Stocks affected by the AI boom will correct in the short term, given that productivity increases translate to leisure rather than more output and earnings

Assessment as of September 26, 2026

The claim combines three propositions: AI saves work time; workers take the savings as leisure instead of producing more output and earnings; and AI-affected stocks therefore correct in the short term. Evidence is strongest for the first proposition and a context-dependent version of the second, not for the proposed market consequence.

A representative Korean worker survey found that 51.8% used generative AI for work, with users reporting a 3.8% reduction in working time. Its reported correlation between time savings and output changes was close to zero, and the authors estimated a 1.3-percentage-point increase in the share of time users spent on on-the-job leisure. This is direct support for the proposed mechanism, but it is survey evidence about workers, not observed corporate earnings or stock returns. [1] An international evidence review likewise finds that task-level gains have often not appeared in measured output or earnings, while emphasizing variation across tasks, workers and implementation. [3] In a randomized experiment across 66 firms, users spent less time on email and working outside regular hours, without a detected change in task quantity or composition from individual access to the tool. [4]

There is an important contrary observation: Danish workers reported reallocating roughly 80% of chatbot-related time savings to other job tasks, and fewer than 10% to additional breaks or leisure, even though the study found little aggregate effect on earnings or hours. No earnings gain is not the same finding as leisure absorbing most savings. [2] A separate customer-support study found 14% more issues resolved per hour on average with an AI assistant, demonstrating that at least some AI use increases measurable output. [7]

The financial link is also two-sided. Microsoft warns that disappointing adoption or underused AI infrastructure could prevent expected investment returns. That identifies a plausible route from insufficient monetization to weaker earnings, without attributing such an outcome to worker leisure. [8] Conversely, NVIDIA had already reported second-quarter fiscal-2027 revenue of $96.2 billion, up 106% year over year, including $89.0 billion of Data Center revenue, up 117%. Those are realized supplier sales—not a forecast and not proof that all customers' investments are profitable—but they challenge a blanket claim that the boom has produced no earnings. [6]

ECB economists judge a correction of elevated stock valuations likely and note that it could happen even if AI succeeds. Their proposed mechanisms include changing risk premiums; they also say the timing is unknowable. Their warning therefore supports correction risk, but not this claim's particular leisure-to-earnings-to-prices causal chain. [5] Over the requested 2–5 years from September 2026, a correction remains possible; the word short term does not identify a testable point within that horizon. The available evidence does not justify predicting that such a correction will occur specifically because productivity gains become leisure rather than output and earnings.