The stable IMF forecast that is really a bet

The headline from the IMF's July outlook reads like comfort. Global growth, it says, holds at 3.0% in 2026 and 3.4% in 2027, a notch below the 3.5% average of 2024–25 yet "broadly unchanged" from a season earlier. Investors are meant to conclude that the world is fine. They should be wary of the number's apparent calm, because a forecast is assembled, not discovered, and this one is the arithmetic of two forces pulling so hard in opposite directions that the average looks stable only because it sits between extremes.

Start with the drag. A war in the Middle East that broke out in late February has delivered the textbook version of a negative supply shock: disrupted energy flows, higher costs, households and firms postponing spending while they wait to see how far the fighting spreads. The IMF cut its 2026 projection to 3.0% from 3.1% in April, admitting that the artificial-intelligence boom has not fully offset the conflict's fallout. Then consider the lift. The fund credits the same forecast with positive "technology momentum" from AI, an investment surge large enough to offset much of the war's damage in its own column. Two shocks, one number.

The trouble is that an average of two forces is not a fact about the world; it is a guess about how they will net out. Independent modelling suggests the plausible range of 2026 outcomes is wide — global growth could come in anywhere from roughly 1.4% to 4.6% depending on how the war and the AI cycle weigh on each other. The IMF's staff have chosen a weighting in the middle and presented it as a forecast. Its projections are contingent on two things at once: that the conflict stays contained, and that the AI investment wave keeps rolling. Downside risks, the April report conceded, dominate; a longer or broader war, a reassessment of AI-driven productivity, renewed trade tensions, all tug the number toward the low end.

A number that hides a dividing world

Beneath the single figure sits a split that matters more to a portfolio than any global total. The two crosscurrents do not cancel out country by country; they separate winners from losers. Whoever sells the machinery of the boom — chips, servers, the hardware of data centres — gains; whoever imports energy and sits near the fighting loses. In the first quarter of 2026, the top four net exporters of AI hardware recorded a positive growth surprise of about 4.4 percentage points while all other countries ran slightly negative. This is why a stable global growth number is compatible with very unequal fortunes beneath it.

The most consequential question is whether the rebound to 3.4% next year is built on foundations the fund has actually assumed. It is not clear that it is. The IMF appears to count the AI boom as demand — capital spending that lifts the level of output for a year or two — rather than as productivity, which would lift the rate of growth permanently. Treasury-and-wonder economic histories suggest the real payoff from a technology like AI shows up slowly, in efficiency gains spread across the economy, not as an immediate burst of investment. If the fund is right to treat it as investment-only, the current strength is partly borrowed from the future, and the 2027 rebound rests on the war ending and the investment still flowing. If the productivity gains arrive after all, the forecasts are too low on the upside. Either way the baseline is a conditional statement, not a plan.

That conditionality has already been stressed. The July estimate was finalised before the freshest exchanges of fire between America and Iran, so the war risk inside the number is the mild version of itself. The cost of the boom, meanwhile, is visible in markets: long-term interest rates across the big Western sovereign curves have drifted up even as inflation cools, because financing investment in digital, environmental and defence transitions demands a higher neutral rate. A higher discount rate is a tax on long-dated growth, a reminder that the "technology momentum" the fund is counting on is not free.

What the reader does with the number

The practical moral is to give the 3.0% little weight. Global GDP is not a tradeable asset, and this particular summary conceals a wide range of outcomes, an uneven distribution of gains and losses, and a rebound whose foundations are assumed rather than earned. The useful question is not whether growth is "on track" but where a holding sits in the split: on the side of the hardware makers, the data-centre suppliers and their customers, who are being carried by the boom; or on the side of the energy importers and war-adjacent economies, who are paying for it. That is where the profit, and the risk, actually live. The IMF's stable headline is a reminder that an average can hide a dividing world — and that the safest reading of an official forecast is often to ask who is expected to lose.