AI Borrowers Make Up Nearly Half of Long-Dated Investment-Grade Bond Supply, PIMCO Says

The AI buildout is increasingly being paid for with borrowed money, and the longest-dated corner of the corporate bond market is where it shows most.

AI-related borrowers account for nearly half of year-to-date gross supply of non-financial U.S. investment-grade debt with maturities beyond 15 years, PIMCO said in a note dated Nov. 10. The note was written by Lotfi Karoui, a managing director and multi-asset strategist at the firm.

Hyperscalers, the cloud giants building the data centers, favor the long end. PIMCO says they have captured more than 30% of the 30-year bond market in certain periods, against roughly 13% for 10-year paper.

The cash flow gap

Spending that was once funded from operating cash flow is shifting to debt. Figures compiled by BigGo Finance show Microsoft, Alphabet, Amazon, Meta and Oracle generated $542.9 billion in operating cash flow in the first nine months of 2026, a record and up 29% from a year earlier. Their capital spending doubled to $556.9 billion over the same period.

That leaves a free cash flow deficit of about $14 billion and, per the same figures, an 83% cut in buybacks. S&P has cut its Meta outlook to negative.

Goldman Sachs projects about $250 billion of dollar-denominated bonds from hyperscalers this year. Its estimates for the wider AI ecosystem, including chipmakers, utilities and data center developers, run from $300 billion to $570 billion.

Goldman also expects debt to fund about a third of Big Tech capital spending this year, peaking at around 35% next year, according to CNN's account of the bank's forecast. JPMorgan projects AI spending has reached $1.6 trillion since 2023 and will hit $1 trillion a year for roughly the next decade.

Stocks sold off first

Equity investors reacted before bond investors did. In the week ended Oct. 30, the Nasdaq Composite fell 2%, the worst of the major indexes. Alphabet closed down 12.29% on Thursday, Oct. 29, after weak advertising revenue raised questions about its AI costs.

On Friday, Oct. 30, Intel plunged more than 28% on weak fourth-quarter guidance. Amazon, which fell more than 7% in early trading after missing estimates and slipping to negative free cash flow because of AI spending, closed down 1.41%. The Nasdaq ended that session down 2.86%.

Richard Saperstein, chief investment officer at Treasury Partners, framed the question this way: whether spending "can continue to be financed by cash flows, or if equity and debt issuance will be necessary, too."

Political risk

The Democratic wins in the Nov. 3 midterms added a new variable. PIMCO's Libby Cantrill noted that many new Democratic governors-elect and other officials ran on platforms with "varying degrees of restrictions" on data center development.

Kevin Thozet of Carmignac said markets are pricing in the risk that Democratic wins in the states with the most projects will make construction "more expensive and longer to complete." Local opposition to data centers has centered on air pollution, higher energy costs, water use and noise.

PIMCO's own read is less dire for the buildout itself. The firm says restrictions could slow construction and near-term issuance, but are "more likely to change the journey than the destination." Demand for compute remains strong, it says, and the ultimate scale of the buildout is unlikely to shift materially.

That is the opposing view to the selloff's premise. A slower pace would spread financing over a longer period, PIMCO says, but "does not eliminate" the funding need. The concentration problem for credit investors stays.

What PIMCO tells investors to buy instead

Karoui's recommendation is to diversify without giving up quality. He names agency mortgage-backed securities, higher-rated non-agency residential mortgage bonds and emerging-market debt.

Financials and "old economy" sectors like energy offer less AI exposure, he says, but both are expensive, with spreads that look tight against their risk-adjusted return potential.

Agency MBS offered a 33-basis-point yield pickup over investment-grade corporate spreads as of Nov. 5, per PIMCO and Bloomberg data cited in the note. Karoui says agency MBS has held up better against rate volatility than in 2023.

The traditional hedge is also shaky. Longer-dated Treasurys have been rattled by volatility for much of 2026, though PIMCO says they have rallied since late September while risk assets sold off.

Stakes beyond Wall Street

Economists tie the exposure to the broader economy. Bruno Cavalier of Oddo BHF says data center spending, up 150%, contributed 30% of U.S. growth in 2025 and 2026. He also says 43% of household wealth is now exposed to the stock market, up 10 percentage points from 2000.

Alexandre Baradez of IG France says the richest 10% of Americans account for half of consumer spending. He argues a market downturn could hit that spending just as average households have pulled back since the inflation spike he attributes to the U.S. war against Iran.

The warnings are not limited to equity analysts. PIMCO CEO Emmanuel Roman told a Hong Kong investment summit that an AI downturn is coming, "and then things are going to break." He named U.S. rate hikes, war or a eurozone crisis as possible triggers. His firm has estimated the buildout could require $2 trillion in financing over five years.

PIMCO is also a participant. BigGo reports the firm helped arrange CoreWeave's first chip-backed financing in 2023 and bought $4 billion of SpaceX bonds last month.

Jim Zelter, Apollo Global Management's president, said society will use "dramatically more computing power in the future," but that "what we don't know is how pricing and economics will evolve."

It remains unclear how long lenders will keep absorbing the supply. CNN's analysis argues the bond market has largely shrugged off the equity selloff so far, and that its appetite will decide whether the buildout succeeds. If Goldman's forecast holds, debt funding of hyperscaler spending will reach about 35% next year, which would be the first real test of that appetite.