Kevin Warsh, Donald Trump�s nominee to lead the Federal Reserve, wants to lean into an AI-driven productivity boom to justify faster rate cuts – echoing Alan Greenspan�s famous 1996 call to hold rates amid a then-misunderstood tech surge. The question for markets and boardrooms: Is the AI lift already showing up where it counts, or is Warsh front-running a payoff that�s still years out?
The pitch: Productivity first, inflation later
Warsh has called the current wave of artificial intelligence �the most productivity-enhancing� of our lifetimes, arguing that stronger underlying output gives the Fed more room to cut without reigniting inflation. Greenspan�s precedent looms large: he persuaded colleagues to wait on hikes as the 1990s tech boom quietly accelerated productivity, keeping growth strong and prices stable.
Some current Fed voices agree in principle. Chair Jay Powell has said repeated tech waves ultimately raise productivity and wages. Governor Lisa Cook recently said growing evidence shows AI could �significantly boost productivity.� But the crucial caveat from former Fed official Vincent Reinhart: the long-run direction looks positive, yet �not much� shows up in productivity right now.
What the data actually shows
The freshest evidence is suggestive rather than decisive. A review of U.S. industry-level data finds that sectors where workers report saving the most time with AI – information, professional, scientific, and technical services – have posted unusually fast labor productivity growth since the launch of ChatGPT. That correlation strengthened into late 2025. Still, adoption remains shallow: self-reported AI use at U.S. businesses was still below 20% by the end of 2025, and correlation isn�t causation.
Another long-view study ties the 2017 advent of �transformers� (the architecture behind modern generative models) to a meaningful pickup in U.S. productivity growth, estimating software-related gains contributed roughly half of the increase between 2017�2024 versus 2012�2017. That�s real momentum – but it�s not the broad-based surge that would let the Fed bank on a rapid disinflationary boost.
The demand shock Warsh must price in
On the ground, the AI boom looks inflationary before it�s disinflationary. Alphabet plans to outspend Wall Street�s 2026 capex forecasts by at least $55 billion, guiding to a massive $175�$185 billion as it races to meet AI compute demand. Its cloud revenues jumped 48% last quarter. That is a classic demand surge: data centers, chips, land, power – everything gets bid up before productivity gains diffuse.
Economist Anil Kashyap captures the timing risk: if companies spend heavily now but efficiency dividends arrive later, near-term inflation pressure follows. That�s a tricky backdrop for aggressive, early rate cuts.
Jobs and the �blue-collar boom� narrative
Politicians have sold AI infrastructure as a blue-collar renaissance. The real picture is spikier. Data center construction needs a small army for 12�18 months – but operations of even the largest sites often require only a few hundred staff. U.S. manufacturing has shed jobs since 2023, and Germany�s post-pandemic manufacturing losses remain sizable despite defense hiring. Recruiters report intense competition for a narrow set of critical-facilities skills; widespread, durable job creation looks limited.
That matters for the Fed. In the 1990s, Greenspan sensed a broad productivity lift coincided with robust but non-inflationary wage growth. Today�s AI buildout is capital-intensive and geographically concentrated. The diffusion into everyday workflows – the bit that expands aggregate supply – looks uneven so far.
Enterprise adoption: fast headlines, slower reality
Markets can over-read splashy demos. When new AI tools spooked investors this year, software and data providers slumped – but the adoption reality is slower. Larger firms, in particular, have pulled back on near-term AI rollouts due to legal, security, and compliance frictions, according to U.S. Census data cited in recent analysis. That inertia tempers near-term productivity uplift, even as standout use cases (like AI-assisted coding) demonstrate big efficiency gains in pockets.
The policy test for a new Fed chair
Warsh says elite firms will do �things that are unimaginable� within a year. Maybe. But Greenspan didn�t win the 1996 argument with vibes; he solved a data puzzle – rising wages, high profits, and low inflation – by unearthing real productivity mismeasurement. If Warsh wants a fast-cut strategy, he needs similarly hard evidence that AI�s supply effects are here and large enough to offset the very visible demand surge from capex, compute, and power.
For executives planning budgets, the takeaway is practical:
- Expect continued pressure on AI-related input costs (compute, talent, power) in 2026.
- Target near-term productivity in functions already showing gains – software development and technical services – while preparing for slower lift elsewhere.
- Watch for clearer diffusion signals – sustained productivity gains beyond tech-centric sectors – before betting on a lasting disinflationary trend.
Greenspan�s lesson still holds: anecdotes start a thesis; data seals it. Warsh�s AI bet will rise or fall on the second.