TL;DR
· Goldman Sachs estimates that the five largest AI cloud giants have issued about $194 billion in debt this year, accounting for roughly 9% of global investment-grade bond supply.
· Apollo’s charts show that the order coverage ratio for related new bond issuance has fallen from nearly 5x in February to below 2x in July.
· This is not yet a default crisis, but rising rates or weaker credit sentiment could magnify the divergence between stocks and bonds.

According to a Goldman Sachs report, AI capital expenditures are pushing the U.S. large technology companies’ financing needs toward the investment-grade bond market.

As of early July 2026, among the five mega-scale cloud service providers—Alphabet, Amazon, Microsoft, Meta, and Oracle—these firms have collectively issued about $194 billion in debt this year, representing about 9% of global investment-grade bond issuance. About 32% of this issuance was completed in non-U.S. dollar markets, indicating that financing pressure is not confined to the U.S. dollar bond market.

The more sensitive change is on the demand side. Market charts cited by Apollo show that the order coverage ratio for this type of bond—how many investor orders there are for each $1 of issuance—has dropped from nearly 5x in February to below 2x in July. Bonds are still being sold, but investors are no longer as eager as at the start of the year.

The AI story remains one of growth and upside expectations in the stock market. In the bond market, it turns into a more realistic question: who will absorb the growing supply of bonds, and what level of yield will investors require to buy?

Stock investors bet on AI to rise, while the bond market first looks at financing pressure

Equity investors look at AI capital expenditures and see compute capacity expansion, cloud revenue growth, and AI applications rolling out. Bond investors see something more direct: the higher the capital spending, the faster free cash flow is consumed, the greater the external financing needs, and the more investment-grade bonds the market must absorb.

The focus of this report is not on whether the long-term AI upside story holds, but on highlighting that the capacity of the credit market is being tested.

The credit quality of large technology companies has not deteriorated suddenly. Alphabet, Microsoft, Amazon, Meta, and others are still among the strongest companies globally in terms of cash flow and balance sheets. The issue is that even with high issuer quality, when supply size is large enough and issuance frequency is high enough, the bond market will demand higher compensation.

The free cash flow of the five largest cloud companies is expected to fall meaningfully over the next 12 months, and Oracle’s free cash flow will turn negative.

Since the start of this year, debt issuance by mega-scale cloud service providers has reached about $194 billion and has covered markets including the U.S. dollar, euro, British pound, and Japanese yen. By share, this group has already contributed nearly one-tenth of global investment-grade bond supply.

A comparison of stocks and bonds’ performance shows that in the first two months of 2026, the S&P 500 and cumulative returns on U.S. dollar investment-grade credit were close to each other, but after mid-March bond performance weakened.

Order coverage slips below 2x—investors aren’t as eager as before

More indicative of stress than issuance size is the rapid decline in new-bond order coverage.

In February this year, when mega-scale cloud service providers issued bonds, the order coverage ratio was close to 5x. By July, that number had dropped to below 2x. For the investment-grade bond market, this still does not mean “nobody is buying,” but it is clearly no longer in the easy, heavily subscribed state seen at the beginning of the year.

Thinner demand for orders usually leads to two outcomes. Issuers need to offer more attractive coupons or spread levels, and investors become more selective about maturity, issuer, and price.

Public market cases are also validating this shift. Grant’s and Bloomberg’s related paraphrases mention that for an Amazon bond issuance in July at roughly the $25 billion level, the final order book was about $41 billion—coverage of about 1.6x, below the average level of U.S. high-grade bonds in 2026.

This is also why these credit signals have relevance for stock investors. The stock market is willing to pay higher valuations for AI capital expenditures because these outlays may translate into future revenue. The bond market does not directly participate in upside gains; it first focuses on cash flow, leverage, supply, and the refinancing environment.

When demand for new bonds thins out, what it signals is not “the end of the AI story,” but rather that financing costs and market absorption capacity are becoming more important.

The order coverage ratio for bond issuance by mega-scale cloud enterprises has fallen from nearly 5x in February 2026 to below 2x in July.

Capital expenditures before 2030 may reach $5.8 trillion—issuance pressure isn’t over yet

Goldman Sachs’ stress test pushes the issue out over a longer timeline.

Based on its estimates, for fiscal years 2025 to 2030, total capital expenditures by mega-scale cloud service providers could reach about $5.8 trillion. If a substantial portion of that is financed through investment-grade bond issuance, the market would have to face a round of sizable new supply every year. Some paraphrased versions of the estimates indicate that if roughly half comes from bond issuance, annual incremental debt financing could be close to $550 billion.

This number needs to retain the “stress test” framing. It is not an already-confirmed financing plan, and it does not mean that all companies will issue bonds in the same proportions. The real question is that if AI infrastructure investment continues to be revised upward, the bond market will need to provide increasingly more funding.

Breaking it down, Goldman Sachs says that if about two-thirds of incremental financing comes from U.S. dollar markets and one-third from non-U.S. dollar markets, both the U.S. dollar investment-grade and the euro investment-grade markets would face additional supply pressure. The proportions may not look out of control, but investment-grade bond markets are not guaranteed an unlimited inflow of capital.

In the past few years, net inflows into the U.S. dollar and euro investment-grade markets have generally been able to cover net issuance. If the interest-rate environment worsens or credit-market sentiment weakens, inflows could slow, reducing the bond market’s ability to absorb incremental AI financing.

This isn’t a default crisis—the real trouble is that financing costs constrain spending in return

This round of stress is different from traditional credit crises.

Large technology companies are not in a high-leverage cycle industry; investment-grade ratings remain solid, and behind AI capital expenditures there are genuine business needs. A more accurate description right now is that the bond market is facing a supply digestion test: issuer quality is good, but there are simply too many bonds. Investors are willing to buy, but they need higher compensation.

Morgan Stanley also expects that AI and data center capital expenditures will reach the multi-trillion-dollar range. This kind of forecast points to the same reality: the faster AI infrastructure expands, the harder it is to avoid rising external financing needs.

This is not just a bearish conclusion for tech stocks. Stocks and bonds face different return structures: equities can bet on revenue upside elasticity from AI, while bonds take on more of the downside and supply pressure. In the short term, the two could continue to diverge: share prices may rise on AI expectations, while credit spreads widen due to increased bond issuance.

What to watch is that when credit spreads rise, order coverage declines, and capital inflows slow all occur at the same time, financing costs will feed back to affect the pace of capital expenditures. For AI trading, the originally more long-term growth narrative can be pulled back into a more realistic question: where does the money come from, and what price will the bond market be willing to pay?

Click to learn more about Lydong BlockBeats when recruiting for positions

Welcome to join the official Lydong BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group chat: https://t.me/BlockBeats_App

Twitter official account: https://twitter.com/BlockBeatsAsia