10-minute read
Global bond markets have produced no shortage of drama in recent weeks, with long-term yields in several major economies reaching multi-decade highs. A sharp repricing has pushed 10-year US Treasury yields to around 5%, putting concerns about government borrowing and debt sustainability back in focus.
Part of the move reflects a more hawkish Federal Reserve (Fed) – which raised rates by 25 basis points this month amid persistent inflation – and a higher expected path for short-term interest rates over the next couple of years.
But it also raises a bigger question: are investors demanding higher returns to hold long-term bonds as governments borrow more and uncertainty around inflation and fiscal policy increases? And how does the rise in debt fit into that picture alongside the AI investment boom?
In this month’s CIO Insights, we look at what’s driving higher bond yields, whether rising government borrowing and the AI investment boom could keep them elevated, and why what matters for your portfolio isn't only whether rates stay high, but whether growth keeps pace.
Key takeaways
- Why are long-term yields higher? It’s not just about the Fed. The latest pick-up partly reflects the Fed’s more hawkish stance and expectations that policy rates will stay higher for an extended period. But over the past several years, the bigger structural shift has been a rise in the term premium – the extra return investors demand for holding longer-term bonds. That likely reflects greater uncertainty around inflation, fiscal policy, and the growing supply of government debt investors need to absorb.
- US fiscal pressures make higher yields more likely to persist. The government is borrowing far more than it did before COVID, with large deficits and a rising debt burden projected to continue for years. Rising interest costs add to those borrowing needs as the debt stock grows and older debt is refinanced at higher rates. That persistent need to issue more Treasuries puts upward pressure on long-term yields, making higher levels more likely to endure.
- AI is adding to the amount of borrowing markets need to absorb – but it could also help ease the debt problem. The AI buildout isn’t the main reason yields are higher today, but its scale is becoming large enough to move the needle as companies invest heavily and increasingly borrow to finance that spending. If those investments lift productivity and growth, however, they can also expand the economy and tax base, helping to slow the rise in debt relative to GDP. Stronger growth could therefore improve the fiscal outlook even while keeping real rates higher.
- Higher yields don’t have to derail equities, but they remain a challenge for long-duration bonds. If AI continues to support stronger growth and earnings, equities can still benefit from rising nominal profits even with higher rates. But even in that more optimistic scenario, US deficits are likely to remain wide and debt levels are likely to continue rising – keeping upward pressure on the term premium and leaving long-duration bonds vulnerable. Our base case is that markets will likely swing between pricing stronger AI-led growth and greater fiscal risk, depending on whether the growth payoff is strong enough to offset worsening fiscal pressures.
(See our Glossary at the end for a breakdown of the terms used in this article.)
Why are long-term yields higher? It’s not just about the Fed
To understand what’s been driving long-term bond yields higher, it helps to break the 10-year Treasury yield into its underlying components. There are a few ways to do that.
One way splits the yield into two parts: 1) where investors expect short-term interest rates to average over the next decade, and 2) the term premium – the extra return they demand for holding a longer-term bond. As the top panel in Exhibit 1 shows, most of the increase in 10-year yields since the start of the year has come from higher expected short-term rates.
That fits with the Fed’s more hawkish turn in recent months, with policymakers signalling that rates may need to rise further following September’s hike and remain elevated through 2027. The term premium, meanwhile, has been relatively stable recently.
Another way to look at the 10-year yield is by separating the real yield – which is adjusted for inflation – from inflation expectations (see the bottom panel in Exhibit 1). Here, too, most of the recent increase has come from real yields, while longer-term inflation expectations have moved much less. Taken together, the latest rise in yields is mostly a repricing of the expected path of short-term interest rates, rather than a renewed rise in long-term inflation expectations.
Over a longer horizon, though, there has also been a more persistent shift. In the past few years, the term premium has steadily risen from the very low – and at times negative – levels seen over the last decade or so. That means investors are once again demanding more compensation to hold long-term bonds. That partly reflects the fading of the era of ultra-easy monetary policy, with heavier government borrowing becoming a more important force in long-term bond markets.
So while some of the latest rise in yields could reverse if expectations for Fed policy change, it’s hard to imagine a return to the exceptionally low yields of the 2010s.
Fiscal pressures make higher yields more likely to persist
That brings us to one of the biggest changes in the backdrop for bond markets: the scale of US government borrowing. As the top panel in Exhibit 2 shows, net Treasury issuance has averaged roughly US$1.8 trillion a year between 2023-2025, compared with around US$800 billion after the Global Financial Crisis (GFC) and $150 billion prior to it.
