Could AI lift euro-area output 3.8% by 2030? The ECB result is a model scenario, not a forecast
The ECB’s new assessment mixes current investment and survey evidence with a simulated worldwide productivity shock. It finds broad wage gains but concentrated IT job losses only under those assumptions.
The short version
- What happened
- The ECB models a 30% productivity gain in IT services occurring worldwide and finds euro-area output 3.8% above baseline by 2030; this is a conditional scenario, not a forecast.
- Who it affects
- Workers may experience faster tasks without their employer converting every saved hour into extra output. The model also suggests that gains can coexist with concentrated adjustment in IT and clerical roles, making training, job design and income transitions more important than a single headline output figure.
- What remains unproven
- ECB policy speech
Keep in mind
ECB policy speech
30% global IT-productivity shock is assumed
Model paper forthcoming
2025–26 digital investment partly proxied
Worker time savings self-reported
Employed-worker response denominator not disclosed
No causal estimate of economy-wide AI productivity
Scenario is not a forecast
Source analysis · 2 linked sources. Primary analysed source date: 30 October 2026. Earliest linked supporting source: 26 August 2026.
Read the full report · 5 min The evidence, methods, implications and limitations
The 3.8% figure begins with a large assumption
The European Central Bank has mapped how AI investment, workplace use and finance are moving through the euro-area economy. Its most eye-catching result is conditional: if total-factor productivity in software, IT consulting and data services rose 30% in every country as an unexpected, near-permanent change, the ECB’s model puts euro-area output 3.8% above baseline by 2030 and 4.1% higher in the long run. Long-run consumption is 5.5% higher.
That is not an ECB forecast of what AI will deliver. The MANKIO simulation represents 20 euro-area countries, the United States, an Asia group and the rest of the world, with 75 industries per country and 2023 input-output relationships. The assumed productivity shock is applied globally and all at once; different adoption speeds, prices, business responses or policy settings would change the result. The underlying model paper is described as forthcoming, so the complete specification has not yet received public journal review.[1]
Current data show spending and use—not a measured macro payoff
Some inputs are observed rather than simulated. The ECB estimates that real digital investment has grown 73% since 2014, from 2.3% to about 3.3% of euro-area GDP by the first half of 2026. The latest two years extend official annual data with a proxy based on digital-services output, so they are estimates rather than final investment accounts. Firms in an ECB telephone survey said AI added about 10% to their 2026 capital-expenditure plans, but the speech does not disclose the response count for that question.
Worker evidence also needs care. The ECB’s monthly Consumer Expectations Survey covers roughly 20,000 people across 11 euro-area countries; 52% of employed respondents reported using AI at work in 2026, up from 26% in 2024. Users reported a median three hours saved per week. The public analysis does not give the employed-worker denominator, and reported time saved is not the same as extra output: capacity may be unused, shifted to other tasks or offset by checking and correction work.[1][2]
The distinction that matters
Three evidence layers answer different questions
The ECB assessment combines measurements, reported experience and a macroeconomic scenario; they should not be read as one observed effect.
Observed
Investment, trade and market data
Shows where spending and financing are moving
Reported
Firm and worker survey answers
Shows stated use, plans and perceived time savings
Modelled
Assume a global 30% IT-productivity shock
Produces conditional 2030 output, wage and employment changes
The model spreads wage gains but concentrates job losses
Under the ECB’s assumed shock, cheaper IT services raise demand and wages across the model, but adjustment is uneven. IT employment is 4.4% lower by 2030 while real wages in that sector are 4.3% higher. Clerical employment outside IT is 0.9% lower and real wages 3.2% higher. Managers, professionals and technicians outside IT have no modelled employment change and a 4.4% wage gain; manual and service workers have no job change and a 5.4% wage gain. These are scenario outputs, not observed people losing or gaining jobs.
For workers, employers and policymakers, the practical message is to separate three evidence layers: recorded spending, self-reported use and conditional modelling. Confidence would rise with final national accounts, disclosed survey response denominators, measured output and error rates in real workplaces, and model results under slower or uneven adoption. Until those checks arrive, the ECB assessment is useful for stress-testing choices about skills, finance and infrastructure—but it does not establish that Europe is on a guaranteed path to 3.8% more output.[1]