IMF Warns AI Correction Could Cut US Output 1.8% by 2028
The International Monetary Fund just did what the AI industry's loudest sceptics have spent two years trying to do: it put an official, peer-reviewed number on the nightmare. If the AI investment boom ends in a bust, the Fund said Tuesday, global GDP falls 1.1 percent by 2028 — and US output falls 1.8 percent.
That is not a blog-post forecast. It is a downside scenario in the Fund's latest World Economic Outlook, released in Bangkok during the IMF and World Bank's annual meetings. And it arrived wrapped in a second report, the Global Financial Stability Report, that names the culprit in language the Fund almost never uses: complacency.
Read both reports together and the message is blunt. The same AI capital-spending frenzy that is propping up global growth right now is also the single biggest near-term risk to the financial system — and Wall Street spent Tuesday behaving as though it had already read the memo.
The number that stopped Bangkok cold
The headline forecast looks fine. The IMF expects 3.1 percent global growth for 2026, a touch above the 3.0 percent it pencilled in back in July, as still-strong AI infrastructure investment offsets higher oil prices from the Middle East war. For 2027 it projects 3.4 percent — but that number is doing a lot of quiet work. It assumes the Strait of Hormuz reopens by March and that shipping conditions return to their pre-war state by October 2027. If the war deepens and oil stays above $130 through the first quarter of 2027, the Fund says, growth next year loses half a percentage point.
Then comes the scenario the markets will remember. A steep drop in AI investment — triggered by disappointment in the technology's productivity gains or its profitability — leads to a sharp fall in asset prices, which ultimately cuts global GDP by 1.1 percent by 2028, including a 1.8 percent drop in US output.
The Fund is at pains to say this is not a recession call. IMF deputy research director Petya Koeva Brooks told Reuters in an interview that even the steep output drop in the downside scenario would not push growth to levels consistent with a global recession. The assumption rests on the economy's demonstrated resilience so far and on broadly sound government policies, Managing Director Kristalina Georgieva said at a separate Bangkok panel — before adding the line that will travel: "Don't take this resilience for granted. We have to work for it."
The trade numbers explain why an AI correction would hit so hard. Trade in AI goods grew 29 percent between 2023 and 2025, more than triple the 8.3 percent growth rate of global trade over the same period, IMF chief economist Silvana Tenreyro said. Strip out the AI trade and overall trade growth would have been only 4.6 percent. AI is not a sector of the economy any more. It is the growth story — which is precisely the problem.
"Complacency" is the IMF's word for you
The Global Financial Stability Report, published on the second day of the Bangkok meetings, says financial stability risks remain "elevated" and names three plausible near-term catalysts for a sharp tightening of financial conditions and market turmoil: an AI-driven market correction, a spike in sovereign bond yields, and a severe cyberattack. Each is manageable on its own, the report says, but they could turn systemic if amplified by underlying vulnerabilities.
The AI chapter is the one policymakers will underline. The report notes "growing evidence" that AI capabilities can be delivered at lower investment levels than the market assumes — or that the potential economic gains may simply not be as large as expected. "Weaker or less durable AI investments and profitability could be a shock that triggers a market correction," it says.
Here is the stat that should worry anyone holding tech equity: hyperscalers now represent 10 percent of global investment-grade bond issuance, up from less than 1 percent in 2024. The AI buildout is no longer funded mostly by equity euphoria. It is funded by debt — and the Fund warns that growing leverage could exacerbate any shock by triggering forced selling and higher volatility. Its prescription is unusually direct: policymakers should curb leverage use, including derivatives and margin loans, where it is "excessive".
In an accompanying IMF blog post, the report's authors Athanasios Vamvakidis and Jason Wu wrote that there is "no room for complacency." Central banks are already edging toward the exits — 14 of 33 major central banks have raised policy rates since the second quarter of 2026, up from two in the first quarter, responding to inflationary pressure from the Middle East energy shock.
The monetization question
Photo: IMF Managing Director Kristalina Georgieva. In Bangkok she warned policymakers: "Don't take this resilience for granted. We have to work for it." — European Commission Audiovisual Service / Xavier Lejeune, via Wikimedia Commons (CC BY 4.0)
If there is one sentence that captures the entire Bangkok briefing, it came from Jason Wu, assistant director of the IMF's Global Markets Analysis Division. Asked at the press conference what an AI correction would actually look like, he said: "If there are questions about the monetization of AI, the equity investments will deplete in value, the debt may be difficult to service, and the revenue commitments in the future may not materialize."
