AI INFRASTUCTURE, GDP, AND DEBT

U.S. GDP annual growth, as measured in the last quarter was 3.4%. This projects as an annualized GDP of just over $33 trillion. This was an increase over the previous quarter when GDP growth was 2.2%. Consumer spending, decreased exports, and increased imports contributed to this number.

One of the most important sectors affecting GDP growth was, and continues to be, AI infrastructure buildout. Information processing equipment spending is up 37.4%, and software sales are up 9.6% this year. Much of the increase in imports is due to these expenditures.

AI spending is estimated to have contributed over a third of the GDP growth seen in the last quarter.

The value of AI to almost all industry sectors is so great that companies are investing heavily and quickly in AI infrastructure so as to make sure that they are not left behind. But not all will survive.

I think it is important then, to understand how the spending on AI infrastructure has been financed.

Between 2020 and 2024, the major investors in AI (Alphabet, Meta, Amazon, Oracle, and Microsoft) used their operating cash flows to finance their investments.

However, in 2025 this changed. About 27% of the investment in AI was funded with debt.

And in 2026, that number increased to about 33% of that investment being funded through debt.

That translates into 10% of the GDP growth having been funded by debt.

In the third quarter this year, the GDP sat an annualized value of $33 trillion. So, the annualized debt would be around $3.3 trillion added to the economy.

And this reliance on debt only projects to get higher. Goldman Sachs projects that over the next 3 years, debt-financed AI investment will reach up to 50% of the spend on those projects.

HOW DO COMPANIES FINANCE THEIR AI INVESTMENT?

Rather than borrowing directly against their balance sheets, the large AI investors use joint ventures and project finance structures to borrow against specific data centers and power plants. This protects the corporations, but focusing on individual projects transfers risks to the lenders.

The increase in debt from the AI investors has resulted in an increase in corporate bond issuance. While those bonds amounted to about 2% of the total bonds issued between 2022 and 2024, they now amount to 9%, an increase of 450%. This places upward pressure on long-term borrowing costs across the entire economy.

WHAT ARE THE RISKS TO THE U.S. ECONOMY?

Most investment booms inevitably lead to corrections. Not every infrastructure project will succeed. Not every AI startup will survive.

Current technology requires large amounts of power and potable water to run the server farms. Advancing technology innovations may dramatically change one or both of those requirements, making currently built projects no longer profitable. Chips will continue to evolve making newer server farms more efficient, and therefore, less costly than those being built today. Companies may develop business models that supplant the current investment endeavors. So, what will happen when those corrections occur?

Economists and credit analysts have looked at this issue, and, not surprisingly, they conclude that there will be severe pain for tech equities, AI startups, and chip manufacturers.

However, there is added risk to the economy from such an event. A default on a large share of investment-grade bonds that have been issued to these AI investors would ripple through the capital markets.

If these AI investors stop spending, the demand for new bonds will decrease; credit spreads would tighten and bond prices would rise.

The big losers would be expected to be the smaller, leveraged data-center operators and private credit funds. These smaller firms rely on floating-rate credit and off-balance sheet loans backed by rapidly depreciating chips.

A decrease in capital investment by large AI investors might trigger a tech stock selloff, erasing tens of trillions of dollars of market value. And because higher income households own the largest amount of these stocks, they would be hit the hardest by such a correction. As this population group has been bolstering consumer spending despite soft wage growth, such a result would quickly cool retail and services spending.

Franklin Templeton suggests that the risks to the U.S. economy would be tempered a bit. They project a recession that would be more similar to the 2001 dotcom bust than the 2008 financial services bust. Their estimate is that even cutting AI Capex in half would only shave off about 0.25% from GDP growth.