Cambridge, Massachusetts. How are bond markets responding to the relentless advances in artificial intelligence? Economists at the Massachusetts Institute of Technology (MIT) say the answer has been unexpectedly striking.
Between 2023 and 2025, bond prices rose on average with each new development in artificial intelligence – whether a fresh version of a chatbot or another technological milestone – while yields declined. The pattern held across long-term government bonds, inflation-protected government bonds and corporate bonds alike. The response was not merely short-lived, but appeared to be lasting.
Researchers at MIT express surprise because the prevailing narrative suggests that artificial intelligence will generate broad-based economic growth. Under conventional assumptions, that prospect would be more consistent with rising long-term bond yields.
At the same time, market behaviour does not indicate that investors expect artificial intelligence to reshape the economy so fundamentally that it would, for the first time in history, overcome the longstanding problem of scarcity – the economic principle that human wants are unlimited while the resources available to satisfy them are finite. Elon Musk has also argued that artificial intelligence could bring scarcity to an end.
In short, bond markets do not appear to believe that artificial intelligence will trigger extreme and pervasive disinflationary or deflationary pressures.
What the new Federal Reserve chief has in mind
However, Kevin Warsh, the likely next head of the United States central bank, is betting on strong disinflation driven by artificial intelligence. He believes that the surge in productivity brought about by artificial intelligence will be so pronounced that it will lead to a marked easing of inflationary pressures, enabling the central bank to lower its key interest rates – as United States President Donald Trump would prefer.
Bond markets, however, appear to be signalling a different scenario. Investors seem to expect that artificial intelligence will deepen inequality, with consequences that dampen overall economic growth. In certain sectors, productivity may rise sharply – perhaps to an unprecedented degree – resulting in above-average profitability, primarily for leaders in artificial intelligence such as Amazon, Alphabet, Meta and, to a lesser extent, Microsoft. From that perspective, the vast sums those companies are currently investing in artificial intelligence infrastructure and data centres may well be justified.
Yet artificial intelligence may weigh on the broader economy. There is a tangible risk of large-scale job losses and rising unemployment. Such developments would constrain household consumption to an extent that materially slows aggregate growth. The outcome would be pressure in the form of so-called bad disinflation – or even bad deflation – not driven by technological progress, productivity gains and the resulting reduction in unit production costs, but by a more generalised slump in demand.
Two countervailing forces would therefore act against any pro-inflationary impulse arising from stronger growth and higher profitability enabled by artificial intelligence. The first is the pressure of bad disinflation – or even deflation – stemming from weakened demand. The second is the pressure of good disinflation – or even deflation – resulting from lower unit production costs. According to that interpretation, the former is likely to dominate.
The call for redistribution
One part of the economy may benefit enormously from artificial intelligence, while another may be seriously harmed by it. Those on the losing side are likely to demand more extensive redistribution from the ‘AI winners’ to the ‘AI losers’, a call that could well attract political backing. At the same time, pressure may mount for the broad introduction of measures such as a universal basic income.
The segments of the economy displaced by artificial intelligence could weaken growth to such an extent that bond markets no longer respond to advances in the technology as one would expect if it were set to unleash clear, robust and widespread economic expansion.
At the same time, bond market behaviour suggests that the gains in productivity and output generated by artificial intelligence may not be sufficient, in aggregate, to offset the economic burden of demographic ageing.
Artificial intelligence may represent a significant technological breakthrough in the eyes of bond investors, but not one capable of ‘feeding’ a newly emerging cohort of unemployed workers and an expanding retired population. Such a development would depress overall consumption to a degree that strengthens the previously described pressure of bad disinflation – or even deflation – driven by weak demand.
The larger the pool of unemployed, the stronger the political pressure for substantial redistribution from the ‘winners’ to the ‘losers’ of artificial intelligence. That pool could expand further if current patterns of immigration persist.
It is for that reason that, at this year’s World Economic Forum in Davos, figures such as Larry Fink, head of BlackRock, and Alex Karp, head of Palantir Technologies, argued that countries which have not embraced large-scale immigration are better positioned for the age of artificial intelligence.
Both men are plainly among the ‘AI winners’. They may already recognise that sustained mass immigration could translate into stronger political demands in future to share the profits generated by artificial intelligence with those who lose out. Their comparatively restrictive tone on immigration in Davos would have been far less likely only a few years ago.
This text, which has been shortened, was originally published on the website lukaskovanda.cz.