When bond yields get too high: The risk to AI and global markets

Two dynamics playing out in financial markets will determine what happens to the global economy for months, perhaps years, to come.

The first is the relentless rise of government debt.

The second is the effect of rising borrowing costs on the artificial intelligence boom which has almost single-handedly driven equity markets for the past two years.

The two are converging, and their meeting point is the yield on the US 10-year government bond.

Warnings about the US debt burden are almost as old as the fiat dollar. They began in earnest in the 1970s, when the US under Richard Nixon abandoned the gold standard. Fiscal hawks were quick to predict that unconstrained borrowing would eventually culminate in a dollar and debt crisis.

But thanks to the dollar’s reserve currency status, it never arrived. The doom-mongers were drowned out by the bull market of the 1980s and the dotcom rush of the 1990s, and even the financial crisis of 2008 proved no more than a brief hiatus.

The government debt kept rising. Yet the quantitative easing of the 2010s, combined with an economy that never quite recovered its former productivity oomph, pushed yields ever lower, to a nadir of almost zero during the pandemic.

Deficits and debt have driven corporate profitability

What is perhaps underappreciated is how important this era of fiscal largesse and cheap money has been for profitability. Investors defending today’s stretched equity valuations argue that this is not at all like 1999; then, the dotcom bubble was populated with companies with little revenue and often zero profits. Today’s technology giants, by contrast, are among the most profitable companies yet.

This may be true, but it is not the whole story. In the last quarter, the circa 28% rise in average corporate profits – the sort of growth that is usually seen in a rebound from recession – has been taken as evidence that corporate America can shrug off almost anything, from trade wars to geopolitical ones. By almost any historical comparison, today’s US profit machine appears invincible.

But persistent government deficits have played an enormous role in these profits. A fiscal deficit above 6% of GDP, during a period of full employment, represents an enormous transfer of income to the private sector. It reaches the private sector either directly, through social security and government spending, or indirectly through tax cuts. Ruchir Sharma of Rockefeller International estimates that up to half of the apparent exceptionalism of US earnings is being financed by federal spending or tax cuts, a direct transfer from government borrowing to shareholders.

This arrangement has been relatively painless when the government could borrow cheaply. It becomes more problematic when it cannot. Since the pandemic the bond market has undergone one of its most extreme repricings in decades. US government bonds have sold off from their 2020 highs with the US 10-year Treasury now approaching 4.8% (yields move inversely to prices).

More troubling is what this means for how the US government continues to fund itself. Federal interest expense has more than doubled as a share of GDP in the past five years, to above 3%. Rising debt, combined with the need to refinance debt at ever higher rates, is converting what was once an abstract long-term fiscal concern into an immediate budgetary expense.

For those familiar with emerging markets, this will sound familiar; it is called a debt spiral.

AI is driving record borrowing in the private sector

This brings us to the second force: AI.

For most of the period between the 2010s and the early 2020s, rising government debt did not coincide with an equally dramatic borrowing binge in the corporate sector. Following their disastrous experience with real estate leverage in the financial crisis, most households and companies were cautious on debt.

That has all changed. The largest technology “hyperscalers” are now spending astronomic sums on data centres, chips, power infrastructure and everything else needed to power artificial intelligence. Until recently this was funded by their extraordinary cash flows. Increasingly however they are having to turn to debt markets. Barclays analysts estimate that hyperscalers could raise up to $1-trillion from public debt markets over five years, all against barely $200-billion of annual revenue.

The AI boom may be a bubble, but such bubbles can continue for surprisingly long, provided money remains cheap enough. This makes the 10-year Treasury yield perhaps the most important number in financial markets. At about 4.8% it is already close to the top of the range of the post-dotcom era. A decisive move above 5% would matter both economically and psychologically, with borrowing at such stratospheric levels.

We have been here before. In late 2023, the 10-year yield briefly approached 5% before retreating. As inflation moderated and investors began anticipating interest rate cuts, yields softened back towards 4%. Risk assets went on the tear they have been on ever since.

Bitcoin, always a useful barometer of liquidity, captured the shift perfectly. After trading at about $25,000 near the 2023 peak in yields, it subsequently ripped higher towards $125,000 by late 2025 as financial conditions loosened.

The question is: what happens now?

If yields subside as they did then, the financing costs for technology companies should ease and the AI investment cycle can continue. In that case, expect equities to power on. But if the combination of vast sovereign borrowing and rapidly increasing AI private sector demand for debt capital pushes yields higher, the consequences could be disastrous.

There is nothing magical about 5%. Markets do not collapse because a number on a Bloomberg terminal changes from 4.99 to 5.01%.

But thresholds matter because they alter behaviour and expectations. At some point, government bonds become sufficiently abundant that buyers demand a higher return to own them. And if investors can get more than 5% risk-free, the hurdle rate for a data centre rises with it, and a great many projects stop clearing it.

For the current cycle, that point may be uncomfortably close. Most investors know this and are holding their breath. The most important dynamic in finance is no longer one between bulls and bears. It is between Washington’s seemingly endless demand for debt and Silicon Valley’s investment boom.

Both are begging the bond market for money and patience. It is unclear how much it has left. DM