Some records are worth reading twice. Investment-grade corporate bond issuance in the United States reached roughly $187 billion in August 2026 — the third consecutive monthly record. Three straight record months has never happened before in the history of this market, which has been tracked since 1995. The reason is not mysterious: companies are borrowing to build artificial intelligence infrastructure, and they are doing it fast.
But the real story is not the borrowing itself. It is what the borrowing reveals about how capital structure strategy has quietly changed — and about the scale of the wager that corporate America is making on the AI buildout.
Why August, of all months
August is traditionally the quietest month in the primary bond market. People are on holiday, the desks are thin, and most issuers wait for September. That convention has been suspended. The previous monthly record was set in July 2026, which itself broke the record set in June. Issuers are front-loading their funding — locking in rates before the Federal Reserve’s next decision, and securing multi-year capital for projects that are already under construction.
The scale is unusual even for a record. The typical August has averaged around $95 billion over the previous five years. This August came in at nearly twice that. Year to date, issuance totals roughly $1.32 trillion, putting the market on pace to approach the full-year record of $1.93 trillion set in 2020 — which was, to be clear, the pandemic emergency year. The difference between 2020 and now could not be starker: then, companies borrowed to survive; now, they are borrowing to build.
Who is borrowing, and for what
The composition of the borrowing matters as much as the volume. Technology accounts for roughly 32 percent of August’s issuance, followed by energy at 18 percent and industrials at 15 percent. Compare that with the diversified pattern of earlier record months, and the AI-specific nature of this capital need is unmistakable.
Technology sector issuance alone reached about $60 billion in August, more than double the $28 billion of a year earlier. The average deal size is up to $1.8 billion from $1.4 billion. These are not companies topping up working capital. They are funding data centres, semiconductor fabs, energy contracts and networking — the physical plumbing that the AI boom actually runs on.
Notice also what is not in the mix: very little of this money is for the kind of M&A that drove previous issuance cycles. This is not consolidation debt. It is construction debt — money for assets that will take years to build and longer to depreciate. That makes the cycle different in a way credit investors are still calibrating: the collateral is not an acquired business with an earnings history, but a facility whose economics will be written in the coming years.
The balance sheet consequences
Here is where the strategist’s eye has to narrow. Median net leverage among investment-grade technology firms has climbed to about 2.1 times EBITDA from 1.7 times a year ago. That is a real move, but it has not yet caused stress: interest coverage remains strong, above 8 times, because profitability is high. The credit picture is ‘more debt, but manageable’ — so far.
The words ‘so far’ are doing a lot of work. What makes this different from a normal borrowing cycle is that much of the debt is being used to build assets whose returns are still being proven. That is not by itself a red flag — railroads, electricity and the internet were all built on borrowed money before they returned it. But it does mean the discipline test comes later than the spending decision. The market’s verdict on these balance sheets will be delivered in future earnings seasons, not at the closing dinner.
What the market is telling us
Credit spreads have stayed remarkably stable despite the supply surge — around 123 basis points over Treasuries on the main index. That is the market’s way of saying it believes the story: demand is absorbing the record issuance without demanding a big concession. Real-money investors, including pension funds and insurers, remain net buyers, particularly of longer-dated paper that matches their liabilities.
The stability is itself information. When a market absorbs record supply without blinking, it is signalling that lenders have decided the risk is worth pricing in the ordinary way — not that the risk is absent, but that they are comfortable owning it at these levels. That is the confidence you would expect at the top of an investment cycle, and it is worth holding in mind.
The next test arrives in September, when the Federal Reserve meets. A signal of rate cuts would reduce the urgency to tap markets and slow the pace; a signal of prolonged higher rates would keep issuers rushing to lock in. Either way, the direction of travel is set for the rest of the year: the pipeline of AI-related funding needs has not shrunk, only the timing of when it comes to market.
The strategy takeaway
Strip the AI excitement away, and what you are watching is a structural change in how capital is being allocated. Companies that historically borrowed conservatively are now choosing debt over equity to fund a multi-year buildout, because they want the leverage now and believe the returns come later. That is a deliberate strategic choice, not a panic — and it is the kind of choice boards only make when they believe the alternative (falling behind in the buildout) is more expensive than the leverage.
The honest question is the same one every capital-structure analyst should ask: what does this look like in three years? If the infrastructure returns, the leverage will look prescient. If it does not, the leverage will look like what it is — risk taken early for a bet that has not yet paid. The bond market has effectively placed its own bet by absorbing the supply calmly. For everyone watching from the outside, the signal is clear enough: the era of ‘AI as an experiment’ is over. It is now a line item on the balance sheet, with all the discipline, leverage and consequences that implies. Whether that turns out to be the smartest borrowing of the decade, or the most expensive, depends on returns that have not arrived yet. The strategy is committed. The verdict is still out.