AI's Market Path Will Get Bumpier

The US economy and its financial markets are in a generational race whose implications extend well beyond the prospects for future prosperity and financial stability. The race is between AI and related technologies that hold massive potential to unlock productivity and earnings growth, on the one hand, and the unintended consequences of these technologies, on the other.

Most of us need AI to unlock those significant productivity and earnings gains, especially as the economy carries weighty legacy issues. But while the race has been remarkably smooth so far, it is likely to become far more volatile.

Count me among those who are genuinely excited about what’s happening in the AI revolution and its potential interaction with equally exciting advances in the life sciences, robotics, and eventually quantum computing. By enhancing the input that goes into so much of what we do—intelligence—AI’s economic potential is enormous. It is, as US Federal Reserve Chair Kevin Warsh framed it in his Jackson Hole speech last month, potentially a “new factor of production.”

Google’s James Manyika has gone further in conversations with me and elsewhere. AI is not just a “general purpose technology” (like electricity) that can improve how we do what we do today. With recursive self-improvement and other advances, it is also increasingly an “invention machine for inventions” that will allow us to do even better things. Such observations have fueled the promise of massively higher corporate earnings, which in turn has driven stock-market indices to record highs and channeled a massive amount of equity and bond financing to the tech sector.

Nowhere is this dynamic more vividly demonstrated than in Nvidia’s staggering earnings last month. Its eye-popping revenue growth continues to validate the size of the sector’s capital expenditures. The same dynamic is also apparent in Anthropic’s IPO preparations, where, like SpaceX, the “total addressable market” it hopes to tap is enormous.

Yet my excitement comes with concerns, not so much about the longer-term risks of shock events or “extinction” (though that is obviously important), as about AI’s more immediate unintended consequences and their collateral damage. Fears of the latter are already spilling into the political and social arena and could well affect the pace and depth of the technology’s adoption.

These changing conditions carry the risk of crowding-out effects, especially in the short term, owing to a classic intertemporal mismatch. The potential longer-term benefits of AI could address many of our current problems, from low growth and high debt to climate change and extreme inequality; but more immediately, AI’s development and deployment is being complicated by issues that, if we are not careful, could make these very problems worse.

Consider what has been happening in the government and corporate bond markets. In a world where deficit financing and traditional corporate borrowing already claim a large share of available bond funding, the massive increase in hyperscalers’ demand for credit has been a factor in pushing borrowing costs, including mortgage rates, higher.

This is especially the case now that historically reliable large buyers of US securities seem less inclined, owing to domestic considerations (in the case of Japan and Gulf countries) and geopolitics (China and Canada). Moreover, in certain parts of the United States, incremental energy demand has contributed to higher electricity prices, further exacerbating a mounting multiyear affordability crisis that falls hardest on lower-income households.

The tension between the march of AI and the need to manage its immediate consequences is not the only risk. Another is the mismatch between exciting, rapid AI advances and uneven adoption rates across the economy. A particularly insightful and comprehensive study from Google shows that while diffusion has been very broad, it has been rather shallow. Because AI has yet to penetrate deeply into workflows, much of its productivity potential remains untapped.

All this suggests that the AI trajectory—which encompasses advances in science, adoption across the economy, and mitigating the risks—is likely to become more challenging. That is why those leading the revolution are looking to policymakers to play a larger role in facilitating adoption and minimizing collateral damage. Given the dynamics at play, such interventions would require agile public-private partnerships. For example, Demis Hassabis, the founder of Google DeepMind, recently proposed a Financial Industry Regulatory Authority-like arrangement whereby a cooperative private body would wield publicly delegated regulatory powers in consultation with government agencies.

So far, the financing of AI (which is US capital markets’ superpower) has been relatively smooth. Equity investors have been able to make money from intensifying promises of future-wide economic prosperity. But the road is about to get bumpier, especially now that both major political parties are leaning into the populist side of the debate. Meanwhile, the group of economists and behavioral scientists working on AI and studying the tensions mentioned above is still too small, given what is at stake.

The era of relatively friction-free financing and seemingly effortless market gains may soon give way to one marked by stronger, multi-directional headwinds. Navigating this challenging terrain will demand far greater agility and collaboration between governments and tech firms, both domestically and internationally.

AI has the power to solve some of our major economic challenges and unlock previously unimaginable levels of growth. But if it is mismanaged, it could bring damaging costs.

Project Syndicate