The Bull Case Against the AI Bubble Thesis: The Productivity J-Curve

The strongest counterargument to an AI bubble thesis is not that share prices have risen. It is that new technology can require substantial complementary investment before its benefits become visible. Early costs may be easier to measure than future organizational improvements.

What the J-curve argument says#

NBER's overview of generative-AI economics discusses research on the productivity J-curve: training, workflow redesign and other complementary investments can precede observable productivity improvements. The argument explains why early measured results can understate potential benefits. It does not establish a guaranteed payoff date.

How to turn the bull case into a test#

For a business adopting AI, identify a measurable operating result: time to complete a task, errors requiring rework, customer retention or cost per completed transaction. Track the result after including implementation and review costs.

A model completing work faster is not enough if employees spend the saved time checking unreliable output. Conversely, a modest task-level gain can be valuable when applied consistently across a large workflow. The evidence needs to describe the actual operation.

Ask who keeps the productivity gain#

Suppose a hypothetical service provider lowers cost per unit from $10 to $8. If competition drives the selling price down by the same $2, customers receive the direct benefit and the provider's unit margin is unchanged. If pricing holds, the provider captures more of the gain.

This is why broad economic progress cannot be translated mechanically into a stock-price target. Ownership of the benefit is part of the investment thesis.

Make the argument falsifiable#

Set milestones for paid adoption, renewals and operating outcomes. Revise the thesis if those milestones repeatedly move farther away without supporting evidence.

The J-curve offers a credible reason not to dismiss AI because benefits lag spending. It should be a framework for patient measurement, not an excuse to treat every disappointment as proof that success is just around the corner.

More in this analysis series#

Start with AI stocks versus railroad mania. Continue with AI Picks-and-Shovels Stocks: Are Suppliers Safer in a Bubble?.

Analysis published October 1, 2026. Historical comparisons and hypothetical examples are not forecasts. Educational information, not investment advice.