Nobel Economists Warn: AI’s Productivity Payoff Falls Far Short of the Hype

Christopher Pissarides doesn’t mince words. The 2010 Nobel laureate in economics sees artificial intelligence as useful in spots. Yet he dismisses any notion it will spark the sort of rapid productivity surge that powered economies in the 1980s and 1990s.

“It’s just not practical to talk about high productivity growth,” Pissarides told The Next Web. “I think we should be resigned to the fact that the days of fast productivity growth are over, whatever we do.”

His assessment lands at a moment when executives, policymakers and investors chase the next big AI breakthrough. Nvidia’s Jensen Huang and OpenAI’s Sam Altman paint pictures of transformative change. Central bankers listen closely at forums like the Sintra gathering. But data and history tell a more measured story. One that two Nobel winners echo with growing insistence.

Pissarides, a professor at the London School of Economics, points to a stubborn fact. Up to 40% of jobs in the UK and US remain largely untouched by current AI systems. Nursing, hospitality and many hands-on roles offer little exposure. No productivity lift there. No economy-wide boom follows.

“There is up to 40%, or at least a big number of jobs in the UK, which are not exposed to AI so they are not going to get productivity gains because of AI,” he said.

And he doubts a fresh computer-driven expansion will match the personal computing wave decades ago. “I doubt there will be a new computer boom equivalent to what we had in the 1980s and 1990s.” Given current trends, productivity won’t reach those levels. Resignation, in his view, beats unrealistic expectations.

The Skeptics’ Numbers

Daron Acemoglu goes further. The MIT economist, who shared the 2024 Nobel for work on institutions and prosperity, crunched the models. His conclusions appear in “The Simple Macroeconomics of AI,” published in Economic Policy. Over the next decade, he projects AI will deliver between 1.1% and 1.6% total GDP growth in the US. That works out to roughly 0.05% additional productivity growth per year. Or, in some refinements, about 0.55% in total factor productivity gains.

Compare that to post-1947 US averages. GDP growth has run near 3% annually. Productivity around 2%. Industry forecasts from Goldman Sachs once talked of 7% global GDP lifts. McKinsey envisioned 3 to 4 percentage point annual boosts. Acemoglu calls those figures disconnected from evidence.

“I don’t think we should belittle 0.5 percent in 10 years. That’s better than zero,” Acemoglu told MIT Economics. “But it’s just disappointing relative to the promises that people in the industry and in tech journalism are making.”

His math rests on task-level realities. Only about 5% of tasks look profitably automatable soon. AI shines on easy, verifiable work. It stumbles on jobs needing judgment, context or creativity. Integration costs add friction. Marginal improvements rarely justify the spend at scale. Hulten’s theorem, which weights task importance by economic share, caps the aggregate effect.

Recent studies reinforce the caution. A February 2026 Economist analysis found America’s 2025 growth at 2.2% with slowing hiring. Productivity showed some lift but no clear AI signature yet. The Economist reported that despite fast AI progress, output effects remain muted. Treasury Secretary Scott Bessent and Fed nominee Kevin Warsh had hoped for quicker impact. Data hasn’t cooperated.

Even optimistic projections temper over time. Wharton Budget Model researchers estimated AI could raise productivity and GDP by 1.5% by 2035, rising to 3.7% by 2075. Yet the annual growth contribution peaks early then fades. Permanent effects shrink to under 0.04 percentage points once adoption saturates. The Wharton analysis highlights sectoral shifts but stops short of promising a new golden age.

Acemoglu stresses the distinction between automation and complementarity. Current systems mostly speed existing work or replace narrow tasks. Real gains come when AI creates new tasks. When it expands what humans can achieve. Podcasts, he notes, grew demand for news and commentary. AI could do something similar. But it hasn’t. Not widely.

“We’re using it too much for automation and not enough for providing expertise and information to workers,” he added in the MIT discussion.

Executives appear to agree, at least in surveys. A April 2026 Fortune report found thousands of CEOs reporting little measurable impact on productivity or employment from AI. The pattern echoes the IT productivity paradox of earlier decades. Fortune detailed the disconnect.

