Everything About AI
NextFin News - Kristalina Georgieva, the managing director of the International Monetary Fund, spent Friday in Bangkok telling the world that artificial intelligence could deliver negative economic shocks if safeguards do not keep pace with the technology, while the United States pushed back with a very different message: the global economy is resilient, American innovation is driving a positive shock, and the Fund's outlook is too gloomy. The clash, on the final day of the IMF and World Bank annual meetings, is more than a diplomatic disagreement. It is the central tension of this cycle: a $1 trillion-a-year AI investment boom is propping up global growth at the same time it is loading the financial system with the kind of concentrated, fast-moving risk that has ended badly before.
The IMF this week raised its global growth forecast to 3.1% for 2026, up from the 3.0% projected in July but still well below the 3.5% pace recorded in 2025 and below the historical average. The upgrade rests on AI-driven capital spending offsetting the damage from a Middle East war that has pushed oil toward an average of $96 a barrel. Growth is then seen rebounding only to 3.4% in 2027, and inflation is projected at 4.7% this year before easing to 4.1% next year. In other words, the world is buying a modest amount of growth with a large amount of risk, and the two people in the room with the most influence over the global rulebook disagree on whether that trade is sound.
"The need of the hour is for urgent and practical international cooperation," said Kristalina Georgieva, the IMF managing director, at a plenary session on Friday. Georgieva, the first IMF managing director from an emerging market economy, added that she had spent sleepless nights worrying about how to regulate AI and how to keep poorer countries from being left behind by its breakneck development.
On a subsequent panel that included Anthropic co-founder Jack Clark, she went further:
"If there are no safeguards, then the risks may exceed the opportunities. And above all, we do need an ethical foundation, and I'm not sure we are there as a society."
The AI Boom Is Real, and It Is Concentrated
The scale of the buildout is what makes the argument worth having, and it is large enough that macroeconomists can no longer treat it as a sector story. Goldman Sachs Research estimates global AI-related investment at roughly $1 trillion in 2026, equal to 0.9% of global GDP, rising to 1.3% in 2027 and 1.4% in 2028. About $581 billion of that is in the United States, and cumulative AI investment since 2022 is estimated to reach $1.8 trillion by the end of this year. These are not marginal numbers. They are large enough to move aggregate demand, and the IMF is explicitly crediting them with offsetting the war-driven energy shock that would otherwise have pushed global growth lower.
But the benefits are not shared, which is why the Fund's caution is not just institutional pessimism. The IMF's latest outlook notes that the United States and China account for more than three-quarters of net demand for imports of computer chips and other AI-related equipment.
"The outlook is highly uneven," said Silvana Tenreyro, the IMF's economic counsellor and director of research, speaking at the Fund's World Economic Outlook briefing.
The Fund expects US growth to hold at 2.3% in both 2026 and 2027, while advanced economies as a group grow 2.1% this year and slow to 1.9% next year. China is forecast at 4.5% in 2026 and 4.0% in 2027; India at 7.0% in both years. The countries paying more for oil and being priced out of the AI supply chain are the ones most exposed if the boom turns.
The transmission mechanism is straightforward, and it runs through the bond market. AI investment lifts growth through the capital-spending channel today, the same way any investment cycle does. The second-order effect is what keeps Georgieva awake: if the productivity payoff arrives more slowly than the spending, equity valuations reset, and bond yields that have already climbed to multi-decade highs do the rest of the tightening for the central banks. US, German and Japanese 10-year sovereign yields have reached their highest levels in decades, and the IMF's Athanasios Vamvakidis, deputy director of the Monetary and Capital Markets Department, flagged an AI-related stock-market correction, rising bond yields and cyberthreats as the trio of financial-stability risks. A 10-year Treasury yield above 5% is not a market that expects a smooth, self-financing boom.
History offers a template for how this ends. General-purpose technology buildouts — railroads, electricity, the internet — do eventually deliver the productivity gains they promise, but they also deliver capital write-downs for the investors who overpaid for the infrastructure along the way. The gap between the social return and the private return is where the cycle breaks. The IMF is not arguing that AI will fail; it is arguing that the financing of AI may fail before AI fails, and that the two are not the same thing.
Washington Disagrees — and It Is Not Just Optimism
"The IMF needs to take a more balanced approach to assessing the global economic outlook," said Erin Browne, US Treasury under secretary, in a call with reporters.
"The global economy has demonstrated significant resilience with strong US growth helping to support broader economic and financial stability."
She argued the Fund should give equal consideration to "opportunities for positive economic shocks, particularly those driven by American innovation and super intelligence, energy abundance, and productivity-enhancing reforms," citing US tax, trade, energy and deregulation policies as the path to stronger investment, higher productivity and durable growth.
The US case is not empty, and it should not be dismissed as political messaging. American equities have been hitting record highs, the US is the single largest destination for AI capital, and if the productivity acceleration is real and durable then the Treasury's framing wins outright: the world is not at the top of a cycle, it is at the start of a new one, and the IMF is applying an old playbook to a new technology. That is the strongest version of the counter-thesis, and it deserves to be answered on its own terms.
