Overview
Economists caution that an earnings bust in artificial intelligence (AI) would not automatically trigger a broader macro‑economic downturn; the overall effect depends on the underlying cause, with some scenarios merely shifting income between AI producers and users rather than delivering a net hit to global growth.
Trigger Scenarios
Jennifer McKeown, chief global economist at Capital Economics, outlines three possible triggers for an AI earnings bust. The first scenario involves declining AI model prices driven by intensified competition, which would pressure providers’ margins but could also spur wider adoption. In this case, income would shift from AI producers to AI users, especially in financial services and healthcare, resulting in a redistribution of earnings within the U.S. economy.
The second scenario centers on rising electricity or commodity prices that squeeze AI firms’ profitability. This would also redistribute income, benefitting energy exporters such as the Gulf states and Norway, as well as metals exporters like Chile and Australia, while the United States could experience a pullback in data‑centre and cloud‑spending. McKeown notes that the effect on global activity would likely be modest.
The third and most macro‑economically risky scenario is weaker‑than‑expected AI demand. A pullback in investment would directly hit semiconductor manufacturing, cloud computing, data‑centre construction, electrical equipment and power generation. The economies most exposed under this scenario are those that have benefited most from the AI investment boom—namely Taiwan, Korea, Mexico, China and the United States. Europe, by contrast, has relatively little near‑term exposure due to its limited role in the AI infrastructure build‑out.
Severe Permanent Demand Weakening
If AI demand were to weaken permanently because of a fundamental reassessment of AI’s technological potential, the repercussions would extend beyond the AI supply chain. Lower productivity growth would reduce potential output across advanced economies, with the United States, United Kingdom and Switzerland identified as the biggest long‑run losers, given they stood to gain the most from AI adoption.
Sectoral Implications
Cheaper AI models could still support semiconductor demand in chip‑producing economies such as Taiwan and Korea, even as they diminish investment incentives for producers. Conversely, higher input costs could benefit energy and metals exporters while curbing U.S. data‑centre spending. Overall, the redistribution effects dominate the first two scenarios, whereas the third scenario poses a broader risk to global growth through reduced productivity.