AI Is Eating Treasury Demand, Patterson Tells Bloomberg

Rebecca Patterson, former Chief Investment Strategist at Bridgewater and a Senior Fellow at the Council on Foreign Relations, said on Bloomberg's 'The Close' on Sept. 25, 2026 that artificial intelligence is crowding out demand for US Treasuries, as the S&P 500 proxy SPY closed at $771.35, up 0.54%.

Rebecca Patterson, the former Chief Investment Strategist at Bridgewater and now a Senior Fellow at the Council on Foreign Relations, made a claim on Bloomberg's "The Close" on Friday that cuts against the way most investors read the bond market: artificial intelligence, she said, is crowding out demand for US Treasuries.

That is a different argument from the one that usually gets made when yields rise. The familiar explanations are fiscal — too much issuance chasing too few buyers — or macro, with inflation expectations refusing to settle. Patterson's version is a competition-for-capital story. If the private sector is absorbing an enormous quantity of savings to build data centers, power them and stock them with chips, the marginal dollar that once rolled quietly into a Treasury auction has somewhere else to be.

What the crowding-out claim actually asserts

Crowding out is a textbook term with a specific mechanism behind it. In its classic form, government borrowing absorbs savings and pushes up the cost of capital for everyone else, squeezing private investment. Patterson is describing the mirror image: private capital demand large enough to move the price of government money.

For that to hold, the AI buildout has to be financed on a scale that registers in national savings, not just on individual balance sheets. The channels are visible even without putting a number on them. Hyperscalers and their suppliers have been issuing corporate debt to fund capital expenditure. Private credit funds have stepped into data-center and power projects. Equity allocations have tilted toward the same theme. Each of those uses a dollar that could otherwise have sat in the front or the belly of the Treasury curve.

The test is not whether yields are high. It is whether the composition of demand at auction is shifting — whether the buyers who show up are increasingly price-sensitive, and whether corporate spreads behave as though issuers are competing for a finite pool rather than tapping an infinite one. Those are checkable things, and they are the right place to look before accepting or dismissing the thesis.

Four markets, one story or several

Patterson's remarks, delivered to Romaine Bostick, sat inside a wider question the segment was built around: whether bonds, equities, commodities and energy are telling the same economic story, or whether one of them is carrying more information than the rest. You can watch the discussion via Bloomberg Technology.

That framing matters because the crowding-out thesis, if correct, breaks the usual translation. Investors are trained to read higher long-term yields as a signal about growth or inflation. A supply-and-demand imbalance driven by private capital formation is neither. It would mean the bond market is transmitting information about where money is going, not about how fast the economy is running or how hot prices are — and that reading a recession or an inflation scare out of the curve would be a category error.

Equities, in that world, are not contradicting bonds when both rise. They are describing the same phenomenon from opposite sides: capital flowing into productive assets, and the price of the safe asset adjusting to reflect its absence.

How the tape closed on the day she spoke

Friday's session gave no sign of stress. The SPDR S&P 500 ETF Trust (NYSEARCA: SPY) finished at $771.35, up 0.54% from the prior close of $767.18, having traded between $766.29 and $772.28. The Invesco QQQ Trust (NASDAQ: QQQ), the Nasdaq 100 proxy and the most direct equity expression of the AI trade, closed at $744.50, a gain of 0.46% against $741.10, inside a $739.64–$745.91 range. The SPDR Dow Jones Industrial Average ETF Trust (NYSEARCA: DIA) led, closing at $517.49, up 0.94% from $512.68. All figures are as of the last trade at 20:00 GMT on Sept. 25, 2026.

The notable detail is that the Dow proxy outpaced the Nasdaq 100 proxy on the day. That is one session and proves nothing on its own, but it is the kind of rotation worth tracking if the crowding-out thesis has legs: broad industrial and financial exposure keeping pace with, or beating, the concentrated technology complex that is doing the borrowing.

What would confirm it, and what would kill it

Three things are worth watching over coming auctions and quarters.

