Do Markets Believe in Transformative AI?

When OpenAI released GPT-4 in March 2023, did it send shockwaves through financial markets? According to a new NBER working paper by Isaiah Andrews and Maryam Farboodi, the answer is yes—but perhaps not in the way many would expect.

The researchers examined US bond yields around major AI model releases in 2023-2024, focusing on how markets priced long-term Treasury bonds, Treasury Inflation-Protected Securities (TIPS), and corporate bonds. What they found was striking: economically large and statistically significant movements concentrated at longer maturities, with the median and mean yield responses across releases being negative.

In plain terms, this means that long-term interest rates fell and remained lower for weeks following major AI announcements. This pattern held across different types of bonds, suggesting a broad market response rather than isolated reactions to specific instruments.

To understand why this matters, consider what interest rates represent. They reflect investors' expectations about future economic growth, inflation, and risk. When rates fall, it typically signals that investors expect slower growth or are seeking safer investments.

The authors interpret these findings through the lens of a simple, representative agent consumption-based asset pricing model. The downward pressure on yields corresponds to two potential narratives:

First, investors may be revising downward their expectations for future consumption growth. This could reflect concerns that AI, while transformative, might disrupt existing economic structures in ways that temporarily reduce overall economic output.

Second, and perhaps more intriguingly, the results suggest investors are adjusting their assessment of extreme outcomes—both existential risks (the possibility that transformative AI could pose catastrophic threats) and the arrival of a post-scarcity economy (where AI eliminates scarcity so completely that traditional economic measures become less relevant).

Interestingly, the researchers found that changes in consumption growth uncertainty did not appear to drive their results. This suggests that markets aren't simply becoming more uncertain about AI's impact, but are making more concrete assessments about its likely trajectory.

These findings challenge some common narratives about AI's economic impact. While many tech enthusiasts and investors have been bullish on AI's potential to supercharge economic growth, bond markets appear to be telling a different story—one where the transformative potential of AI is accompanied by significant uncertainty and risk.

The paper's methodology is particularly noteworthy. By focusing on discrete events (major AI model releases) and examining market reactions in the immediate aftermath, the researchers isolate the signal from the noise. This approach helps address the challenge of separating AI-related market movements from broader economic trends.

However, the study does have limitations. The sample size is relatively small, covering only a handful of major AI releases in a single year. Additionally, the interpretation of market movements is inherently speculative—while the researchers provide plausible explanations, alternative narratives cannot be ruled out entirely.

What makes this research particularly relevant is its attempt to bridge two worlds that often seem disconnected: the breathless optimism of AI development and the sober analysis of financial markets. While venture capitalists and tech companies pour billions into AI development, bond markets—representing the collective wisdom of institutional investors managing trillions in assets—appear to be signaling caution.

This disconnect raises important questions about how we should think about AI's economic impact. Are financial markets being overly conservative, failing to appreciate AI's transformative potential? Or are they correctly identifying risks and uncertainties that enthusiasts are overlooking?

The answer likely lies somewhere in between. AI almost certainly will transform the economy in profound ways, but the path to that transformation may be more turbulent and uncertain than many realize. The bond market's reaction suggests that investors are grappling with this complexity, balancing the potential for revolutionary change against the risks of disruption.

As AI development continues to accelerate, understanding how markets interpret and price these changes becomes increasingly important. The findings from Andrews and Farboodi suggest that even as AI capabilities advance rapidly, the economic implications remain deeply uncertain—and that uncertainty is being reflected in the most fundamental of all economic indicators: interest rates.