Markets have finally run out of patience with sovereign risk.

That’s the overarching message from developed market bond yields, which legged higher still this year as a shifting buyer base demanded more compensation to loan money to unapologetically profligate governments.

At the same time, wave after wave of AI-related corporate issuance, much of it long-tenor investment grade, suggests supply could be structurally higher going forward. Governments, then, will need to compete for duration buyers with the world’s best companies.

Although the AI issuance story’s a recent development, the rest isn’t new. “Tighter funding conditions for DM governments shouldn’t be a surprise,” SocGen’s Wei Yao wrote, in a Monday note.

“Debt is high, structural deficits persist, the interest-rate cushion is fading as debt rolls over, spending demands are becoming more rigid and the marginal buyer of sovereign debt is more price-sensitive and sometimes more leveraged,” she added.

The figure on the left, above, gives you some context for the annualized pace of hyper-scaler debt sales (that pace lines up almost perfectly with the actual USD issuance total since Oracle kicked off the AI borrowing spree a year ago next month). In 2026, the five hyper-scalers’ net issuance is tracking to exceed all DM sovereign borrowing except America’s.

Needless to say, AI borrowing’s going to continue. Indeed, company analysts reckon the hyper-scalers alone will borrow at least $800 billion between them across 2027 and 2028. If this year’s issuance is any indication, half (or more) of that will be 10-years and out.

Again: That’s just five companies. If you pan out and guesstimate total AI-related, longer-tenor corporate supply, the figures will be considerably larger, and the competition for capital that much more intense.

“While the primary drivers of rising long-end sovereign yields are investors’ concerns around fiscal sustainability and reaccelerating inflation, in our view, part of the rise is attributable on the margin to hyper-scaler issuance [which] has pushed the price of money substantially higher, both outright and on the curve,” BNY Mellon’s David Tam said, adding that there’s a “circular” dynamic in play. “Higher yields attract the very investors buying hyper-scaler debt, enabling further issuance,” he wrote.

Recall the figure above, which shows you the breakdown of AI and data center debt issuance across 2025 and 2026.

The hyper-scalers are “inordinately issuing at the long-end relative to other issuers [as] their voracious cash needs forc[e] them to spread out their issuance across the curve,” Tam wrote, in the same piece, noting that although the group’s “outstanding debt only represents around 4% of the IG universe, they have accounted for one-third of issuance 30 years and longer.”

The good news is, demand for AI-related debt’s still quite strong, with offerings heavily oversubscribed even as some recent deals have seen larger concessions versus previous sales. There’s no buyers’ strike, Wei said. “Markets can still absorb both sovereign and corporate supply, but only at a higher real return.” (Emphasis mine.)

Recall that substantially all of the rise in 30-year US yields is reals. The figure below, which I highlighted 10 days ago, shows you the breakdown for 2026. If you pan out, you discover that this dates to “Liberation Day.” Since then, 30-year US reals are up ~50bps, while breakevens are unchanged.

Of course, as Wei went on to say, higher real rates “can reflect good news, such as stronger growth expectations,” but they can, and in this case at least partially do, reflect “bad news, such as fiscal risk and policy uncertainty.”

Scott Bessent’s tentative intervention to cap US bond yields last week may presage a more concerted effort to reinaugurate financial repression given the long odds (particularly in the US) of credible fiscal retrenchment. “Given Washington’s firmly pro-growth mindset, few expect any substantial spending cuts or tax rises, with the burden probably placed on efficiency gains,” ING’s Chris Turner remarked.

Bessent’s stepped up buybacks don’t constitute financial repression on their own, but they could be a step down the slippery slope, particularly if he follows through on the “threat” to upsize the program beyond the increases announced last week.

“[E]xpanding buybacks during a long-end selloff blurs the line between liquidity management and active control of borrowing costs,” Wei went on, in the course of expounding on a path paved with good intentions. In the event US long-term borrowing costs don’t come down, repression “could deepen incrementally,” she said, starting with more Treasury buybacks funded by more Bill issuance.

From there, the US (and other DM economies hard-pressed to rein in borrowing costs) could push for “regulatory preferences” that favor sovereign bonds while instituting policies to create captive pools of capital, with retirement savings being one classic example. Eventually, central banks could step in, citing the necessity of ameliorating market dislocations.

As Wei wrote in her Monday piece, it won’t necessarily be obvious, even to policymakers, that there’s an overarching plan. “Each step may have a defensible rationale,” as she put it. “The regime changes when the combined effect” of those steps “is to create captive demand and stop sovereign funding costs from fully reflecting inflation, duration and fiscal risks.”

So much for Kevin Warsh’s contention that markets should be encouraged (and allowed) to “play the ball, not the referee.”