Those levels aren’t unprecedented – issuance also jumped during the GFC and COVID pandemic – but they were temporary responses to economic emergencies. What stands out today is that borrowing has stayed elevated even with the economy still holding up.
There’s also little sign that borrowing will fall back toward its pre-COVID norm. The bottom panel in Exhibit 2 shows the federal deficit projected to remain at around 6% of GDP over the coming decade. That means the government will need to keep issuing debt to cover the gap between spending and revenues, pushing debt held by the public from about 101% of GDP today to a projected 120% by 2036, according to the latest Congressional Budget Office (CBO) outlook¹.
What’s more, as that debt stock grows, so does the cost of servicing it. More government debt is being refinanced at higher interest rates than in the 2010s, pushing its net interest costs from around 3.3% of GDP in 2026 to 4.6% by 2036, based on CBO estimates. Those higher interest costs, in turn, add to the government’s borrowing needs.
In short, large deficits are keeping Treasury supply high, and rising interest costs are reinforcing those borrowing needs. Together, those forces put upward pressure on long-term yields and mean higher levels are more likely to persist.
AI is adding to bond supply, but it could also improve the fiscal outlook
Government borrowing isn’t the only source of increased demand for capital. As we’ve shared in recent months, the AI buildout is also requiring increasingly large amounts of investment. (Read more about AI spending in CIO Insights: Earnings season, brought to you by AI.)
For the five major US hyperscalers – Alphabet, Amazon, Meta, Microsoft, and Oracle – capital spending expectations continue to rise in the years ahead, from around US$400 billion in 2025 to nearly US$800 billion this year, before reaching about US$1.1-1.2 trillion a year from 2027 through at least 2031.
Not all of that spending needs to be borrowed: the largest tech companies still generate substantial cash flow, while equity, leases, private credit, and project financing are also being used to fund the buildout. But bond issuance has increased sharply, as shown in Exhibit 3 below. J.P. Morgan estimates that hyperscaler and other AI-related bond issuance has climbed from just US$17 billion in 2024 to around US$280 billion in 2026, and upwards of US$350 billion in 2027².
While still smaller than US government borrowing, AI-related issuance already represents a sizable share of the corporate bond market. According to Goldman Sachs³, AI-related borrowers accounted for just 1% of US investment-grade bond issuance in 2024. That figure rose to 7% in 2025, and to around 18% as of early August of this year.
The shift is even more pronounced at the longer end of the yield curve: AI-related borrowers have accounted for roughly 40% of investment-grade issuance with maturities of 15 years or more, putting them in direct competition with long-term Treasuries for investors' capital.
To be sure, that increase is too recent to explain the broader rise in long-term yields since their 2020 lows. But sustained borrowing on this scale could add to the upward pressure on yields from government financing needs.
Faster AI-driven productivity could improve the debt outlook
If the AI investment boom ultimately delivers sustained productivity gains, it could lift the economy’s long-term growth potential (as we discussed in our 2026 Macro Outlook: Just the FACTs). Stronger growth would expand the economy and lift tax revenues, reducing the need for new borrowing. If those gains prove large enough, GDP could grow faster than government debt, bringing the debt ratio down.
The CBO already assumes that AI will add around 0.1 percentage point (ppt) a year to productivity growth in the period to 2036. Using its sensitivities as a rough guide, the implied average boost to growth per year is equivalent to about US$40 billion of additional output – or about 7% of the roughly US$600 billion of US AI-related investment estimated for 2026⁴.
That contribution could prove conservative if AI adoption spreads faster or delivers larger efficiency gains than the CBO assumes. For context, long-term returns on invested capital (ROIC) exceed 20% in chipmaking and parts of software⁵. (This comparison is only illustrative: company returns measure profits on total capital invested, while our estimate compares broader economic gains to annual investment spending.)
Exhibit 4 explores what an additional 0.3 or 0.5 ppt of annual productivity growth could mean for growth and public finances.
Under these scenarios, real GDP growth would average around 2.2% and 2.4% a year over 2027-2036, versus 1.8% under the CBO baseline. That stronger growth would help reduce borrowing needs and leave debt at around 113% and 109% of GDP in 2036, compared with 120%.
Faster or more effective AI adoption could improve the debt outlook further. But even under both scenarios, debt would still rise above the projected 2026 level of about 101% of GDP. AI could substantially ease the US’s fiscal challenge, but stabilising its debt burden would require either larger gains in growth or adjustments to government spending and revenues.
Here’s the kicker: stronger productivity could also keep yields elevated. Higher returns on investment would encourage businesses to invest and increase demand for capital, supporting higher real interest rates. And those higher borrowing costs would offset some of the fiscal benefit from faster growth. Even so, AI could improve the fiscal outlook while still contributing to a higher-rate environment.
