The AI boom has reached a new stage.
For the past few years, investors mainly had to ask one question: Is demand real?
Increasingly, the answer is yes. AI spending, server demand, backlogs, and hyperscaler capex remain enormous.
Take Dell’s (DELL) latest earnings. The company’s Infrastructure Solutions Group—the part of the business benefiting most directly from AI—grew 89% year over year. Its AI backlog nearly doubled in a single quarter, from roughly $51 billion to more than $90 billion.
Simply put, AI is having no trouble finding customers.
But the sheer size of this buildout is creating a different problem: Someone has to pay for it all.
The biggest technology companies have funded much of the AI buildout so far with their enormous cash flows. Now, capital spending is growing faster than the cash they have available to fund it. As a result, companies are increasingly turning to debt and other outside financing.
The AI buildout is getting too big to fund with cash alone
For years, companies like Microsoft, Alphabet, Meta, and Amazon could finance huge investments almost entirely from the cash generated by their existing businesses.
But even their resources have limits. Let’s look at the numbers…
At the beginning of this year, Wall Street expected the major hyperscalers to spend roughly $515 billion on capital projects in 2026.
That estimate has since surged to around $775 billion. Alphabet, Amazon, Microsoft, and Meta alone are expected to spend around $700 billion in 2026.
And by 2027, annual spending is projected to top $1 trillion.
Meanwhile, aggregate free cash flow across several of the largest hyperscalers peaked near $400 billion in late 2024. Current estimates put it at roughly $21 billion by the end of 2026 as capital spending absorbs more and more of the cash coming in the door.
Remember, free cash flow is what’s left after a company pays its operating expenses and capital investments. The hyperscalers aren’t suddenly generating less cash from their businesses. They’re spending so aggressively on AI that very little may be left over afterward.
Alphabet offered a striking example last quarter. The company spent $44.9 billion on capital expenditures while generating $39.1 billion in operating cash flow, pushing quarterly free cash flow below zero for the first time since it went public.
Already, tech companies are issuing more bonds, bringing in outside investors, structuring leases and joint ventures, and using other forms of financing to keep construction moving.
AI-related debt issuance has already topped $220 billion this year, and Morgan Stanley estimates that number could reach roughly $570 billion by year-end—more than double last year’s level.
Looking forward, Sycamore Tree Capital estimates that roughly $2.9 trillion will be spent on data centers globally through 2028—and about $1.5 trillion will need to come from external financing.
None of this is inherently alarming. Borrowing money to invest in a fast-growing, highly profitable business can be a fantastic use of capital.
But it also changes the equation. Once companies begin relying heavily on outside financing, AI growth no longer depends solely on demand. It also depends on the cost and availability of money. And both are moving in the wrong direction.
The 10-year is making the AI boom more expensive
The yield on the 10-year Treasury has surged from roughly 4.2% to around 4.8% in only a few months. That might sound like a small move. But for financial markets, it’s significant.
The 10-year Treasury is one of the most important benchmarks in the global financial system. It influences mortgage rates, corporate borrowing costs, and the return investors demand from other assets.
As Treasury yields rise, businesses generally have to pay more to borrow. That’s manageable when you’re financing an ordinary expansion.
It becomes much more tenuous when an entire industry is trying to fund one of the largest capital spending booms we’ve ever seen…
Say investors are willing to lend a company money at 1 percentage point above the comparable Treasury yield.
With the Treasury at 3.5%, that debt might cost roughly 4.5%. With the Treasury at 4.8%, the same borrower could pay closer to 5.8%, even if investors view the company’s creditworthiness the same way.
Apply that difference across hundreds of billions of dollars of new financing and the cost adds up quickly.
More importantly, projects have to earn enough to clear a higher bar.
A data center expected to generate a 7% return looks attractive when financing costs 4%. That’s a healthy spread between what the project earns and what the money costs. But at a 6% financing cost, most of that cushion disappears.
Companies can respond by accepting lower returns, raising prices on their customers, contributing more of their own capital, or deciding some projects no longer make economic sense.
Any of those outcomes changes the profitability of the AI buildout.
And this isn’t a theoretical future problem. Long-term rates are rising while hyperscalers are simultaneously asking the bond market for unprecedented amounts of capital. Credit investors are beginning to price in the fact that companies once famous for massive cash piles and relatively little debt are becoming much bigger borrowers.
In fact, the AI borrowing boom may itself be contributing to higher rates. Reuters reported this week that AI-related corporate borrowing itself is adding to the pressure on yields as tech companies compete with the federal government and other borrowers for investor money.
In other words, AI is adding to the enormous demand for capital… which, in turn, is making borrowing more expensive.
Financing is becoming part of the AI thesis
The fundamental AI story remains incredibly strong.
Dell’s backlog shows that customers still want more infrastructure than suppliers can currently provide… Hyperscalers continue raising their spending forecasts… And companies are finding more ways to use AI throughout their businesses.
But that success has produced an infrastructure buildout unlike anything the technology industry has attempted before.
Big Tech’s cash flow financed the early stages… The next stage requires trillions of dollars for data centers, power, chips, networking, and everything surrounding them.
Increasingly, a large portion of that money will come from investors and lenders. And as long-term rates rise—or if expected AI revenues begin slipping—the economics get tighter.
That’s the connection investors need to watch.
Editor’s note:
As this buildout gets bigger, more expensive, and more dependent on outside capital, simply buying anything with an AI connection becomes a much riskier strategy.
In Curzio Alpha, we’re focused on finding the companies with the strongest economics, clearest demand, and best positioning as trillions of dollars continue flowing into AI infrastructure.
If you’re looking to profit from the next stage of the AI boom—without chasing the hype—check out Curzio Alpha.
