When the International Monetary Fund (IMF) last brought its annual meetings to Bangkok, in 1991, the Gulf War had only just ended and East Asia’s fast-growing economies were the envy of the developing world.
Thirty-five years on, the Fund was back in the Thai capital with another Gulf war on its hands, and once again, it is Asian factories that are holding the line.
The October World Economic Outlook, titled Resilience under Strain, Urgent Choices, was published in Bangkok on October 13.
It puts global growth for 2026 at 3.1%, a tenth of a point above the 3.0% the Fund pencilled in during July, and exactly where it stood in April.
The forecast for 2027 holds at 3.4%. Inflation stays at 4.7% this year before easing to 4.1% next year, a fifth of a point higher than the figure expected earlier in July.
On paper that looks like a forecast standing still. Underneath, it is anything but.
The world economy is being pulled in two directions at once, by a supply shock from the war in the Middle East that has kept oil above USD 100 a barrel, and by a demand shock from the global rush to build artificial intelligence infrastructure.
Which force a country feels more now decides whether its forecast rises or falls.
“The global economy has been resilient, but that resilience is under strain,” said Silvana Tenreyro, the Fund’s new chief economist.
How the forecast has travelled this year
In January, before the war, the IMF expected the world to grow 3.3% in 2026, helped by easing US tariffs and an AI investment boom.
The April outlook, published in the shadow of the conflict, cut that to 3.1% and raised the inflation forecast to 4.4%, up from 4.1% in 2025, as energy and food prices jumped. Even that assumed a short war.
It was not short. By July, the Fund had trimmed growth to 3.0%, lifted inflation to 4.7%, and declared that global disinflation had stalled. Yet, the July update also carried the first clear sign of what would become this autumn’s theme.
The economies at the heart of the AI hardware supply chain, namely Taiwan, South Korea, Thailand and Malaysia, were beating forecasts despite their heavy reliance on imported energy.
South Korea’s economy grew at an annualised 7.5% in the first quarter on the back of a semiconductor export boom.
The upgrade since July, the IMF says, comes largely from continued heavy investment in AI infrastructure, which has lifted forecasts for the advanced economies building AI platforms and, above all, for the countries supplying the hardware.
The winners
Taiwan is the outlier of outliers. The IMF now expects its economy to grow 10.7% this year, a pace more familiar from China’s boom years than from a mature, high-income economy.
South Korea is forecast to grow 3.3%, up from 1.1% in 2025, on the strength of chip demand. In Southeast Asia, Vietnam is projected to grow 8.4%, Malaysia 5.2%, and Singapore 5%, with the Fund tracing the strongest spillovers from Korean and Taiwanese tech exports through Singapore, then the factories of Vietnam, Malaysia, and the wider ASEAN group.
India is the big upgrade among large economies. Its 2026 forecast rose 0.6 points to 7.0%, with 2027 also at 7.0%, helped by strong domestic demand, services activity, and AI-related investment. It remains the fastest-growing major economy.
Perhaps, the most striking shift is in Europe, a region long described as a laggard in the technology race. Euro area growth for 2026 was raised from 0.9% to 1.1%.
Germany jumped 0.6 points to 1.3%, thanks to higher defence spending and, more unexpectedly, its chemical industry’s role in AI supply chains. Britain gained 0.3 points to reach 1.3% on the back of larger AI investment. Advanced economies as a group were raised 0.4 points to 2.1%.
The US sits in a curious middle position. Its 2.3% forecast for 2026 is unchanged, and 2027 has been nudged up to 2.3%.
American firms are spending more than anyone on data centres and chips, but much of that money leaves the country.
The IMF notes that the boost from AI investment is offset by its large import content, which is another way of saying that a good share of America’s AI boom is being booked as growth in Taipei and Seoul.
The losers
On the other side of the ledger sit the economies nearest the fighting, and those that import energy without exporting technology.
The Middle East and Central Asia regions are now expected to shrink 0.3%, a full point worse than in July.
Saudi Arabia’s forecast has swung from growth of 1.7% in July to a contraction of 3.3%.
Iran, Iraq, Kuwait, Qatar, and the United Arab Emirates are all projected to fall into deep recessions, with Iraq and Kuwait each expected to shrink by 16% or more.
Beyond the Gulf, the cuts are thinner but more widespread. Since April, the IMF has downgraded 47% of the economies it covers, many of them energy importers with little stake in AI value chains. China’s forecast slipped 0.1 points to 4.5%, as costlier oil compounds weak domestic demand despite strong exports.
