AI’s $1 Trillion Bet Is Driving Up Real Interest Rates

The 10-year Treasury yield recently climbed to its highest level since November 2023. This shift is not driven by resurgent inflation or a loss of Federal Reserve control. Instead, it reflects the massive capital expenditures of Microsoft, Amazon, and Nvidia as they invest aggressively in AI infrastructure. Their spending is reshaping the bond market’s playbook by pushing real interest rates higher.

The scale of this spending is unprecedented. The Minneapolis Fed projects that AI-related capital expenditures will reach $1 trillion by 2027. This isn’t merely a corporate line item—it’s a macroeconomic force. By absorbing a significant portion of available capital, these investments are crowding out small businesses, municipalities, and other borrowers. The result is a structural rise in real interest rates, as demand for capital outstrips supply.

The Capital Call No One Can Ignore

Nvidia’s data centers and Microsoft’s cloud expansions are not funded by incremental adjustments. These companies are issuing debt, repatriating cash, and redirecting every available dollar toward AI infrastructure. The QQQ ETF, which tracks the Magnificent Seven, serves as the primary vehicle for this spending spree. As these firms deploy billions into AI, their capital demands are directly influencing macroeconomic conditions.

The dynamics are straightforward: when the largest spenders in the economy commit to AI, demand for capital surges. This isn’t a temporary spike—it’s a structural shift. Smaller borrowers are being squeezed out, and real interest rates are rising as a result.

For investors, the critical question is whether this shift is permanent. If AI delivers on its productivity promises, the long-term disinflationary effects could offset today’s higher rates. If not, the bond market may face a lasting repricing.

Productivity: The $1 Trillion Gamble

David Zervos’s thesis hinges on a single assumption: that AI-driven capital expenditure will translate into broad-based productivity gains. So far, the evidence is inconclusive. Federal Reserve researchers note that while GDP growth remains strong, employment growth has been tepid. This dynamic suggests rising productivity, but the data lacks clear signs of a transformative shift in aggregate output or labor markets.

Projections from the National Bureau of Economic Research (NBER) estimate Total Factor Productivity (TFP) gains of less than 0.66% over the next decade. This forecast falls far short of a revolution—it’s a marginal improvement. The Minneapolis Fed reinforces this view, stating that AI is currently heating up the economy rather than cooling it. While capital is being deployed, the productivity payoff remains elusive.

The bond market, however, is already pricing in a future where AI’s productivity gains materialize. If these gains fail to arrive on schedule, today’s high yields could become a trap rather than an opportunity.

Why Long-Term Bonds Could Still Be a Steal

Zervos’s argument for long-term bonds is straightforward: buy now, before AI’s disinflationary effects take hold. Today’s high real yields present a rare opportunity to lock in returns before the market recognizes the productivity payoff. If AI delivers, these yields could appear exceptionally attractive in hindsight.

However, there’s a counterargument. Rising real rates pose a structural challenge for long-duration assets. The experience of ARKK, which saw a 71% peak-to-trough decline in assets under management, serves as a cautionary tale. If AI’s productivity gains fail to materialize, the same forces that make bonds attractive today could pressure them tomorrow.

The key is distinguishing between near-term opportunity and long-term risk. If AI delivers on its promises, today’s yields could prove to be a bargain. If it doesn’t, the bond market may face a prolonged adjustment, with real rates settling at levels unseen in the pre-AI era.

What Could Go Wrong?

Zervos’s thesis relies on three critical assumptions: that the $1 trillion in AI spending materializes, that it translates into measurable productivity gains, and that those gains drive disinflation. Any of these could fail.

If mega-cap tech scales back its AI investments, the capital call could evaporate, causing real rates to fall just as quickly as they rose. If productivity gains never materialize, the bond market could be left with a structural shift toward higher real rates—and lower bond prices. If inflation expectations resurface, the entire narrative could unravel.

For now, the spread between nominal Treasury yields and inflation breakevens is the metric to watch. If real rates continue to drive the move, Zervos’s thesis remains intact. If inflation expectations rise, the bond market’s repricing could take on a far more ominous tone.

Sources

• Reassessing market concentration in the AI era — invesco.com (primary source)
• How is AI influencing interest rates? Investment, productivity, prices, and more — minneapolisfed.org (primary source)
• Watch CNBC's full interview with Jefferies' David Zervos — cnbc.com
• ARK Innovation ETF (ARKK): The Rise, Fall, and What the Filings Reveal — developer.stockfit.io