Look at any corporate balance sheet from a technology-heavy company these days and you will find something that did not exist in the same shape five years ago: a large, growing line of debt tied to physical infrastructure. Data centres. Semiconductor fabs. Power contracts. Networking. The AI buildout has moved from the research budget to the fixed-asset schedule, and the financing behind it is now visible in the corporate ledger.
The scale is worth stating plainly. Investment-grade corporate bond issuance in the United States reached about $187 billion in August 2026, the third consecutive monthly record. Technology firms account for roughly a third of the volume. These are treasury teams borrowing for multi-year projects, locking in rates and front-loading funding in a way that would have been unthinkable in a normal August.
What the money is actually for
It helps to be concrete about where the dollars land. A data centre is not one purchase; it is a cascade. Land, shell, cooling, power delivery, networking, servers, and the software to run it all. The bond proceeds flow into capital expenditure budgets that then feed an entire industrial chain — construction firms, equipment makers, power infrastructure, chip-tool suppliers. When a company raises a billion dollars of debt for AI infrastructure, it does not sit in a bank account; it becomes orders placed across a dozen industries.
That is why the borrowing records matter beyond the finance pages. They are a forward indicator of physical investment that will arrive over the next three to five years. The debt is contracted now; the factories, data centres and power plants it funds will be producing for a decade. The balance sheet is, in effect, the delivery schedule for the infrastructure economy.
It is also worth noticing what is missing from the picture: very little of this borrowing is for acquisitions. Previous issuance booms were driven by companies buying companies. This one is different in kind. It is construction financing — money for assets that will take years to build, and whose economics will be written in the coming years rather than read from an acquired earnings history. That changes how credit investors must evaluate the debt, and the adjustment is still under way.
The discipline question inside the treasury
What should interest anyone who reads company accounts is the management of this. Median net leverage among investment-grade technology companies has risen to about 2.1 times EBITDA from 1.7 times a year ago. That is a deliberate choice — borrowing to build assets whose returns are not yet proven — and treasury teams are managing it with varying degrees of care.
The careful ones are doing the unglamorous work: laddering maturities, mixing fixed and floating, keeping interest coverage healthy, and making sure the debt is matched to assets with long useful lives. The less careful ones are betting that the current enthusiasm justifies the leverage, which is exactly the bet that has historically gone wrong. The difference does not show up in the borrowing year. It shows up in the year the buildout either pays for itself or starts to weigh.
Interest coverage above 8 times across the sector buys time, and time is the thing the strategy needs most. If the infrastructure starts generating returns before the debt matures, the leverage was smart. If it does not, the balance sheet becomes a constraint instead of a platform. That is the wager, and it is a genuinely new kind of wager for conservative borrowers who, until recently, treated debt as something to avoid except in emergencies.
Why borrow at all, instead of issuing equity
It is worth asking why companies choose debt when equity markets are strong. The answer is a mix of math and signalling. Debt is cheaper than equity in most tax systems, because interest is deductible. Debt is also non-dilutive: it does not hand a share of a business that management believes will be worth far more in five years to today’s investors. And in a winner-take-most race, there is a real cost to moving slowly — a company that waits to raise equity may simply fall behind in the buildout.
That reasoning is sound up to a point. The risk is that every management team making it believes the same thing, which is exactly what a credit boom looks like from the inside. The market’s calm absorption of the supply — spreads stable around 123 basis points over Treasuries — suggests lenders have decided to go along. But lenders are price-setters, not prophets. The credit that looks sensible at issuance always does.
There is a subtle signalling element as well. A company that funds its buildout with debt is telling the market it expects the returns to be large enough to service the debt and then some — a more binding statement than selling equity, which carries no promise of future performance. Debt is a commitment device, and the boards taking it on are, perhaps deliberately, tying their own hands.
The line item to watch
For anyone tracking the corporate ledger, the AI infrastructure debt is the line item to watch over the next five years. Not because it is necessarily a mistake — it may well be the most productive borrowing of the decade — but because it concentrates a decade of investment decisions into a short window, financed with obligations that arrive on a fixed schedule.
The treasury teams that manage this well will be the quiet heroes of the next cycle: laddering the maturities, keeping the coverage ratios honest, resisting the urge to borrow one more billion because the market is willing. The companies that manage it badly will not be distinguishable today — only in the year the repayments land. The ledger has a new line, and like every line that ever mattered, it will be paid in full — either by the returns, or by the companies themselves.