• Auction internals. Bid-to-cover ratios and the split between direct bidders, indirect bidders and primary dealers show who is actually absorbing supply. A persistent rise in dealer takedown would suggest genuine buyer scarcity rather than a pricing quirk.
• AI-linked corporate issuance. The size and frequency of investment-grade deals funding data centers, chips and power is the clearest proxy for private capital demand. If that calendar stays heavy while Treasury demand thins, Patterson's mechanism has evidence behind it.
• Spread behavior. Corporate spreads that widen alongside rising Treasury yields point to a scarcity of capital overall. Spreads that tighten while yields rise point to preference — investors choosing AI credit over government credit — which is the more specific version of her claim.

The thesis fails if Treasury demand holds steady while yields move on inflation prints and fiscal headlines, or if AI capital expenditure turns out to be funded largely from operating cash flow rather than from new borrowing. Cash-funded capex does not compete for savings in the same way; it recycles profits that were already inside the corporate sector.

Why the distinction matters for portfolios

For anyone holding duration, the difference between the two explanations is not academic. If yields are elevated because of inflation risk, they fall when inflation cools and bonds do their traditional job in a downturn. If yields are elevated because a generational capital buildout is absorbing savings, they stay elevated until that buildout slows — and the hedging properties investors expect from Treasuries during an equity drawdown become less reliable, because the same force is setting both prices.

Patterson did not put a number on the effect, and the honest position is that the evidence is still being assembled. But the question she raised — whether the bond market is pricing an economy or pricing a competition for capital — is the one that determines what every other market is saying.

Key facts

• SPY close: $771.35, +0.54%, as of 20:00 GMT Sept. 25, 2026
• QQQ close: $744.50, +0.46%, as of 20:00 GMT Sept. 25, 2026
• DIA close: $517.49, +0.94%, as of 20:00 GMT Sept. 25, 2026
• Speaker: Rebecca Patterson, ex-Bridgewater CIS, CFR Senior Fellow

Frequently asked questions

What did Rebecca Patterson say about AI and Treasuries?

Speaking with Romaine Bostick on Bloomberg's 'The Close' on Sept. 25, 2026, Patterson said artificial intelligence is crowding out demand for US Treasuries. The argument is that private capital being absorbed by the AI buildout reduces the pool of savings available to buy government debt, affecting yields through supply and demand rather than through inflation or growth expectations.

What does 'crowding out' mean in bond markets?

Crowding out traditionally describes government borrowing absorbing so much of a country's savings that private investment becomes more expensive. Patterson is describing the reverse: private capital demand — for data centers, chips and power — large enough to compete with the Treasury for the marginal investor dollar, which would put upward pressure on government borrowing costs.

Who is Rebecca Patterson?

Rebecca Patterson is the former Chief Investment Strategist at Bridgewater, one of the largest hedge fund firms in the world, and is currently a Senior Fellow at the Council on Foreign Relations. She appeared on Bloomberg's 'The Close' with Romaine Bostick on Sept. 25, 2026 to discuss her broad read on bonds, equities, commodities and energy.

How did US equity benchmarks close that day?

As of the last trade at 20:00 GMT on Sept. 25, 2026, the S&P 500 proxy SPY closed at $771.35, up 0.54% from $767.18. The Nasdaq 100 proxy QQQ closed at $744.50, up 0.46% from $741.10. The Dow proxy DIA closed at $517.49, up 0.94% from $512.68, leading the three on the day.

Why does it matter whether yields reflect inflation or capital competition?

The distinction changes how bonds behave in a downturn. If yields are high on inflation risk, they typically fall as inflation cools, restoring Treasuries' role as an equity hedge. If yields are high because a capital buildout is absorbing savings, they may stay elevated until that spending slows, weakening the diversification investors expect.

What evidence would test the crowding-out thesis?

Three markers: Treasury auction internals, including bid-to-cover ratios and the share taken by primary dealers versus direct and indirect bidders; the volume of investment-grade corporate issuance funding AI infrastructure; and whether corporate credit spreads tighten or widen as Treasury yields move. Capex funded from operating cash flow rather than new debt would undercut the argument.

Sources

• Patterson: AI Is Crowding Out Demand for US Treasuries — Bloomberg Technology