The answer is that the Treasury is right about the first-order effect and Georgieva is right about the second. AI spending is genuinely supporting growth today; the question is whether the return on that spending justifies a permanent re-rating of risk. The bond market is already pricing that uncertainty. The 10-year Treasury yield traded around 5.1% in mid-October, more than a percentage point above where it sat a year earlier, and that repricing is not a detail — it is the discount rate against which every AI project is being underwritten. If yields stay here while earnings growth disappoints, the mathematics of the boom change even if the technology does not.
Emerging Markets Are the Transmission Belt
The place where this tension becomes concrete is the emerging world, and the warning signs are already flashing. South Korea's Kospi fell 5.5% on Tuesday, after sinking as much as 7% intraday, as investors questioned the sustainability of the AI boom. Japan's market lost 5.4% over the week. On Friday, Indonesian stocks dropped 11.6%, their steepest one-day fall in 28 years, after global index provider MSCI said the Indonesian market no longer qualified for emerging-market status.
"Investment poured into Asia, and then you had the crisis because it wasn't really a miracle," said Eli Remolona, governor of the Bangko Sentral ng Pilipinas. "We have not really been making a lot of effort looking at that kind of crisis."
The parallel is not exact — Asia's foreign-exchange reserves and current-account positions are far stronger today than they were then — but the mechanism is the same: capital flows in on a story, the story gets repriced, and the borrowers are left holding the bill.
The IMF's own balance sheet shows the stress building beneath the surface. Christian Mummsen, the Fund's strategy chief, said low-interest lending to low-income countries has quadrupled since the pandemic and demand is expected to remain strong; about 60% of IMF financing since COVID has gone to low-income countries. Barbados this week joined a group of countries calling for an international panel to address inequality, citing geopolitical tensions, climate change and rapid technological change. When the lender of last resort is seeing demand for concessional lending quadruple, the system is telling you something about the limits of resilience.
The jobs channel adds a slower-moving version of the same risk. Georgieva has estimated that AI will affect roughly 40% of jobs globally, with positions augmented, downgraded or eliminated, calling the labor-market impact "a slow moving tsunami, but a tsunami nonetheless" — a structural redistribution of income, not a cyclical fluctuation, and it lands hardest in the countries with the weakest fiscal capacity to retrain workers or support displaced labor.
Cyclical Boom, Structural Divide
The cyclical-versus-structural call matters because it determines the conclusion, and this story has both forces at work. The AI investment boom itself is cyclical: it is a capital-spending wave that will peak and slow, and Goldman's own forecast implies it moderates as a share of global GDP after 2028. Spending that is 0.9% of global output this year cannot grow at the same rate forever without consuming an implausible share of world investment. When it slows, the growth it has been supporting slows with it. That part reverts.
The structural part is the divide the boom is carving. The concentration of AI capability, chip demand and productivity gains in a handful of countries is not a cycle; it is a change in the distribution of economic power. The 40% of jobs worldwide exposed to AI is a labor-market restructuring, not a fluctuation. And the fiscal position of the countries that borrowed through the war and the energy shock is a balance-sheet problem that does not self-correct. The right call is therefore split: expect the investment leg to mean-revert, and expect the inequality and debt leg to persist.
What to Watch
Three signals decide which side of this argument wins. First, the productivity data: if US non-farm productivity and unit labor costs show sustained acceleration through the next two quarters, the Treasury's "new era" case strengthens. Second, the bond market: if the 10-year Treasury yield holds above 5% while earnings growth disappoints, the discount-rate channel will do the tightening that central banks are trying to avoid. Third, the AI capital-expenditure guidance from the hyperscalers: any sequential deceleration in announced spending is the first sign the cycle is peaking.
The base case is that growth holds near the IMF's 3.1% for 2026 while volatility rises, because the spending continues but the risk premium widens. The upside case is a productivity surprise that validates the spending and lets yields stabilize below 4.5%. The downside case is an AI capex pullback combined with a prolonged Middle East conflict, which would push the global economy below the 3.1% forecast and toward the 3.0% July estimate or lower. The falsifying signal for the cautious view is specific: if global AI investment stays above 1.5% of GDP through 2028 while core inflation in advanced economies returns to target and the 10-year Treasury yield falls back below 4%, the risk warnings were overdone and the Treasury's optimism was right.
The central judgment is this: the IMF and the US Treasury are not arguing about what is happening today. They are arguing about whether today's growth is being rented or earned. The AI boom is real, but it is being financed by investors who have not yet seen the return, in countries that have not yet felt the benefit, at a moment when the bond market is charging the highest price for long-duration risk in a generation. That is not a foundation for a soft landing; it is a bet on a technology arriving fast enough to outrun its own financing costs.
Explore more exclusive insights at nextfin.ai.
