Artificial intelligence has emerged as a significant macroeconomic force. To appreciate the magnitude, consider the numbers: global AI investment is projected to surpass $1 trillion by 2026, with more than $500 billion allocated in the US alone. These amounts equate to 1.8% of US GDP and 0.9% of global GDP, rising to 2.8% and 1.4%, respectively, by 2028.

In terms of growth contribution, technology and data-centre activity are expected to account for 36%–50% of US year-on-year expansion this year. On a global scale, the direct AI investment impulse represents approximately 10% of overall growth.

From a financial markets perspective, however, analysis of AI has largely been confined to equities or private credit. Equity markets remain preoccupied with the microeconomic dimensions of artificial intelligence—corporate margins, capital expenditure cycles and labour substitution—and their implications for sector positioning, valuation and concentration risk. In private credit, the focus is on the risk profile and pricing of AI-related investments.

By contrast, sovereign debt markets are narrowly focused on supply dynamics such as issuance, deficits and quantitative tightening. Broadly speaking, the market prices the AI productivity gain, but not the fiscal relief that should come with it.

Binary Outcomes
Let us examine whether this is emerging as a profound blind spot in macro pricing, and whether it is time to commit capital to a long forward-to-forward trade in US yields.

--Channels of pricing

Artificial intelligence presents a structural uncertainty for sovereign debt markets, with outcomes that could diverge sharply depending on how productivity gains interact with fiscal dynamics.

--The fiscal dividend (bull case)

If AI-driven total factor productivity outpaces historical trends, GDP growth accelerates, organically expanding the tax base while cyclical welfare spending declines. Structural deficits narrow, long-term Treasury issuance falls, and the result is a powerful rally at the long end of the curve.

A positive productivity shock enhances tax revenues disproportionately relative to the underlying GDP gain, thereby improving the present value of future primary surpluses. In this sense, bondholders are effectively positioned to benefit from AI-driven productivity.

If artificial intelligence delivers a genuine and sustained productivity shock, the traditional calculus underpinning long-term US Treasuries could be transformed in ways that current pricing does not anticipate.

--The displacement burden (bear case)

Conversely, if labour-market dislocation outpaces economic absorption, governments will face intense political pressure to fund expansive safety nets—ranging from universal basic income to healthcare adjustments and retraining programmes. In this scenario, productivity gains accrue disproportionately to capital, while fiscal burdens shift directly to the state, undermining debt sustainability and intensifying supply pressures.

If AI fails to deliver a sustained productivity uplift—or if gains plateau at a stable but modest level—the fiscal channel turns adverse. Tax revenues rise only in line with GDP, or even lag behind if displacement pressures increase welfare spending. In this environment, the present value of future primary surpluses does not improve; it may deteriorate.

Historical Trends
Carlota Perez, in her 2002 book Technological Revolutions and Financial Capital, argues that technological revolutions occur in roughly 50-year cycles. She identifies five such episodes, from the Industrial Revolution in Britain in 1771 to the information and telecommunications revolution in the US in 1971.

In many respects, the artificial intelligence revolution—arguably inaugurated on November 30, 2022, with the launch of ChatGPT, nearly 50 years after the dawn of the Information Age—makes her work appear strikingly prescient.

During the 1970–1980 information and telecommunications revolution, bonds delivered steady nominal returns at a compound annual growth rate of approximately 7%–8%, but inflation meant equities captured the real wealth gains.

This historical parallel is useful when considering how AI-driven productivity shocks might eventually reshape both equity and sovereign debt markets. If history rhymes, the long end of Treasuries may be the most mispriced asset class once AI-driven productivity translates into fiscal repair.

Channels of Transmission
There are two principal channels of transmission. The first is a macroeconomic channel, encompassing potential growth, inflation dynamics and the fiscal position, with spillover effects for the rest of the world. The second is a microeconomic channel, reflecting the impact of AI financing on the structure of fixed income markets.

The latter has gained prominence recently, as market participants debate whether AI-related debt issuance is competing with US government securities and thereby contributing to upward pressure on yields.

Macroeconomic channels capture how long-term interest rates discount fundamental economic variables, while microeconomic channels reflect how the AI funding model—its scale and reliance on self-financing versus debt issuance—shapes the pricing of duration.

These channels operate with different lags, and their effects may shift direction over the course of our 10-year forecast horizon. Broadly, macro channels influence the expected-rate component of long-term yields, whereas micro channels act primarily on the term premium. The notable exception is the fiscal position, which, despite being a macro channel, exerts a direct influence on the term-premium component of long-term government bond yields.

The market is pricing the “growth and rates” side of AI but not the “fiscal repair” side. That could mean either scepticism about the political economy of capturing AI’s benefits or a darker narrative in which AI raises productivity but also increases fiscal burdens through displacement.

The market’s reluctance to price fiscal gains is not really about whether AI can deliver productivity gains, but whether political institutions can translate them into fiscal repair.

Political Economy
History shows that every technological revolution creates enormous potential for fiscal relief—higher GDP, broader tax bases and lower relative deficits. But whether that potential is realised depends on political economy:

• Institutional efficiency: Without structural reform, productivity windfalls are often dissipated through immediate tax cuts or new spending commitments.
• Political capital: Capturing fiscal gains requires governments to resist populist pressures and channel those gains into debt stabilisation rather than redistribution.
• Historical precedent: The IT revolution boosted productivity in the 1990s, but fiscal repair only materialised when paired with political consensus on deficit reduction. Conversely, the oil boom of the 1970s generated huge revenues but was neutralised by inflationary spending.

This is why markets are reluctant to price the fiscal dividend of AI. Investors know that technology alone does not fix debt trajectories—it takes disciplined policy. Without that, sovereign debt markets assume that any windfall will be neutralised before it stabilises long-term issuance.

The unpriced fiscal relief channel suggests potential value in the long end of Treasuries—but only if you believe the US will successfully harness AI to stabilise its fiscal path rather than exacerbate social costs.