The AI boom used to be paid for mostly out of cash. That is no longer true. The biggest technology companies are now some of the heaviest borrowers in the corporate bond market, and they are doing it just as government bond yields hit their highest levels since 2007. The collision between AI financing and a 5% Treasury market is becoming one of the most important stories for anyone trading stocks, credit or rates.
How big the borrowing has become
The five largest cloud and AI spenders, Amazon, Alphabet, Microsoft, Meta and Oracle, are guiding to well over $600 billion of combined capital spending this year, roughly three times what they spent in 2024. Estimates from Bank of America suggest that spending will absorb around 90% of their operating cash flow. That leaves little room for buybacks and dividends, and it means the gap increasingly has to be filled with debt.
The numbers show it. Hyperscaler bond sales hit a record of about $108 billion in 2025, and 2026 is running far ahead of that pace, with several of the largest corporate bond deals on record, including multi-tranche offerings running into the tens of billions and even a rare 100-year "century" bond. Goldman Sachs data point to around $420 billion of gross hyperscaler issuance in 2027, another 60% jump. Add in off-balance-sheet structures such as data-center joint ventures and leases, and the total amount of AI-related borrowing is larger still.
The bond market is getting pickier
A year ago, investors eagerly took anything a top-rated tech company issued. That has changed. The high-grade corporate market has effectively split in two:
The main complaint from bond buyers is visibility. They can see the spending. They cannot yet see the return on it, and a flood of supply from the same few names makes it harder for the market to absorb each new deal.
Why this matters for Treasury yields
AI borrowing does not only affect tech credit. It competes for the same pool of long-term money that buys government bonds. When pension funds, insurers and bond funds are being offered a steady stream of long-dated, high-quality corporate paper at attractive yields, they have less appetite for long Treasuries at the margin. That is one reason analysts cite AI issuance as a contributor to the rising term premium, the extra yield investors demand to hold long government bonds, alongside oil-driven inflation fears and large government deficits.
It works in the other direction too. Higher Treasury yields raise the cost of every new AI bond, which raises the hurdle rate for every data center. A project that made sense when the company could borrow at 5% looks different at 7%.
What to watch as a trader
The bottom line
The AI buildout has moved from being paid for out of profits to being financed in the bond market, and the bond market is starting to set terms. At 5% Treasury yields, capital is no longer cheap, and every AI project now has to clear a higher bar. Follow the spreads and deal flow on AI debt: they will likely tell you how long the boom can keep its current pace before stock prices do.
How big the borrowing has become
The five largest cloud and AI spenders, Amazon, Alphabet, Microsoft, Meta and Oracle, are guiding to well over $600 billion of combined capital spending this year, roughly three times what they spent in 2024. Estimates from Bank of America suggest that spending will absorb around 90% of their operating cash flow. That leaves little room for buybacks and dividends, and it means the gap increasingly has to be filled with debt.
The numbers show it. Hyperscaler bond sales hit a record of about $108 billion in 2025, and 2026 is running far ahead of that pace, with several of the largest corporate bond deals on record, including multi-tranche offerings running into the tens of billions and even a rare 100-year "century" bond. Goldman Sachs data point to around $420 billion of gross hyperscaler issuance in 2027, another 60% jump. Add in off-balance-sheet structures such as data-center joint ventures and leases, and the total amount of AI-related borrowing is larger still.
The bond market is getting pickier
A year ago, investors eagerly took anything a top-rated tech company issued. That has changed. The high-grade corporate market has effectively split in two:
- AI issuers pay more. Spreads on AI-related debt, the extra yield over Treasuries that investors demand, have been running around 115 basis points, against about 78 basis points for the broader investment-grade market. That is a notable premium for companies with some of the strongest balance sheets in the world.
- New deals need sweeteners. Amazon's $25 billion sale in July had to offer roughly 18-21 basis points of extra yield on its longest bonds to get done. A Meta-linked Texas data-center financing priced near 7.5%, well above a comparable deal a year earlier.
- Traditional issuers are being rewarded. Banks and industrial companies are seeing strong demand for their bonds, as investors rotate toward borrowers whose cash flows they can forecast.
The main complaint from bond buyers is visibility. They can see the spending. They cannot yet see the return on it, and a flood of supply from the same few names makes it harder for the market to absorb each new deal.
Why this matters for Treasury yields
AI borrowing does not only affect tech credit. It competes for the same pool of long-term money that buys government bonds. When pension funds, insurers and bond funds are being offered a steady stream of long-dated, high-quality corporate paper at attractive yields, they have less appetite for long Treasuries at the margin. That is one reason analysts cite AI issuance as a contributor to the rising term premium, the extra yield investors demand to hold long government bonds, alongside oil-driven inflation fears and large government deficits.
It works in the other direction too. Higher Treasury yields raise the cost of every new AI bond, which raises the hurdle rate for every data center. A project that made sense when the company could borrow at 5% looks different at 7%.
What to watch as a trader
- AI credit spreads as an early warning. If spreads on AI-linked bonds keep widening while the broader credit market stays calm, the bond market is flagging stress before the stock market does. Credit investors have historically been quicker than equity investors to question overextended spending.
- Deal reception. How new AI bond deals are priced tells you a lot. Big concessions, reduced order books or deals pulled from the market would be meaningful warnings. Deals that price tight and are heavily oversubscribed would suggest the funding window is still wide open.
- Equity funding. If borrowing becomes too expensive, companies may turn to share sales, as Oracle has already signalled with plans to raise up to $50 billion through a mix of debt and equity this year. Share issuance dilutes existing holders and tends to weigh on the stock.
- The chip supply chain. Suppliers such as Nvidia depend on this financing. If the hyperscalers slow issuance, they are likely to slow spending, and the effect would reach chipmakers and data-center suppliers first.
The bottom line
The AI buildout has moved from being paid for out of profits to being financed in the bond market, and the bond market is starting to set terms. At 5% Treasury yields, capital is no longer cheap, and every AI project now has to clear a higher bar. Follow the spreads and deal flow on AI debt: they will likely tell you how long the boom can keep its current pace before stock prices do.
clean
by ai-agent
— Support AI-stocks news pairs, user-directed 2026-09-25