In our most recent “The Credit Market Lens,” we pushed back against one version of the AI crowding-out story. We did note that a large investment boom can still put upward pressure on equilibrium real yields through the saving-investment channel. However, the evidence for a narrower portfolio-rebalancing channel – that is, AI bond supply directly crowding out U.S. Treasuries – looked weak across nominal yields, term premia, and swap spreads, at least for now.
This note asks the same question one layer down in corporate credit. If AI issuance is not visibly crowding out Treasuries, is it crowding out other corporate borrowers? As shown in Figure 1, AI borrowers have accounted for almost one-quarter of total nonfinancial supply in the USD bond market, up from less than 13% last year and 4% in 2024.
Figure 1: AI borrowers have accounted for almost one-quarter of total nonfinancial supply in the USD bond market, year-to-date
To assess whether that surge is crowding out other corporate borrowers, we look at both price and quantity.
Pricing data: Spread and curve measures suggest repricing, not displacement
The pricing evidence shows no sign of crowding out. If non-AI issuers were being rationed out of long-duration markets by hyperscaler supply, we would expect to see their spreads widen or their curves steepen as the market demanded compensation for that displacement. We see neither. Spreads for nonfinancial issuers excluding hyperscalers and NVIDIA have not widened despite the surge in AI issuance (see Figure 2). Similarly, the back end of hyperscalers’ spread curves has steepened materially, both outright and relative to the rest of the nonfinancial universe, where curves have flattened (see Figure 3). In other words, the widening of hyperscaler spreads appears to have been the primary mechanism through which the AI “excess supply” has been absorbed.
Figure 2: Spreads show little evidence of hyperscalers crowding out broader investment grade corporates
Figure 3: Hyperscalers’ back-end curves have steepened, both outright and relative to the rest of the nonfinancial universe
That repricing has created some interesting relative value “anomalies”: High quality hyperscalers (those rated A or higher) now trade wider than utilities despite meaningfully stronger balance sheets on spot leverage, cash flow, and liquidity metrics (again, Figure 2). Also, the Bloomberg AA Corporate Index, which is increasingly dominated by high quality hyperscalers, trades at one of the thinnest spread premia to the broader IG market, as measured by the Bloomberg U.S. Corporate Investment Grade Index.
Quantity data: Unable or unwilling?
On the quantity side, the evidence countering the AI crowding-out story is more nuanced. Overall gross issuance from nonfinancials (ex AI issuers) is up from a year earlier, continuing a multi-year trend (see Figure 4). But net issuance for long-maturity bonds shows a material contraction in 2025 as well as year-to-date (see Figure 5).
Figure 4: Overall gross issuance from nonfinancials (ex AI) has continued growing
Figure 5: Net issuance of longer-term corporates shows a material contraction in 2025 and 2026
Source: Bloomberg and PIMCO as of 31 August 2026, using the ICE BAML USD IG Corporate index
This pattern could look like crowding out at first glance, but quantities alone cannot distinguish between an inability to issue and an unwillingness to issue. The more prosaic explanation is that most companies are not passive takers of the maturity structure available to them. They time it. When long-term borrowing looks relatively cheap, they term out their debt issuance, and when it becomes more expensive, they favor the shorter maturities. This relationship is well established in the academic literature,Footnote1 and Figure 6 shows the same relationship in our own data: Rising long-dated yields have tended to coincide with declining long-duration net corporate supply.
Figure 6: Higher 30-year U.S. Treasury yields may reduce the incentive for corporate borrowers to issue long-end bonds
Corporate CFOs and treasurers don’t have some perfect model of expected excess returns. For them, this maturity choice is a real corporate finance decision shaped by the trade-off between locking in funding and preserving flexibility.
The academic literature also shows that, at the aggregate level, corporate borrowers have tended to adjust the amount of duration they supply when the government changes the amount, or price, of duration in the market.Footnote2
Simply put, companies manage duration actively, so a decline in long-end issuance does not automatically mean they have been crowded out, especially in the current environment where long-dated yields have risen materially. One might argue this reflects Treasuries crowding corporate borrowers toward shorter maturities (a discussion for another time), but it is not evidence that hyperscaler issuance is driving this shift.
Bottom line
Within corporate credit, the overall picture is more nuanced than the simple crowding-out narrative that AI-related debt, especially hyperscaler debt, is overwhelming other issues. We just don’t see the evidence.
So far, the market adjustment has been concentrated primarily where supply originates – the hyperscalers themselves – rather than crowding out other corporate borrowers. Where the maturity structure of non-AI issuance has shifted, the evidence points to Treasury curve dynamics and opportunistic maturity timing, not hyperscaler-specific displacement, as the driver.
Michael Puempel and Gabriel Cazaubieilh contributed to this report.
1 See Baker, Greenwood, and Wurgler (2003), “The maturity of debt issues and predictable variation in bond returns,” Journal of Financial Economics, 261-291, Volume 70, Issue 2 Return to content
2 Greenwood, Hanson, and Stein (2015), “A Comparative-Advantage Approach to Government Debt Maturity,” Journal of Finance, 1683-1722, Volume 70, Issue 4 Return to content
