The Saving Constraint

For much of the past two years, the discussion of artificial intelligence and inflation has favored the disinflationary interpretation. Lower unit costs, a displaced workforce, and narrower margins in commoditized services all point toward softer prices, consistent with the pattern established by earlier general-purpose technologies. The more consequential development is not that this interpretation is being questioned in general terms, but that one of its most credible advocates has changed its position.

Hoisington Investment Management, whose long-standing conviction in long-duration Treasuries is well known to institutional investors, used its second-quarter 2026 client letter to argue that the structural environment for U.S. inflation is shifting higher. In the firm’s assessment, the long-run equilibrium range is migrating from approximately 1.5% to 3.5% toward 3.5% to 4.5%, with a material probability of episodes above 5%. Consistent with that assessment, and according to figures reported by Bloomberg, the effective duration of the firm’s bond portfolio has fallen from nearly twenty-one years last September to less than one year as of June. A manager positioned for lower yields across four decades has moved decisively in the opposite direction.

The reasoning behind that shift does not depend on the first-order economics of artificial intelligence. It depends on the relationship between national saving and national investment.

The United States is moving into a period in which its appetite for capital is expanding much faster than its domestic capacity to save. Net national saving has declined from a long-term average of roughly 6.8% of national income toward historic lows, averaging approximately 0.6% over the last four quarters. On the Bureau of Economic Analysis series measured against gross national income, a broader denominator that reads correspondingly lower, net saving registered 0.1% in 2024 and 0.5% in 2025, against readings of 7.1% and 8.0% in 1980 and 1981 respectively. The measure turned negative for the first time in the postwar period in 2009. The personal saving rate has fallen from above 10% in the 1970s to roughly 3% at present. That decline reflects both the growth of federal budget deficits and the reduction in household saving, and it constrains the private sector’s capacity to finance investment.

This shortfall arrives precisely as the economy confronts exceptional investment requirements, including artificial intelligence and its associated data-center footprint, the expansion and hardening of the electricity grid, semiconductor and energy infrastructure, reindustrialization, defense modernization, and space exploration. Because investment must ultimately be funded either from domestic saving or from foreign capital, a persistently low saving rate points to some combination of greater reliance on foreign financing, higher real interest rates, and the displacement of productive private investment.

A distinction between monetary conditions and real resources is central here. Monetary expansion can supply additional liquidity, but it cannot conjure the previously saved real resources on which capital formation depends. Financing a surge in capital demand through money creation would therefore be more likely to add to inflation and to misdirect resources than to close the underlying capital gap, and the economy would be left contending with both faster price growth and a persistent drag on its long-run growth potential.

The scale of the accumulated imbalance is visible in the debt data. On figures drawn from the Federal Reserve’s financial accounts, domestic nonfinancial debt spanning the public and private sectors has increased from roughly $4 trillion in the early 1980s to more than $80 trillion in nominal terms, and relative to nominal GDP that debt has almost doubled, from about 1.4 times to 2.6 times. An economy that saves progressively less has sustained investment and consumption through a rising reliance on credit, an arrangement that becomes considerably more difficult once the saving shortfall becomes binding and the price of capital is set by scarcity rather than by abundance.

As the saving shortfall becomes binding, the risk of crowding out rises. According to CBO estimates cited by the Peter G. Peterson Foundation, over the long run each additional dollar of federal deficit reduces private investment by approximately 33 cents. For much of the last cycle this effect was muted, because ample foreign demand absorbed Treasury issuance without displacing private borrowing.

That source of demand is now weakening. Foreign and international investors held 30.0% of federal debt held by the public at the end of 2025, down from a peak of 49.0% in 2008 and 47.2% as recently as the end of 2010, based on U.S. Treasury data. While the absolute dollar amount of foreign holdings has continued to rise, from $4.4 trillion at the end of 2010 to $9.3 trillion at the end of 2025, it has grown far more slowly than the debt itself, which increased from $9.4 trillion to $30.9 trillion over the same period. Foreign demand has therefore absorbed a steadily smaller proportion of each year’s issuance.

The composition of that demand matters as much as its level. Foreign official institutions, principally central banks, have historically been price-insensitive buyers, acquiring Treasury securities for reserve-management purposes rather than for return. Their retreat leaves a larger share of issuance to be absorbed by private and domestic investors, who require compensation for duration and inflation risk. As the marginal buyer shifts in that direction, a higher term premium becomes more probable, and with it a higher cost of capital across the economy.

