Artificial intelligence and economic growth: market reaction to the release of new AI models

Since the end of 2022, the markets have been euphoric: each new generation of AI models causes bouts of optimism, and the quotations of AI-related technology companies are growing by tens and hundreds of percent. But beyond the hype and flashy headlines in the media and investor communities, there is another, more objective indicator - the behavior of the government and corporate bond market. Economist Denis Elakhovsky has studied the research and found out that after the releases of the largest AI models, long-term rates are falling - as if investors are laying down not for accelerated growth, but for risks and stagnation.

The real attitude of the capital market to the prospects of artificial intelligence technology is much more wary than it might seem by looking at the stock prices of the leaders of the AI race, argue Isaiah Andrews and Maryam Farboudi of MIT in their study. They examined how long-term bond yields in the US changed before and after the releases of new versions of major AI models in 2023 and 2024. The event list covers all the major announcements from the top players: OpenAI (GPT-4), Google DeepMind (Gemini), Anthropic (Claude 3), xAI (Grok), and DeepSeek (R1).

The method of analysis is a classical event study. Andrews and Farboody studied how the yields of government bonds (10-, 20- and 30-year Treasuries), corporate bonds and TIPS (inflation-protected bonds) changed for 15 trading days before and after the release of the model. It turned out that after the release of AI-models yields of long-term bonds decreased by 12-15 basis points. Moreover, the decline started 2-3 days before the official release: probably due to leaks or preliminary access to the models by experts and investors. The effect persisted for up to two weeks.

Long-term bond yields are an indicator of market expectations about economic growth, inflation and risk. If the market believes in accelerating growth, especially if it is accompanied by accelerating inflation, yields tend to rise. When there are expectations of economic crisis, the dynamics of yields become more complex. They usually decline. But in the case of existential risks, such as threats of uncontrolled development of AI (for example, as in the movie "Terminator"), the opposite effect is possible - yields may rise if the market allows the probability of a catastrophic scenario, but does not consider its realization inevitable.

In other words, Andrews and Farboody's study shows that institutional investors do not yet believe that AI will bring economic prosperity in the coming years. Rather, on the contrary: they see risks of stagnation. There is perhaps only one piece of good news here: borderline scenarios like a "machine uprising" are also assessed as unlikely. Otherwise, all current bond pricing models will cease to be applicable in principle.

A legitimate question arises: why is the stock market reacting strongly to the releases of new AI models, while the bond market is radiating pessimism? The answer may be hidden in investors' goal-setting. Stocks react to the earnings of individual companies - NVIDIA, Microsoft, Google, Amazon, Meta, Palantir and others. Bonds, on the other hand, reflect expectations for the entire economy: growth in consumption, inflation, employment and GDP.

In other words, the market accepts that individual AI companies will make super profits, but this will not transform the economy as a whole. Moreover, it may even harm the economy if AI replaces workers faster than it creates new industries.

This hypothesis is supported by empirical data. In a study published in Science in 2023, it is shown that using ChatGPT in office tasks can reduce turnaround time by 40% while improving its quality by 18%. This explains why the capitalization of individual AI developers and infrastructure providers is growing. But it doesn't explain why it should accelerate the economy as a whole.

The study draws several conclusions at once that should be considered in your strategy.

First, do not overestimate the growth of the economy due to the introduction of AI. The bond market is a mirror of professionals' expectations. If long-term rates are declining, it means that we are not talking about a really noticeable technological revolution from a macroeconomic point of view. It may be risky to build your personal investment portfolio on the assumption that AI will multiply GDP growth in such conditions.

Secondly, it is worth looking for new AI-related investment ideas not in broad market indices, but in specific sectors. AI does generate profits, but locally - in chip manufacturing, cloud platforms, model development. It's not yet a mass-market story like the internet or smartphones were in their day. Betting on the entire market, expecting it to grow due to AI empowerment, may not be the most effective solution. It would be much more sensible to place part of the portfolio in securities related only to the development of AI infrastructure.

For example, last year's State Street report emphasized that the experience of introducing AI into business processes leads to a dramatic increase in productivity. This allows us to hope that we can expect productivity gains at the macro level as well, but this is a slow transformation that the market may not be betting on yet.

Thirdly. You should never ignore the dynamics of bonds. For example, a decline in yields may indicate that investors are not waiting for growth, but for a recession with the disinflation that it is supposed to bring. If such a scenario is realized, even the most successful, powerful and productive AI models will not be able to bring their creators as much profit as they could generate in a steadily growing economy.

Dollar. In addition to the fall in bond yields after the releases of AI models, the study also records a slight weakening of the dollar. This fits into the general logic: lower yields increase the outflow of capital from the U.S., and this already leads to a depreciation of the exchange rate.

Spreads. If the market perceived AI as a risk mitigation mechanism, for example by increasing the productivity of companies, then corporate spreads (the difference between corporate and government bond yields) should be narrowing. But this is not happening. So, for now, the evolution of AI models is not convincing investors that it will allow businesses to significantly improve their performance.

Inflation. A similar effect is seen in the yields of TIPS (inflation-protected bonds). This suggests that the market is not expecting a jump in inflation due to AI - on the contrary, the effect is more likely to be deflationary.

Perhaps the main practical conclusion from this study can be formulated as follows: technological progress should not be confused with economic recovery. AI is indeed capable of transforming entire industries, but its implementation may be slow, the economic benefits are unevenly distributed, and the risks (from substitution of human labor to social upheaval) are great. And yes - rising stock prices do not always mean rising consumption and employment. In general, all these considerations make the capital market feel more doubtful than hopeful about AI development.

In addition, regulators are increasingly warning about the risks. The Bank of England said in April 2025 that the widespread use of one-size-fits-all AI models by traders could lead to market crashes as AI learns to replicate stressful scenarios in search of profits. The IMF last year warned: AI increases the efficiency of markets but makes them more volatile and increases the concentration of risk around a small number of AI service providers.

Study the behavior of rates ahead of key AI events. This can be a leading indicator of capital flows. Evaluate AI companies separately from the macroeconomy. Their stock growth is still a story about market share, not GDP growth. Watch spreads: if they start to narrow, it will be a sign that AI is starting to be perceived as a factor of stability rather than risk.

AI is changing the world, that's for sure. But exactly how it is changing it is still unclear. In the midst of this fog, it is always better to keep as many benchmarks in sight as possible. And the bond market should definitely be among them.

This article was AI-translated and verified by a human editor