France fell to just 0.4%, Brazil to 2.2%, and Russia to 0.9%, each down 0.2 points. Emerging and developing economies as a whole were trimmed to 3.7%.
The host country captures both forces in a single economy. Thailand is a net energy importer and remains exposed to costlier oil, yet its electronics exports have ridden the AI wave. The IMF now projects Thai growth of 2.1% in both 2026 and 2027, modest but well up on the 1.5% it expected in April.
The concentration is stark. Just 32 economies now account for 95% of global growth, according to the Fund, compared with 54 in April. The world is still growing, but fewer places are doing the growing.
Is AI now carrying the world economy?
In the narrow sense, yes. Without the surge in spending on AI hardware and infrastructure, October would very likely have brought another downgrade.
The upgrades are clustered in economies that either build AI or make the parts for it, with India the main exception.
But AI is not doing the work alone, and it is not doing it cheaply. The Fund’s deputy research director, Petya Koeva Brooks, has said global growth would have fallen to 2.5% this year had countries and companies not moved quickly to cushion the oil shock by drawing down reserves, switching suppliers, and cutting demand.
Those cushions are wearing thin. Tenreyro warned that inventory buffers and energy efficiency gains are approaching their limits.
AI also comes with a price tag. The Fund expects the investment boom to keep upward pressure on prices, one reason inflation is forecast at 4.7% this year rather than falling back.
More worrying is what happens if the bet goes wrong. The report models a scenario in which the promised productivity gains fail to appear and investment falls sharply. In that case, US equities could drop 20%, financial conditions would tighten, and global output would end up 1.1% below the baseline by 2028, with the US 1.8% lower and Asia 1.7% lower.
The economies enjoying the biggest upgrades today would take the biggest knocks.
On the upside, faster adoption could add 0.6% to global output by 2029, mostly in the US and China.
So, AI is propping up the world economy rather than carrying it. It is a narrow pillar, concentrated in a handful of supply chains, financed partly by stretched equity valuations, and it stands next to a war that has not ended.
The war assumption
The whole forecast also rests on a timetable for the Strait of Hormuz. The IMF now assumes shipping through the strait resumes in March 2027, with conditions broadly back to pre-war levels by October 2027.
In July 2026, it expected reopening to begin in mid-July. Oil is now assumed to average USD 96 a barrel in 2026 and USD 87 in 2027, against an $89 assumption for this year in July.
In the Fund’s adverse scenario, a worsening war pushes oil to $130 a barrel through the first quarter of 2027.
Growth would then be 0.2 points lower this year and 0.5 points lower next year, and inflation in 2027 would climb to 5.2%.
What the analytical chapters add
The two chapters released a week earlier read differently in light of these numbers.
The cost-of-living chapter, published on October 6, studied three decades of food and energy price spikes and concluded that such crises are likely to become more frequent. Inflation expectations rose markedly during these episodes and were still well above pre-crisis levels three years later.
The chapter found that delivering the same protection to poorer households through consumer subsidies can cost three to six times as much as targeted transfers, while subsidising producers can cost up to 22 times as much.
Chiara Maggi, who co-led the research, said such crises ‘reduce living standards and they worsen inequality’. With war and El Niño both stoking food prices, the warning is hardly academic.
The corporate tax chapter, released on October 5, looks at how tax competition shapes where multinationals book profits and build plants. It found that a one-point rise in a country’s corporate tax rate relative to others cuts foreign direct investment inflows by about 0.5% of GDP over three years.
In a year when AI investment is deciding who grows, that finding will only sharpen the contest for data centres and chip plants.
Bring debt back to sustainable levels
The policy message is blunt. The Fund wants governments to stop delaying, rebuild fiscal buffers, and set out credible plans to bring debt back to sustainable levels, while central banks keep energy-driven inflation from becoming entrenched.
For the finance ministers and central bankers, the October outlook offers a split verdict. If an economy makes the chips, the servers or the chemicals that feed the AI build-out, the war has been a cost it could absorb. If it imports oil and exports little technology, it is paying for both shocks and benefiting from neither.
Many low-income countries and fragile states fall into that second group.
The headline figure of 3.1% looks like stability. It is better read as a balance between two forces, each of which could tip. If Hormuz reopens on schedule and AI spending keeps paying off, 2027 could beat 3.4%. If the war drags on or the AI trade cracks, the narrow base now holding up global growth may prove too narrow to bear the weight.