grantham Is Right About Corporate Profit Margins,” Advisor Perspectives (Baijnath Ramraika and Prashant Trivedi), 11 November 2014.
8 CBO’s Projections of Realized Capital Gains, Congressional Budget Office, February 2023; “Historical US Federal Capital Gains Tax Rates & Collections, 1913-2025,” Tax Foundation, 15 August 2025.
9 Monthly Budget Review: August 2026, Congressional Budget Office, 9 September 2026; “A Third of US Debt Matures in 2026,” Real Investment Advice, 12 December 2025.
10 Table 1.5.2, Bureau of Economic Analysis, 30 September 2026; “Tracking AI's Contribution to GDP Growth,” Federal Reserve Bank of St. Louis (Hannah Rubinton and Bontu Ankit Patro), 12 January 2026.
11 “How AI Debt Is Reshaping Credit Markets,” Goldman Sachs Exchanges (Amanda Lynam and Zach Ablon), 5 August 2026.
12 The Alchemy of Finance, George Soros, 1987.
13 Narrative Economics, Robert Shiller, 2019.
14 MSCI ACWI ex USA IMI Index factsheet, MSCI, 30 September 2026.
15 “Rethinking the Financial Network,” Bank of England (Andrew Haldane), 28 April 2009; Adaptive Markets, Andrew Lo, 2017.
Code: S 10/26 M-007360
