Fed Study: AI's Missing Productivity Boom Looks Familiar-But the Stock Market Can't Wait

AI: no aggregate productivity gain yet, but expectations are already moving

The macro data still shows no broad productivity payoff from AI, even as markets price in a much larger future. St. Louis Fed research scanning nearly 490,000 earnings calls found no measurable bump in aggregate productivity from AI so far. At the same time, AI's share of productivity discussion rose from near zero to roughly 15% by the end of 2025, and about 95% of those AI-related comments described gains expected in the future, not already realized. The result is a clear tension: the economy-wide evidence is still thin, while corporate commentary already sounds confident that the payoff is coming.

Why the gap matters for investors

That gap is no longer just academic. AI spending has helped support the broader expansion, contributing to earnings growth and headline GDP. That is constructive, but it also raises the stakes. If leadership remains concentrated in a handful of large companies, investors cannot wait for every firm to validate the AI productivity case. They have to price the gap between expectation and evidence while market leadership stays narrow.

The upside is easy to see: roughly $27 trillion in AI-related market value has already been created since late 2022. The risk is straightforward as well: if broad productivity still does not materialize, investors may be paying too much for gains that are still only prospective.

Why a real productivity wave can arrive late

The Fed's sequence matches the historical pattern

The missing boom may simply be too early to see in aggregate data. The Fed's framework separates the AI buildout into three stages: capabilities and costs, then firm investment and adoption, and only afterward measurable aggregate productivity gains. That sequence is consistent with how major technologies have diffused in the past. Historical evidence on technology development and diffusion suggests it can take 20 to 30 years for the broadest economic effects to show up.

In practical terms, cheaper tools do not immediately reshape work. They first have to be purchased, integrated into existing processes, and learned by teams that still need new workflows, oversight, and coordination.

Some gains may be real but hard to measure

There is also a measurement problem. The St. Louis Fed research argues that when AI makes some output radically cheaper to produce, that output can simultaneously become less valuable. If the task gets easier but the market price of the result falls, the gain can be masked in aggregate statistics. That helps explain why real improvements at the margin can still disappear in economy-wide data.

The first signs are showing up in specific sectors

The effects are not completely absent. High AI-exposed sectors have posted 3.7% annualized productivity growth since early 2024, versus 1.7% for the rest of the economy. That does not prove AI is fully paying off across the broader economy, but it does suggest the first signals are appearing where exposure is strongest.

For investors, that is the setup to watch. The payoff pattern looks consistent with familiar technology diffusion stories, yet the earliest positive signals are already visible in selected areas. The key question is whether those gains broaden over time before expectations get too far ahead of the data.

The market now: is AI a real productivity wave or an expensive buildout?

Once the productivity delay is accepted, the debate shifts. The question is less whether AI is real and more whether today's valuation already assumes too many winners and too large a share of future profits. The market has assigned roughly $27 trillion in market value to AI-related companies since late 2022, and US tech investment as a share of GDP has surpassed its 1990s peak. That raises the burden of proof. Investors are not just betting that AI will help somewhere; they are betting that the buildout will translate into durable profits for a recognizable set of businesses.

The bullish case: some companies are already converting AI use into cash flow

The bullish case is not only about narrative. In a market this large, investors need evidence that AI is moving beyond buzzwords. Morgan Stanley found that adopters delivering measurable results are seeing cash flow margin expansion at roughly 2x the global average. That is the clearest signal markets care about: not hype, but better cash conversion.

Sector selection also matters more than the broad AI label. High AI-exposed sectors have posted 3.7% annualized productivity growth since early 2024, versus 1.7% for the rest of the economy. That supports a selective view of AI returns rather than a blanket assumption that every part of the economy will benefit equally.

The bearish case: spending may outrun the profit pool

The bearish case focuses on the other side of the equation. If investment is already above 1990s levels and still accelerating, then current valuations may require fairly optimistic assumptions to work. Goldman Sachs has warned that investors could be overestimating how long above-normal profits will last, especially for companies supplying AI infrastructure if the economic gains end up spread across many participants rather than captured by a narrow group of winners.

That makes the investment debate less about owning the AI theme in general and more about identifying the part of the value chain that can retain more of the economic benefit as adoption broadens.