AI Investment Lifts Growth as Leaders Confront Job Risks

Kristalina Georgieva told IMF and World Bank meetings in Bangkok on Monday that AI infrastructure spending is helping keep global growth near 3%, despite an energy supply shock. At the same event, the IMF managing director warned that young people entering work could lose access to the routine jobs that once helped them move up.

The two claims describe different parts of an economy. Spending on chips, equipment and infrastructure can lift output in the places that supply them. A graduate trying to get a first job needs an employer willing to hire someone who is still learning. The policy question raised in Bangkok is whether governments can keep that first step open while AI changes the work done there.

Singapore President Tharman Shanmugaratnam, who co-chairs the World Bank’s Jobs Council, put the concern plainly: “the first rung in employment, your first job upon graduation, is now seeing less demand.” He called for governments to work with firms and offer incentives to hire new graduates. His warning concerns demand, not a measured count of jobs lost to AI.

The Growth is Arriving Through Hardware

Georgieva described the investment boom as a demand shock benefiting Asian economies that supply AI hardware. She estimated that AI adoption could add half to one percentage point to growth in advanced Asian economies and 0.2 to 0.8 percentage point in emerging Asian economies. Those are projected gains, not growth already recorded across either group.

The scale of the supply chain helps explain her emphasis. In remarks in Singapore the previous week, Georgieva said AI hardware and related technology products accounted for more than a tenth of global goods trade. The IMF’s July outlook had already forecast 3% global growth for 2026 and identified AI demand as a boost for technology exporters, while conflicts weighed on energy importers. Monday’s message built on that earlier picture: investment demand is supporting economies that sell into the build-out even as an energy shock drags elsewhere.

For a hardware supplier, a new order is an immediate economic event. For a worker hoping to use AI to do a job better, a productivity gain depends on adoption at a workplace and on how that workplace shares the resulting value. Georgieva’s regional estimates concern potential growth; they cannot tell a prospective graduate whether a local firm will open a junior position. That is why a global growth figure and an entry-level hiring concern can coexist without either cancelling the other.

The employment estimate needs the same care. Georgieva said the IMF expects about 40% of jobs worldwide eventually to be augmented, downgraded or eliminated by AI. That combines several possible outcomes, including jobs made more productive. It describes potential change over time, not a forecast that 40% of workers will be unemployed. The pace and mix of those outcomes remain open.

Georgieva singled out young entrants because routine, repetitive work has often served as an early step up a career ladder. Shanmugaratnam argued that AI-related displacement would proceed “task by task,” giving employers and educators time to respond if workers can learn on the job and universities provide more applied training. That is a more precise problem than trying to guess the date on which a profession disappears: what happens to training when some of its starter tasks can be automated?

A Familiar Policy Tool, with a Different Job to Do

Shanmugaratnam reached for a precedent in his proposed response. Governments used employment support during the Covid-19 pandemic and, in some cases, the global financial crisis. He argued that measures treated as temporary in those emergencies should become part of the regular policy toolkit for getting young people into work. His concrete proposal in Bangkok was to work with the private sector and incentivize firms to hire fresh graduates.

The precedent explains the shift in his thinking. Crisis measures aim to bridge a period of disrupted hiring. A standing graduate-hiring incentive would have to deal with a continuing change in the tasks assigned to junior staff. The earlier interventions were responses to economic crises, while this proposal addresses a technology that firms may keep using. Shanmugaratnam did not specify a scheme or its past results at the panel, so its success as an AI-era policy is still an open question.

An incentive could encourage a firm to offer a first job. On-the-job learning would then determine whether that job still offers a path upward as routine tasks change. Those are separate tests for policymakers: how many graduates get hired, and what skills can they build once inside? A subsidy that increases hiring without creating useful work would solve only the first.

World Bank President Ajay Banga approached the jobs question from another direction. He estimated that 1.2 billion young people will turn 18 over the next 15 years and said governments should build infrastructure, set policy and reduce risks that discourage private investment. He also described the bank’s interest in “small AI” tools that do not depend on large language models, aimed at improving productivity for farmers, healthcare systems and educators in emerging markets.

Banga’s case is that private investment creates jobs at a scale governments cannot supply on their own. Small AI offers a route to practical gains beyond the companies building large models and the economies selling them hardware. But a tool that helps a farmer or educator work better answers a different question from whether a firm will hire a graduate. Infrastructure, useful software and first-job access may reinforce one another; none guarantees the others.

Exposure Depends on Where You Look

The IMF figures Georgieva cited put eventual job exposure at about 60% in advanced economies and 23% in low-income countries. She cautioned that the lower figure partly reflects AI being further away in poorer countries. Lower exposure today therefore offers little basis for calling those workers secure, especially when access to the technology and its productivity gains also differs.

Set those figures beside the Asian growth estimates and the limits of a single AI scorecard become clear. Advanced and emerging Asia are regional growth categories; advanced economies and low-income countries are broader income groups used for job exposure. The percentages measure different things in different populations. They can guide separate questions about growth and work, but combining them into a ranking of winners and losers would manufacture an answer.

Bangkok’s strongest policy idea starts at a less glamorous place than the data center: the first job. If AI investment keeps supporting growth, governments and employers will still have to decide whether beginners get paid work in which to learn. The growth forecast cannot make that decision for them.