The boom in investments in artificial intelligence has powered global economic growth through the shock in energy markets, but that could end abruptly, the International Monetary Fund warns.

If the expected profits and productivity gains from AI don't materialize, the fund forecasts a sharp decline in investment in the technology sector, concentrated in the U.S.

Under a market-correction a scenario, an unwinding of AI-related investment would reduce U.S. investment by 5% and gross domestic product by 0.8% in coming years, the IMF said Tuesday.

That would trigger a selloff in assets and tighter financial conditions, sweeping across economies.

It modelled the scenario using World Economic Outlook assumptions over a 2026-2031 forecast horizon.

AI has sparked unprecedented levels of spending, funnelling billions into chips, data centers and related infrastructure.

That's creating what the IMF called a "front-loaded demand shock," raising demand for capital goods, construction services, land, energy and labor before broad productivity gains are secured.

A reappraisal of AI productivity gains would set off a domino effect.

"Equity prices in the U.S. [could] fall by 20% in 2026 and 15% elsewhere," the IMF said.

That would sour risk appetite, raising corporate risk premiums temporarily--most acutely in the U.S.--but also in other economies, including China. Weak sentiment would bleed into sovereign bond markets, spurring a rise in term premiums, the IMF said.

The concentration of those shocks would manifest in reduced appetite for U.S. assets, including the dollar, it added.

The consequences for broader global economic growth would be grim.

The spillovers in Asia would be sizable, reducing regional GDP by about half a percent by 2029 relative to the baseline, the IMF said. High-end tech exporters like South Korea would be hardest hit.

Asian tech producers would experience a 20% drop in export prices, whereas others would take a 4% hit.

By 2028, the great AI unwind would directly reduce global GDP by as much as 0.4% and investment by as much as 1.3%, the IMF said.

The corresponding tightening of financial conditions will amplify the shock, with reversing wealth effects causing a big contraction in consumption, particularly in the U.S.

Globally, consumption would fall by 0.6% by 2028, with output shrinking by 0.6% and investment by 1%.

But if the bets on AI payoff beyond expectations, the AI buildout would increase the world's GDP by 0.6% by 2029, relative to the baseline, the IMF said.

The biggest winners in the positive scenario are the U.S. and Asia, with U.S. GDP continuing to grow over the forecast horizon, reaching 1.3% by 2031.

Countries exposed to gains in automation and better prepared for AI adoption would log larger productivity gains, fueling the investment surge while supporting consumption.

The polarizing scenarios underline the IMF's message to policymakers to tackle the risks posed by AI--both positive and negative--to deliver more balanced growth that doesn't rely so heavily on select economies, businesses and sectors.

So far, AI has been the strongest offset for the current energy shock. But the resilience has been uneven, and policy room to respond to the fallout of an AI market correction is more limited than ever, the IMF said.

"A lot of shocks can be unpredictable, and that's why it is extremely important for policies to strengthen resilience, to increase buffers and be able to del with an unforeseen scenario," said Athanasios Vamvakidis, deputy director of the IMF's monetary and capital markets department.