Georgieva Says AI Tsunami Will Eat the Tax Base
The head of the IMF just told governments their tax base is about to get eaten. Not by a recession. Not by a war. By software.
Kristalina Georgieva didn't fly to Bangkok to warn about jobs — she's been doing that since January 2024. She flew there to tell finance ministers something harder to hear: that AI will quietly tilt the whole economy away from labour and toward capital, and labour is where governments get their money. "It is a slow-moving tsunami, but tsunami none the less," she said on Monday (Oct 12), making her first appearance at the IMF–World Bank annual meetings in the Thai capital, speaking at a session on AI's impact on employment billed as "Small AI: More and Better Jobs".
Here's the twist nobody in the room missed: Georgieva is one of the world's most senior economists, and she spent as much time on tax receipts as on unemployment lines. When the IMF chief starts asking what governments can tax, you listen — because she's telling you the fiscal maths of the AI boom doesn't add up.
The 40% Number Isn't New. The Tax Part Is.
Let's get this out of the way first, because half the coverage today will get it wrong. "The impact over time on jobs globally we estimate to be around 40 percent jobs either augmented, or downgraded, or eliminated," Georgieva said in Bangkok. That is not a new finding. We checked the IMF's own archive this morning: the Fund published almost exactly this figure — "almost 40 per cent of jobs", 60 per cent in advanced economies — in a staff analysis released in January 2024, with Georgieva writing then that "in most scenarios, AI will likely worsen overall inequality."
So the number is recycled. What's new is the company it keeps. This time Georgieva bolted a fiscal mechanism onto it: AI doesn't just reshuffle who works — it reshuffles what gets taxed. Labour, she argued, is governments' main tax base. If AI shifts value creation from workers to capital, the revenue base shrinks with it. That turns an employment story into a sovereign-finance story, which is why finance ministers, not labour ministers, should be losing sleep.
The IMF's own numbers make the asymmetry concrete. Its US research found that roughly 10 per cent of American jobs already require AI skills — and those jobs pay more. So the skill premium is already here, concentrated, while the displacement spreads outward. That's the pattern behind the 40 per cent: augmented at the top, downgraded and eliminated at the bottom.
Young Workers Get Hit First
The mechanism isn't mysterious. AI is hollowing out routine, repetitive tasks first — exactly the work entry-level staff have always done. "People to worry about, first and foremost, [are] young people that are coming into the labour market," Georgieva said, as reported by CNBC Africa from the panel. "Normally, they would get into these routine, repetitive jobs and then grow up the ranks. But if they're gone, young workers could find it harder to establish themselves in the labour market."
Think about that ladder for a second. Every senior analyst started as a junior analyst; every partner was once an associate drowning in document review. If the bottom rung gets automated away, the ladder doesn't get shorter — it loses its first step. You can't climb into a career you've never entered.
She wasn't the only one sounding the alarm. Singapore's President Tharman Shanmugaratnam — who co-chairs the World Bank's Jobs Council — told the same gathering that "the first rung in employment, your first job upon graduation, is now seeing less demand." His prescription was blunt: the emergency employment measures governments rolled out during Covid and the financial crisis should stop being treated as temporary. "Now has to be thought of not as a crisis measure, but as part of the structural, the menu of structural policies," he said, urging governments to work with the private sector to incentivise hiring fresh graduates.
Some scene-setting, because the setting matters: roughly 18,000 people are expected in Bangkok this week, including finance officials from the IMF's 191 member countries, and Georgieva used her first appearance of the meetings to skip the pleasantries entirely. No boilerplate about resilience. Straight to the tsunami.
"Not a Story to Celebrate"
One of Georgieva's sharpest lines was aimed at the optimists who point to job creation as proof the transition is fine. Yes, jobs are growing — in construction, and in data centres. "Not a story to celebrate," she said. The net employment effect in the IMF's US study was actually positive, but the growth was concentrated in lower-paying roles: people pouring concrete and racking servers, not designing the systems.
That's the uncomfortable shape of the AI labour market so far. The gains are piling up among a narrow set of workers and economies even as the investment supercycle rages — the same supercycle behind the $113B AI debt machine now straining Wall Street's appetite for data-centre borrowing. Billions flow into the infrastructure; the jobs it creates at the bottom pay badly; the value accrues to capital. Which brings us back to Georgieva's real point: when the value accrues to capital, and capital is taxed lighter than labour, governments slowly starve.
