The competition for capital in an AI world

Governments have rarely needed so much capital. The problem is that everyone else needs it too. That competition for capital is showing up at the long end of global yield curves, complicating the backdrop for policymakers as they gather in Wyoming this week.

Economics 101 hinges on supply and demand – when supply exceeds demand, prices fall. That is what is playing out at the long end, where heavy issuance from governments to hyperscalers is pushing yields higher as investors demand greater compensation to absorb it.

The rising competition is captured in global bond yields testing multi-year highs. The US 30y Treasury yield climbed above 5.3% last week, its highest level since 2007¹. Long-end yields have also moved sharply higher across developed markets including Germany, Japan, and the UK. Inflation expectations have been relatively contained through much of the move, leaving a greater role for term premium as markets confront rising risks, from fiscal sustainability to a crowding out as big tech taps debt markets.

Buybacks can ease the pressure, not the supply

The rates selloff prompted the US Treasury to step in. It announced it would at least double the size of some buybacks of 10y to 30y securities, from $2B to at least $4B per operation to drive down yields². The announcement pushed yields roughly 10bps lower with long-end term premium compressing modestly, but this relief-rally proved short-lived³.

The Treasury funds buybacks with new issuance, mostly at the front end. It is swapping the maturity profile, not eliminating the government’s financing requirements. While that can improve liquidity and temporarily relieve pressure in targeted parts of the curve, it cannot fundamentally change the supply-demand balance.

The move also signals that policymakers are paying close attention to the long end, introducing more two-sided risk around future issuance decisions.

The fiscal competition intensifies

Meanwhile, the US federal debt just crossed $40 trillion, larger than the size of the economy, and the Congressional Budget Office estimates the annual deficit to climb above $3 trillion in the next decade after tracking $2.1 trillion for this fiscal year⁴. Global public debt, meanwhile, rose to roughly 94% of GDP last year and the IMF expects it to reach 100% by 2029, as governments face growing spending demands from defence, ageing populations and infrastructure⁵.

More important for markets is what it costs to service what governments owe. Net interest expense now exceeds $1 trillion a year in the U.S. and is projected to roughly double over the coming decade⁶. That would amount to roughly two-thirds of every dollar the federal government borrows being used to service the debt.

As existing debt rolls over, higher borrowing costs feed into interest expense, putting more pressure on deficits and requiring more issuance. The feedback loop matters given much larger public debt stocks and structurally higher borrowing costs.

Everything, everywhere, all AI

At the same time, the AI investment cycle is creating another enormous demand for capital.

Hyperscalers are spending hundreds of billions of dollars on data centres, chips, power and the infrastructure needed to support them. AI-related issuers, including chipmakers and data-centre developers, have already raised close to $340 billion in US markets this year⁷.

Where that borrowing lands on the curve matters as much for rates as the amount being issued.

Analysis from the Dallas Fed estimates that $300B of AI-related investment grade issuance carries the interest-rate risk of about $360B in 10y UST notes because it skews longer-dated, or roughly an eighth of what Treasury issuance supplies⁸.

Hyperscaler capital spending growth may moderate, but the AI buildout still leaves governments and some of the world's strongest corporate borrowers competing simultaneously for long-term capital.

For investors, a 30-year Treasury may look attractive relative to where it traded six months ago. But the question is whether that yield offers enough compensation relative to long-dated IG credit, infrastructure assets and relatively attractive sovereign yields elsewhere.

And that may help explain why the long end has been so difficult to anchor.

Who buys the long end?

For much of the period after the financial crisis, governments benefited from an unusually favourable combination of weak private investment and abundant savings, with central banks willing to absorb ample government debt. Today's dynamics are materially different.

That makes Jackson Hole more than watching Warsh and policymakers for any hints of forward guidance. Markets need greater clarity on the Fed’s reaction function in terms of how much weight it puts on still-elevated inflation, and how it responds when structural forces are pushing long-term yields higher. Especially after Warsh noted the rise in nominal and real yields had done “quite a bit” of the Fed’s work at the July meeting⁹.

Rate decisions still dominate the front-end, but they have less ability to offset a long-end influenced by the volume and composition of debt supply. And central banks that once absorbed large quantities of government debt through QE are no longer playing that role, raising the question of who will be the incremental buyer of long-dated debt.

Higher yields will create their own demand, and Treasury's buybacks can help at the margin. But if markets are being asked to absorb structurally more duration, investors may need to rethink what counts as a high yield. A more price-sensitive buyer base can absorb the supply, but only at higher yields that compensate them.

For now, we prefer taking yield where it is compensated rather than extending duration at current levels. Short and intermediate maturities offer most of the income with far less exposure to a supply and demand story that has yet to be resolved. For longer-term exposure, municipal bonds are particularly attractive due to the relative steepness of their yield curve.

The market will absorb the supply. The question is the price it takes to do so. Economics 101.

With assistance from Quinn Brody, senior macro analyst