But. Implementation lags. Organizational change proves hard. Skills gaps persist. And AI models still fail basic real-world tests. They can’t read a room. They miss non-obvious connections. They falter where judgment matters most.

“It’s not that you cannot get big productivity gains from automation,” Acemoglu told Fortune in June 2026. “It is that it’s not as easy as sometimes it’s presumed.” He labeled much of the surrounding discourse “brainless.” About 80% of it, in his estimate. Speculative. Sometimes fictional. Focused on the wrong questions.

Power concentration worries him more. Hyperscalers dominate. They extract data and attention. They shape regulation. Without deliberate policy, AI risks becoming extractive rather than inclusive. Wages stagnate for many. Inequality widens. Democracy strains.

“What we should be talking about is the displacement and unequalizing roles of AI,” he said. He dislikes the blanket term “capitalism” because it lumps Sweden with Honduras. The real test lies in institutions. Do they spread participation and reward innovation? Or hoard gains at the top?

Pissarides once floated a four-day workweek enabled by productivity gains. He no longer sees AI delivering that outcome broadly. The technology helps in exposed sectors. It falls short elsewhere. Fast growth, he argues, belongs to the past.

Recent Dallas Fed research offers a middle ground. AI could add 0.3 percentage points to annual productivity growth over a decade in conservative scenarios. That lifts GDP per capita by a few thousand dollars by 2050. Noticeable. Not transformative. Goldman Sachs analysts once projected up to 3 points. The gap between forecasts remains wide. The Fed paper underscores how adoption, skills and execution will decide outcomes.

World Economic Forum surveys of chief economists point to 1.5% to 3% labor productivity gains over ten years. Concentrated in tech, finance and professional services. Early time savings appear in St. Louis Fed data. Yet aggregate proof stays elusive. WEF summarized the expert consensus.

So what now? Acemoglu calls for better conversation. One centered on desirable outcomes. Wages. Job quality. Shared prosperity. Global cooperation on safety and standards, even with competitors like China. He sees little appetite in Washington. Bipartisan China criticism crowds out nuance.

Failure of imagination, he says, traps debate. Hyperscalers set the terms. Alternatives stay vague. If AI displaces 30% or 40% of new graduates without creating quality work, social consequences could turn serious. History shows revolutions follow youth unemployment spikes. Social media adds unpredictable fuel.

Fortune 500 leaders, he quipped, should hope his modest forecasts prove correct.

Evidence accumulates slowly. Early 2026 data shows no productivity explosion. Investment runs hot. Returns lag. The 1990s PC boom combined hardware, software and organizational overhaul. AI lacks equivalent diffusion so far. Task substitution dominates over task creation.

Economists at AEI note the debate hinges partly on science. Can AI accelerate discovery itself? Current systems automate known processes better than they invent new ones. An NBER paper suggests that pivot could matter more than pure automation. AEI explored the science angle in March 2026.

Pissarides and Acemoglu speak from different generations and datasets. Their shared caution carries weight. They built careers explaining how technology and institutions interact. Both see AI as powerful within limits. Both reject euphoria that ignores those limits.

Productivity growth matters. It drives living standards, wages and fiscal room. Slower growth means harder choices. Higher debt. Later retirements. Less room for error. Policymakers who bet everything on an AI windfall may face disappointment.

Yet zero isn’t the alternative. Modest gains compound. Targeted applications in medicine, education and design already show promise. Complementary innovations could expand them. The question is whether society steers development toward those paths or accepts whatever hyperscalers prioritize.

Acemoglu puts it plainly. We need to articulate what a human-centered AI future looks like. Then summon the will to demand it. Otherwise the technology delivers what it was built for. Narrow efficiency for a few. Uneven gains for the rest.

The rapid growth era may indeed be over. Not because of AI’s failure. But because expectations ran too far ahead of what the technology, the economy and the data can realistically support. Resignation has its place. So does determined, clear-eyed policy. The difference will shape the next decade.