AI Might Buy Governments Time. It Won’t Pay the Bill.

Artificial intelligence is increasingly seen as a real productivity story, not just a technology story. The OECD’s recent work estimates that AI could raise annual total factor productivity growth by roughly 0.25 to 0.6 percentage pointsover a 10-year horizon, equivalent to about 0.4 to 0.9 percentage points of extra annual labour-productivity growth under standard assumptions. That is meaningful. Faster productivity can lift output, support wages, and expand the tax base. But even optimistic institutional estimates describe AI as a boost to growth — not a cure for fiscal imbalances. Source: OECD, “AI for public finance: potential subnational efficiency gains” (May 2025); OECD, “Miracle or Myth? Assessing the macroeconomic productivity gains from artificial intelligence” (2024). (OECD)

The debt problem governments face is much larger and more entrenched. The IMF said in April 2025 that global public debt was projected to rise by 2.8 percentage points of GDP in 2025, pushing debt above 95% of GDP, with the ratio likely to approach 100% of GDP by the end of the decade. The OECD has also warned that debt levels are still climbing across advanced economies: its Global Debt Report 2025 said the aggregate central government marketable debt-to-GDP ratio in OECD countries could reach 85% in 2025, while government interest payments hit 3.3% of GDP in 2024, exceeding defence spending across the OECD as a whole. Source: IMF Fiscal Monitor blog and press briefing (April 2025); OECD Global Debt Report 2025 and OECD press release (March 2025).(IMF)

Why does that matter? Because debt dynamics are being driven by structural fiscal pressures that productivity alone does not erase. The OECD says near-term budget strains from debt-service costs and military spending are set to be compounded by longer-run costs tied to population ageing and climate mitigation and adaptation. In other words, even if AI lifts growth, governments still face spending commitments and refinancing burdens that do not disappear when software gets smarter. Source: OECD Economic Outlook, Volume 2025 Issue 1; OECD Economic Outlook (May 2024). (OECD)

There is also a second problem: the size and speed of the AI payoff remain uncertain. OECD research says the conditions for large, economy-wide AI gains are still “far from being met,” with adoption uneven across firms and sectors and dependent on complementary investment in skills, data, and organisational change. The BIS has made a similar point, reviewing estimates that range from as little as 0.07 percentage points of annual TFP growth to around 0.9 percentage points in more optimistic scenarios. That is a wide range, and it underlines the core fiscal reality: governments cannot responsibly budget as though the most optimistic AI scenario is guaranteed. Source: OECD, “The impact of Artificial Intelligence on productivity, distribution and growth” (2024); BIS Bulletin No. 121, “Economic impact of AI in emerging market economies” (2026). (OECD)

Even the fiscal upside from AI is not automatic. The IMF says generative AI could improve public service deliveryand help support productivity growth, but it also warns of labour disruption, higher inequality, and pressure on the tax system if labour’s share of income declines while gains concentrate in capital. That matters for debt because stronger GDP does not automatically translate into stronger, broader, or more stable public revenues. Governments may collect more from growth at the margin, but the composition of that growth still matters for tax receipts. Source: IMF, “Broadening the Gains from Generative AI: The Role of Fiscal Policies” (2024); IMF blog, “Fiscal Policy Can Help Broaden the Gains of AI to Humanity” (June 2024). (IMF)

The United States offers a concrete example of the scale mismatch. In its February 2026 outlook, the Congressional Budget Office projected that federal debt held by the public would rise from 101% of GDP in 2026 to 120% in 2036. Even if AI improves productivity and slows the pace of deterioration, that still leaves debt on a historically elevated path. The likely conclusion from the evidence is straightforward: AI may ease the pressure, improve efficiency, and buy governments some time — but it will not do the heavy lifting. Stabilising public finances still depends on policy choices about spending, taxation, and long-term fiscal discipline. Source: CBO, “The Budget and Economic Outlook: 2026 to 2036” (Feb. 11, 2026). (cbo.gov)