IMF Managing Director Kristalina Georgieva used a speech previewing next week's IMF-World Bank Annual Meetings in Bangkok to name the forces she believes are pulling the world economy in opposite directions: an energy shock, record public debt and inequality, and an artificial-intelligence investment boom that is now big enough to move financial stability — not just tech share prices.
Georgieva argued the AI build-out is inflationary in the near term. Data centres, power contracts and chip supply chains compete for the same capital, labour and electricity as the rest of the economy. Investment in AI, she said, is on track to surpass — in relative scale — the historic build-outs of railroads, electricity grids and telecommunications networks.
She then pointed to the financing side of the boom. Roughly $450 billion in AI-related debt was issued in 2026, with trillions more projected through 2030, spreading exposure beyond technology equities into credit markets. In her framing, the danger is not that AI is a bubble by definition, but that a sudden repricing — investors losing confidence and valuations correcting sharply — could spill into the wider financial system.
That warning lands days after the Bank of England flagged energy prices, geopolitical conflict and rising AI-related debt as growing financial-stability risks, and it echoes a debate already running through markets: the S&P 500 and Nasdaq hit records this week on the back of AI names before pulling back, while US Treasury yields climbed to their highest levels in more than two decades.
The IMF's prescription is familiar but pointed: faster action to curb public debt, more room for countries to absorb shocks, and faster work on AI regulation so that the technology's benefits do not arrive with a systemic tail risk attached. Fresh IMF growth forecasts, due during the Bangkok meetings, are expected to show the steepest downgrades in economies damaged by war.
The meetings run from 12 to 18 October and will bring finance ministers and central bank governors together against an unusually crowded risk calendar — energy, debt, and now the machine-learning capex cycle itself.