The challenge of financing wider federal deficits while simultaneously absorbing hundreds of billions of dollars of capital to build data centers is therefore intensifying. Without renewed inflows of foreign capital, higher borrowing costs are the likely result.

A further consideration concerns the effect of labor-force displacement on federal finances, and it runs counter to intuition. When layoffs accelerate, the initial impact tends to be deflationary, as observed in prior recessions. At the same time, as layoffs increase, government deficits tend to widen, because tax revenues fall short while government assistance rises.

The significance of that dynamic lies in the composition of federal revenue. Bloomberg reported in July 2026 that labor income and payroll taxes together account for more than 65% of federal revenue. Analysis of 2024 government data indicates that payroll taxes represent approximately 37% of federal revenue and labor income a further 28%, with corporate taxes, capital income, transfer income, and other sources making up the remainder. The federal revenue base is, in effect, a tax on human labor. If displaced workers are compelled to accept lower-paying roles, or are pushed out of the labor force altogether, the effect on federal revenue could be pronounced. The same displacement that shifts income from labor toward capital would shift it toward the more lightly taxed of the two.

This dynamic is compounded by the trajectory of interest costs. In contrast to the period from 1990 to 2020, during which net interest relative to GDP was declining and a wider deficit therefore did not compound, net interest is now at a record high and rising. Federal interest outlays reached 3.15% of GDP in 2025, effectively matching the postwar high of 3.16% set in 1991, and on the Congressional Budget Office’s February 2026 projections they reach 3.3% of GDP in 2026, exceeding that record, before climbing to 4.6% of GDP by 2036. A significant labor-market dislocation would intensify that trend, as falling revenues coincided with higher spending on unemployment benefits, retraining, and related services. The combination of declining revenues, rising costs, and prolonged uncertainty represents a material source of fiscal and financial instability.

Higher deficits also translate into higher borrowing costs for consumers, businesses, and governments, contributing to cost-push inflation. The Budget Lab at Yale estimates that a permanent increase in the primary deficit equal to 1% of GDP, roughly in line with the cost of fully extending the individual provisions of the Tax Cuts and Jobs Act, raises inflationary pressure after five years equivalent to a reduction in household purchasing power of $300 to $1,250 in 2024 dollars, and increases mortgage-interest payments by $600 to $1,240 a year in the current housing market. Expressed in terms of borrowing costs, the same increase adds roughly $60 to annual interest on a typical auto loan, $600 to interest on a median-priced home, and $1,000 to interest on a typical small-business loan after five years. Over a thirty-year horizon, those figures rise to approximately $200, $2,300, and $3,400 respectively.

The analysis above does not constitute investment advice, and readers should treat it as scenario analysis rather than a recommendation. It does, however, frame the allocation question.

If these dynamics develop as described, the environment that rewarded long-duration fixed income over the past four decades is likely to reverse. Understanding these interacting factors will be central to asset-allocation decisions. On the trajectory outlined here, the bull market in commodities and hard assets would accelerate, while fixed income would increasingly be avoided. The repositioning of Hoisington’s portfolio, from twenty-one years of duration to less than one, illustrates that conclusion in a credible setting. Within equities, valuation would tend to favor businesses that generate cash in the near term and retain the ability to reprice their inputs, over those whose value rests principally on cash flows in the distant future.

The firm summarized the structural case in the following terms. Absent a sustained recession, a favorable supply-side shock, or a prolonged period of monetary restraint, a backdrop of larger structural deficits, heavier capital demands, more fragmented supply chains, diminished gains from globalization, and greater sensitivity to the supply of Treasuries points to both inflation and long-term Treasury yields that, in the firm’s words, “trend upward.”

The debate between the inflationary and deflationary consequences of artificial intelligence is unlikely to be settled quickly. The near-term effect of accelerating layoffs is genuinely disinflationary, and it may prevail for a period. The structural forces described here, however, operate through the accounting relationship between saving and investment rather than through a forecast, which makes them more difficult to offset. The indicators most worth monitoring are the path of net national saving, the term premium, foreign demand for Treasury securities, and the primary deficit. Together they will indicate the direction in which the constraint is resolving.