The global economy is entering an unusual phase: artificial intelligence is helping sustain growth while simultaneously increasing the risks to financial stability. The International Monetary Fund projects world output to expand by 3.1% in 2026 and 3.4% in 2027, but warns that the recovery is increasingly dependent on a narrow group of companies, industries and economies. The IMF’s October outlook describes resilience under strain, with inflation expected to rise to 4.7% this year before easing to 4.1% in 2027.
That tension is the central fact behind the current AI boom. Spending on data centres, chips, cloud infrastructure and software has become a major source of demand, trade and investment. The IMF says stronger AI investment is supporting productivity and expects technology-related gains to contribute to the global expansion. Yet the same concentration that makes the boom powerful also makes it fragile: a small number of firms control much of the capital, computing capacity and market optimism on which the cycle depends.
Why AI has become a macroeconomic story
AI was initially treated mainly as a technology-sector development. It is now influencing the wider economy through several channels. Companies are committing enormous sums to computing infrastructure, governments are competing to secure semiconductor supply chains, and investors are pricing the prospect of productivity gains that may not arrive evenly or quickly.
The IMF’s analysis identifies AI as a positive demand shock, boosting investment and trade. That support matters as other parts of the global economy face weaker momentum, higher borrowing costs and geopolitical uncertainty. But it also creates an unusual dependence on continued spending. If the largest technology companies slow their capital expenditure, suppliers, construction firms and energy providers could all feel the effect.
Recent market signals have sharpened that concern. Deloitte reported that investors are increasingly questioning whether AI-company revenues will be sufficient to service the debt financing the infrastructure build-out. Credit-default-swap prices and yields on AI-related bonds have risen, suggesting that at least some investors are demanding greater compensation for the possibility of disappointment.
Optimism and the case for caution
The bullish case is substantial. If AI raises productivity across manufacturing, services, logistics and research, it could expand economic capacity and eventually help contain prices. Faster innovation could also generate new industries rather than merely automate existing jobs. The IMF’s forecast of stronger global growth in 2027 reflects, in part, expectations that investment and productivity will remain supportive.
Critics do not necessarily dispute AI’s long-term potential. Their concern is timing and valuation. Infrastructure is being built before many applications have demonstrated durable profits at scale. That creates a gap between current expenditure and future cash flow. If businesses conclude that AI tools deliver smaller gains than expected, demand for computing capacity could weaken while debt obligations remain fixed.
The risk is not limited to shareholders. A correction could affect pension funds, lenders, regional economies dependent on data-centre construction and public finances that have offered subsidies or tax incentives. Conversely, moving too slowly could leave countries dependent on foreign technology and deepen an already unequal distribution of productivity gains.
A geopolitical and regulatory fault line
AI supply chains are also becoming inseparable from strategic competition. The IMF warns that dependence on a limited number of countries, sectors and companies creates potential chokepoints. Semiconductor restrictions, export controls and tensions between Washington and Beijing could disrupt access to the hardware and models on which the investment cycle relies.
China’s approval of DeepSeek and MiniMax models for global use, reported by Reuters, illustrates the competing pressures. Beijing is seeking international reach for domestic AI firms while proposing export controls covering technologies related to models, agents and computing clusters. The result may be a more fragmented market, with companies and governments forced to choose between ecosystems rather than using the most efficient tools globally.
Supporters of tighter controls argue that advanced AI has military and national-security implications and that supply-chain dependence is dangerous. Opponents warn that excessive restrictions can raise costs, slow innovation and encourage duplication of infrastructure. Both arguments point to the same conclusion: AI policy is no longer only about privacy or workplace rules; it is becoming industrial and foreign policy.
What comes next
The next test will be whether AI investment translates into measurable productivity and recurring revenue. Markets will watch capital spending by major technology companies, corporate adoption rates, financing conditions for data centres and evidence that AI-generated efficiencies are reaching firms outside the technology sector.
Policymakers face a narrow path. They must preserve competition, monitor leverage and prevent public subsidies from socialising losses, while avoiding regulation so restrictive that it blocks useful innovation. Diversifying semiconductor and energy supply chains may reduce geopolitical vulnerability, but it will also be expensive.
The most plausible outcome is neither an immediate collapse nor an effortless productivity revolution. AI is likely to remain a powerful source of growth, but with increasingly volatile returns and uneven benefits. The IMF’s warning is therefore less a prediction of crisis than a description of the stakes: the world economy has gained a new engine, but it has placed too much weight on too few machines, firms and assumptions.