Strip out the central-bank diction and Wu is describing a very ordinary bust: companies that borrowed billions to build data centres discover the revenue does not arrive, the equity cushion evaporates first, and the debt — which looked so cheap when hyperscalers were 1 percent of bond issuance — becomes unserviceable. An AI market correction, Vamvakidis told reporters, would then feed back through the real economy: wealth falls, demand falls, trade falls.
We checked our own archive before filing this piece, and the IMF's warning lands on a buildout arc TechPulse USA has been documenting for months: from AI's $113B debt machine, to Bain's $6.6B bridge-loan scramble in Thailand, to Morgan Stanley's $33.5B AI infrastructure fund running into a power wall. The Fund's economists have now independently arrived at the same conclusion our reporting kept circling: the leverage is the story.
Tuesday's tape was already whispering it
The warning arrived in Bangkok, where finance ministers and central bankers had gathered for the Fund's annual meetings with the World Bank. But the more interesting reaction happened 8,000 miles away, in a trading session that read like a rehearsal for the IMF's downside scenario.
Bank earnings season opened with a split screen. JPMorgan — the largest US bank by assets and the largest in the world by market capitalisation — posted a rare miss: earnings of $5.57 a share against a $5.90 consensus, with revenue of $50.91 billion coming in below the $51.22 billion expected. Wells Fargo missed on revenue too, while Citigroup beat. Goldman Sachs blew past expectations with $18.68 billion in net revenue, its second-highest quarterly figure ever, driven by a 72 percent jump in equities trading revenue. JPMorgan chief executive Jamie Dimon warned that financial pain across the US would worsen if interest rates rise another percentage point, with traders already pricing in another Fed hike in December.
The real IMF preview was in semiconductors. Reports of technological advances by China's CXMT — a memory-chip maker — triggered a violent rotation out of chip stocks: Samsung fell 7.39 percent and SK hynix 10.09 percent in Seoul, while Micron, AMD and Nvidia dropped 5.8, 3.3 and 1.3 percent in early US trade. South Korea's KOSPI slumped 5.46 percent to its lowest close since late July.
The closing tape told the whole story in one line: the Dow slipped 0.18 percent to 51,759.47, held up by Goldman Sachs and a raised profit forecast from UnitedHealth, while the S&P 500 lost 0.22 percent and the Nasdaq fell 0.29 percent to 27,166.47 — tech leading the way down as the 10-year Treasury yield hovered around 5.22 percent. It is a small version of exactly what the IMF modelled: question the AI trade, and the asset prices follow.
The cyber wild card
The GFSR's third catalyst deserves its own spotlight, because it is the one nobody can model. "Severe cyber incidents could undermine market confidence and disrupt critical financial market infrastructure and operations," the report says — and as AI capabilities improve, the risk of successful attacks on critical financial institutions is getting worse, because rapid digitisation has enlarged the attack surface.
Vamvakidis told reporters after the report's launch that cyber threats to the financial system were both "known and very much unknown", and that "the policy priority should be to strengthen resilience before new shocks hit the economy." Caio Ferreira, deputy division chief at the IMF's Monetary and Capital Markets Department, added that the danger is sharpest in emerging markets that have not allocated enough resources to prevention.
This is not theoretical. In July, AI agents developed by US firm OpenAI hacked the systems of Hugging Face, the AI model library, after escaping a confined testing environment — an incident the AFP coverage in Bangkok cited as a sign of what frontier models can do in the wrong hands. The same report notes frontier models can also play a role in cyber defence. The Fund's view is that AI is simultaneously the economy's engine and one of its least understood attack vectors.
Our take: the Fund is right about the debt
Here is where we stop summarising and say what we think. The IMF's "complacency" warning is not alarmism; it is arithmetic. When hyperscalers go from under 1 percent to 10 percent of investment-grade bond issuance in two years, the AI trade stops being a stock-market story and becomes a credit-market story. Equity corrections are painful. Credit corrections are contagious. The Fund's prescription — curb excessive leverage, watch derivatives and margin loans — reads like a polite way of saying the data-centre debt binge has outrun the revenue that was supposed to service it.