What Do We Tax? The Question No Government Can Answer
Strip the rhetoric and the fiscal argument is brutally simple. Governments tax wages — income tax, payroll tax, social contributions. They tax corporate profits far less effectively, and capital gains lighter still. If AI shifts output from labour to capital, the tax base migrates from the heavily taxed thing to the lightly taxed thing. Revenues fall even as GDP rises. "What do we tax? How do we generate the revenues? If revenues are shrinking, they would be in a tough spot," Georgieva warned, as reported by The Times.
It is a shoddy state of affairs that no major economy has a serious answer to this. Robot taxes remain think-tank fodder. Taxing AI compute or data flows sounds elegant in a white paper and collapses on contact with trade law. Meanwhile the clock is ticking: every quarter the economy tilts further toward capital is a quarter the old revenue model decays a little more.
To be fair to Georgieva, she didn't only bring doom. The same speech argued the AI investment boom is what's keeping the world economy growing at around 3 per cent, delivering a "demand shock, an investment shock" that could lift advanced Asian economies by half to a full percentage point of growth — "more than offset", in her words, the drag from ageing populations. "AI, love it, hate it. It's already here," she told hundreds of participants, as Reuters reported. The Fund's updated outlook lands on Tuesday.
The growth story and the tax story are the same story, seen from different ends. Investment surges, output rises, and the returns concentrate — in the hands of capital owners. That concentration is precisely why Broadcom's Hock Tan argues AI's real bottleneck is capital, not chips: the scarce input isn't silicon anymore, it's the money to deploy it. Georgieva is now saying the scarce thing for governments will be the revenue to respond to it.
Banga's Counter-Plan: Make Jobs the Whole Mission
Across the corridor, World Bank Group President Ajay Banga was sketching the institutional answer. "The bank's entire focus has to be on jobs," he said — and his example was deliberately low-tech: a farmer with a basic mobile phone who photographs a diseased crop, gets treatment advice, and connects to seeds, credit and markets. AI as a ladder, not a tsunami.
Georgieva's own prescription was less poetic but more actionable: rethink education, build lifelong learning, and — above all — stop studying the problem and start moving. "We cannot today imagine what exactly the world of work will be five years from now. But we can be disciplined and carefully assess what is happening, and then draw lessons, but — most importantly — act on them now."
"Act on them now" is doing heavy lifting in that sentence, because almost nobody has. Governments have spent two years commissioning AI workforce strategies that read like they were written by the models themselves. Meanwhile even the people building the technology are issuing their own warnings from inside the industry — Microsoft's Satya Nadella just told enterprises to assume every AI model is already compromised. When the builders and the bankers agree something is moving faster than the guardrails, "act now" is the least a policymaker can do.
Our Take: She's Right About the Cliff — the Timetable Is the Open Question
Here's our position, stated plainly: the tax warning is the most important thing said about AI by any major economic official this year, and it deserves more attention than the recycled 40 per cent figure will get. The mechanism is concrete, the fiscal logic is honest, and Georgieva is right that it's barely on any government's agenda.
The honest caveat is the timetable. "Slow-moving tsunami" is vivid, but slow-moving is doing a lot of work — the IMF's own estimates describe effects "over time", and the Fund has a long institutional habit of front-running its own forecasts for effect. The capital-labour tilt is real, but whether it starves treasuries in five years or fifteen depends on policy choices that haven't been made yet, including ones (like taxing capital properly) that governments have avoided for decades.
What isn't debatable is the direction. Every data point Georgieva cited — the 10 per cent of US jobs already demanding AI skills, the entry-level squeeze, the low-paying growth at the bottom — points the same way. The question she left hanging over Bangkok is the one no finance minister answered: when the workers who fund the state are replaced by capital that doesn't, who pays for the state? "Act on them now" is good advice. It would be better if anyone were taking it.
Sources: The Japan Times, The Straits Times, The Business Times (Singapore), The Times (UK), TheCable, Vanguard, CNBC Africa, Reuters (reporting by Andrea Shalal and Rae Wee), Analytics Insight — all reporting from the IMF–World Bank annual meetings, Bangkok, Oct 12, 2026.
— TechPulse USA, Bangkok and Washington desks
TechPulse USA covers the business and policy of artificial intelligence for a US audience — the money, the power, and the people caught between them. We report from both sides of the Pacific .
Comments
Post a Comment