What higher yields would mean for portfolios
The investment implications that follow depend on which force dominates:
- If AI investment lifts productivity and economic growth, long-term yields can stay elevated alongside stronger nominal earnings growth. That would still be a constructive backdrop for equities, even if higher yields put some pressure on valuations.
- The less favourable outcome is one where long-term yields stay elevated because of persistent inflation, large fiscal deficits, and a higher term premium, without a comparable improvement in growth. That would be tougher for equities because valuation pressure would not be offset by stronger earnings growth. Meanwhile, long-duration bonds would remain vulnerable to a further rise in yields.
Our base case is that both forces remain in play: AI supports stronger productivity and earnings, while large deficits and rising debt keep fiscal pressure elevated. Markets are therefore likely to move between those two narratives depending on which force is dominating at a given time. Exhibit 5 summarises how the investment implications differ under each scenario.
Even if AI delivers stronger productivity and growth, US deficits are still likely to remain wide and debt to keep rising relative to GDP. That could keep the term premium elevated and leave long-duration bonds vulnerable, even as strong earnings should continue to support equities.
The broader takeaway is that higher yields aren’t inherently negative for investors. Stronger growth can come alongside persistent fiscal pressures, and vice versa – and the forces pushing yields higher can change over time.
What’s most important is staying diversified across asset classes, so your portfolio isn’t overly dependent on either outcome as the balance between growth and fiscal risk continues to shift in the years ahead.
Authors
Stephanie Leung, Chief Investment Officer
Stephanie and her team oversee the full spectrum of investment products and portfolios offered at StashAway. She brings more than two decades of investment expertise across multiple asset classes. Prior to joining StashAway in 2020, she managed investment portfolios at institutions such as Goldman Sachs and multi-billion dollar family offices in the region.
Justin Jimenez, Head of Macro and Investment Research
Justin brings nearly 15 years of experience in economic and investment research to StashAway, where he contributes to shaping the investment office’s views on the global economy and financial markets. Before joining StashAway in 2022, he was an economist at Bloomberg, and holds degrees in international economics and finance from Columbia and UCLA.
Jim Tai, Head of Quantitative Research and Portfolio Management
Jim brings nearly two decades of quantitative research and systematic trading expertise to StashAway. Having managed quantitative strategies across leading financial institutions in Hong Kong and New York, he holds advanced degrees in Applied Mathematics and Engineering from Columbia and Princeton.
Glossary
Term premium
An estimate of the additional return investors demand for holding a long-term bond instead of repeatedly investing in shorter-term securities. It compensates investors for uncertainty around future interest rates and inflation.
Real yield
A bond yield after removing compensation for expected inflation. In this article, we use the yield on 10-year Treasury Inflation-Protected Securities as a proxy.
Duration
A measure of how sensitive a bond’s price is to changes in interest rates. Bonds with longer duration generally lose more value when yields rise and gain more when yields fall.
Inflation expectations
The rate at which households, businesses, and investors expect prices to rise in the future. In financial markets, these are often approximated by the breakeven inflation rate – the difference between the yields on conventional and inflation-protected Treasury bonds of the same maturity.
Net Treasury issuance
The amount of new Treasury debt issued beyond what is needed to replace securities that mature. It therefore measures the additional Treasury supply that investors must absorb.
Net interest costs
The interest the federal government pays on its debt, after subtracting the interest income it receives. These costs rise when the amount of debt increases or when debt is refinanced at higher rates.
References
- Congressional Budget Office. (2026). The Budget and Economic Outlook: 2026 to 2036. Available at: https://www.cbo.gov/publication/62105
- Jackson, J. and Liu, J. (2026). Can credit markets absorb the AI buildout? J.P. Morgan Asset Management. Available at: https://am.jpmorgan.com/wr/en/asset-management/institutional/insights/market-insights/market-updates/on-the-minds-of-investors/can-credit-markets-absorb-the-ai-buildout/
- Goldman Sachs. (2026). How AI Debt Is Reshaping Credit Markets. Available at: https://www.goldmansachs.com/insights/goldman-sachs-exchanges/how-ai-debt-is-reshaping-the-credit-market
- Goldman Sachs. (2026). Global AI Investment Is Forecast to Exceed $1 Trillion in 2026. Available at: https://www.goldmansachs.com/insights/articles/global-investment-is-forecast-to-exceed-1-trillion-in-2026
- Damodaran, A. (2026). Return on Capital by Sector (US). New York University Stern School of Business. Available at: https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/roc.html