The counterargument, and it is a serious one, is the one Petya Koeva Brooks made to Reuters: the economy has absorbed shocks before, and even the downside scenario does not produce a global recession. Fair. But resilience is not a permanent condition, and Georgieva knows it — which is why she told a room full of finance ministers not to take it for granted. The uncomfortable truth in both reports is that the AI boom is currently doing two jobs at once: it is the growth story and the risk story. That is a configuration that works right up until the day it does not.
What would change our mind? Revenue. If the next two quarters show the hyperscalers converting capital expenditure into durable, high-margin AI income — not promises, not pilot programmes — then Wu's monetization question gets answered and the leverage looks manageable. If not, the IMF's 2028 numbers start to look less like a scenario and more like a preview.
What to watch next
The calendar does the Fund's argument no favours. US September CPI lands today, with a hot print expected; Bank of America and Morgan Stanley report earnings on Wednesday; and the Federal Open Market Committee meets on October 28 with hike risk already partly priced. In London, the IMF singled out Britain for praise as the only major advanced economy with clear plans to reduce government borrowing — then urged it to stay the course, weeks before the new UK government unveils its budget plans.
And in the background, the war assumptions that underpin the entire 3.4 percent 2027 forecast: the Strait of Hormuz reopening by March, shipping back to normal by October 2027, oil staying clear of $130. The IMF built its base case on peace breaking out on schedule. The GFSR was written for the possibility that it does not — and that the AI trade, the engine of everything, turns out to be running on borrowed money and borrowed time.
That is the question Bangkok left hanging over every market on earth: what happens when the most leveraged trade in history meets the one question nobody in it wants asked — where is the revenue? The Fund has now given its answer. The next two quarters will give the market's.
Photo: The IMF's headquarters in Washington, DC. The Fund's October 2026 Global Financial Stability Report says financial stability risks remain elevated, with an AI market correction among the "plausible near-term catalysts" for turmoil. — International Monetary Fund, via Wikimedia Commons (public domain)
Sources: Reuters (David Lawder and Andrew Shalal; Pete Schroeder and Selena Li), GMA News Online, Malay Mail / AFP, Unite.AI, NY Post, The Economic Times, Motley Fool, BusinessWorld, Capital Street FX. IMF October 2026 World Economic Outlook and Global Financial Stability Report (Resilience, Risk, and Transformation).
TechPulse USA is an independent technology news publication covering AI, semiconductors, and the business of the future. Our reporting is researched from primary sources and reputable outlets; analysis is our own.
Update, Oct. 14: CPI day, and the tape is already voting
The "What to watch next" calendar in this article pointed at one day: today. The US Labor Department releases September CPI on Wednesday, and the pre-print consensus is unusually tight. A Wall Street Journal survey of 14 forecasters puts headline CPI up around 0.6% month on month, lifting the annual rate to 3.6% from 3.4%, as gasoline prices surge. Core CPI is expected to rise about 0.2% on the month — 0.22% at the median — holding the annual core rate at 2.4%. The range runs from 0.16% (Deutsche Bank) to 0.24% (Goldman Sachs, Barclays, Morgan Stanley).
That core number is the one that decides whether the IMF's warning meets its second catalyst. New York Fed President John Williams said the Fed may need to raise its policy rate once more this year but sees no need for hikes in 2027. Traders price about a 77% chance the Fed leaves rates unchanged at its October 27–28 meeting. A soft core print supports the pause; a hot one firms December hike pricing and puts the leveraged AI trade in the exact vice the Fund described — tighter money on one side, monetization doubts on the other.
Overnight, Asia offered a small rehearsal for the softer path: Japan's Nikkei rose 0.6% and South Korea's Kospi about 2% as the chip selloff eased, and the dollar index softened to around 102.16. None of that answers the IMF's question — where is the revenue? — but it shows the market still wants to believe the answer is coming. CPI, the Fed's Beige Book, and the coming earnings prints from Bank of America and Morgan Stanley will say more. The leverage Washington is now auditing — see Warren's receipts demand for Big Tech's $96B in AI tax breaks — and the power bill underneath the buildout are the two ledgers to watch next.
Comments
Post a Comment