On 1st October, Britain’s 30-year gilt yield briefly crossed 6%. It was the first time it had reached that level since 1998. Across the Atlantic, the 10-year US Treasury yield rose to 5.34%, its highest level since 20021. For more than a decade after the financial crisis, investors became accustomed to unusually cheap money. Central banks kept interest rates near historic lows, while quantitative easing created enormous demand for government bonds. Governments could borrow cheaply, companies could refinance cheaply, and investors were pushed towards riskier assets in search of returns. That environment is now disappearing.

The question is no longer simply whether interest rates are high, but rather who will pay for the growing demand for capital. Governments are one obvious source of that demand. Global public debt reached almost 94% of GDP in 2025 and the International Monetary Fund expects it to reach 100% by 20292. Governments are simultaneously facing greater spending pressures, from defence and energy security to ageing populations and infrastructure. The result is an uncomfortable combination: more borrowing, just as investors demand greater compensation for holding long-term debt.

That compensation is reflected in bond yields. When investors demand more, bond prices fall, and because prices and yields move in opposite directions, the yield rises. A higher yield therefore means a higher cost of borrowing for the government. But this is where the story becomes more interesting. Governments are no longer the only borrowers competing for investors’ money: the artificial intelligence boom is creating an enormous new demand for capital. Building the infrastructure required to train and operate increasingly powerful AI models means constructing data centres, buying chips and expanding electricity generation. Estimates put the investment required for AI-related data centres over the next three to five years at between $3 trillion and $5 trillion. Much of the initial investment by the largest technology companies was funded internally, but companies are increasingly turning towards debt markets3.

This affects bond investors because AI companies are now borrowing for much longer periods. Data centres are expensive, long-lived assets, making long-term fixed-rate financing attractive. Economists at the Dallas Federal Reserve note that Wall Street expects around $300 billion of AI-related investment-grade bond issuance in 2026, which could add as much as $360 billion of 10-year-equivalent duration to US bond markets3. In other words, the AI boom is beginning to enter the bond market. This creates an unusual collision. Both governments and technology companies need to issue more debt. Investors, meanwhile, do not have an unlimited supply of capital with which to buy it all.

When the supply of bonds rises faster than demand, borrowers may have to offer higher yields to attract investors. This does not mean that AI companies are directly “causing” government borrowing costs to rise. Inflation expectations, central-bank policy, fiscal deficits and investor demand are all contributing to the current sell-off4. But the sheer scale of investment required by AI adds another source of demand for financing at precisely the moment when governments are already borrowing heavily.

There is a second problem. Higher yields do not merely make new borrowing more expensive. They gradually make existing debt more expensive as governments refinance maturing bonds. This is why bond yields can become self-reinforcing: higher borrowing costs increase government interest payments, which can widen deficits. Larger deficits require more borrowing, increasing the supply of bonds that investors must absorb. If investors demand still higher yields, the cycle continues.

The IMF estimates that global interest payments have risen from around 2% of GDP to nearly 3% in just three years. That may sound small, but applied across the global economy it represents trillions of dollars that governments cannot spend elsewhere5.

Britain provides a useful example of how quickly this can become a financial constraint. On 1st October, the 30-year gilt yield reached 6.029%, while the 10-year gilt yield climbed to 5.51%6. These are not simply abstract numbers for bond traders. They feed into the wider cost of money, influencing mortgages, corporate borrowing and the returns investors require across financial markets.

For equities, the implications are particularly important. The value of a company today depends partly on the profits investors expect it to generate in the future. Those future profits are discounted back to their present value using an interest rate. When that rate rises, the present value of distant profits falls. This is particularly relevant to technology companies, whose valuations often depend heavily on expectations of profits far into the future.

There is an irony here. AI could ultimately make economies more productive, raise output and improve governments’ ability to service their debts. The IMF notes that stronger-than-expected AI-driven productivity could improve debt dynamics, provided real interest rates do not rise much in response2. But getting there requires enormous upfront investment. If that investment is financed through debt while governments are simultaneously expanding their own borrowing, the financial system has to absorb a huge amount of additional debt before those productivity gains materialise.

That creates a tension: investors may believe that AI will transform economic growth over the next decade, but bond markets are concerned with the return they receive today for lending money. A promise of higher productivity tomorrow does not eliminate the need to pay interest today.

The same tension exists in government finances. Governments can borrow to invest in infrastructure, defence or productivity-enhancing projects, potentially raising future growth. But investors still have to be convinced that the resulting debt is sustainable. Once confidence in that sustainability weakens, the cost of borrowing can rise before any economic benefit from the spending arrives. The recent bond sell-off should therefore not be viewed simply as a temporary market tantrum. That said, the IMF has described global bond markets as continuing to function in an orderly manner, despite the sharp rise in yields7. There is no immediate reason to treat the move as a repeat of a financial crisis.

However, there has evidently been a big change. For much of the post-2008 era, the defining financial problem was an excess of cheap capital. Today, the problem may increasingly be an excess of borrowers competing for expensive capital. Capital is no longer almost free. And when everyone wants to borrow at once, someone has to pay its price.

Footnotes

  1. Reuters, Why are world bond markets selling off again? (opens in a new tab), 1st October 2026.

  2. International Monetary Fund, Fiscal Monitor: Fiscal Policy under Pressure: High Debt, Rising Risks (opens in a new tab), 15th April 2026. 2

  3. Federal Reserve Bank of Dallas, How AI debt financing impacts duration supply and interest rates (opens in a new tab), 10th February 2026. 2

  4. Reuters, Bond markets take a drubbing again, 10-year Treasury yields highest since 2002 (opens in a new tab), 1st October 2026.

  5. International Monetary Fund, Fiscal: Growing Pressures on the Public Purse (opens in a new tab), 2026.

  6. Reuters, UK 30-year gilt yields top 6% for the first time since 1998 (opens in a new tab), 1st October 2026.

  7. Reuters, Bond markets functioning in orderly manner, IMF says (opens in a new tab), 1st October 2026.