Europe 2045: About and Methodology

Europe 2045 is a fictional scenario in which the European Union implements a series of reforms that address its current problems, eventually bringing about a strong middle-power coalition built around the principles of growth, technological development and social progress.

The project started as a series of conversations with friends and peers around a set of fundamental questions:

  • What would it take to build a technologically dynamic Europe with a coherent industrial policy and a stronger geopolitical position?
  • Would it be possible to build this Europe without sacrificing the European social model or its sense of identity?
  • In fact, might technological and economic development, notably the automation brought by AI, allow us to combine these two objectives and advance social progress?

While there are sections where we sacrifice a degree of verisimilitude in exchange for painting a picture of what such a Europe could be like, the scenario remains plausible as one outcome in the diffuse light cone of all possible future worlds.

Margin is committed to social equity and progress. We believe these goals are best achieved through economic development and the advancement and diffusion of new technologies.

At present, this means achieving a strong level of sovereignty, investing in infrastructure and developing the European economy — while ensuring the fruits of innovation and automation are shared by every citizen. It does not mean sacrificing the European social model for the benefit of a minority who own equity in technology companies or increasing dependence on those without European interests at heart.

This leads us to four priorities:

  • COMMON INVESTMENT. A pan-European industrial policy, even where that strains the four freedoms and state-aid rules. Investment for growth, with the returns shared broadly across the population.
  • ACCELERATION FOR ALL. Embrace automation. Reinvest gains from technological acceleration into reinforcing the European social model.
  • CLEAN, SOVEREIGN GRID. Continued clean-energy build-out, now as much a matter of sovereignty and economic security as of climate policy.
  • MIDDLE-POWER COALITION. Enlargement and deeper partnerships, with the aim of building a strong middle-power bloc that can defend its interests and bargain from a stronger position between the superpowers.

We welcome feedback and input.

Authors

Margin is a new-media and research studio based between London and Berlin founded by Sam Diamond and Son La Pham.

In a world where culture and politics have become inseparable from finance and technology, we explore questions of power and sovereignty, technological acceleration and economic justice.

We specialise in new media, research and narrative design.

If you want to stay up to date on our work, follow us on X and Instagram, and join our mailing list. You can also reach us directly by email.

Thanks

Special thanks to Arthur Röing Baer, Beth von Undall, Cullen Miller, Francis Tseng and Guy Mackinnon-Little for their advisory help in developing this report.

About the model

While the scenario is fictional, it is underpinned by a comprehensive quantitative model based on current baseline numbers and research by a number of individual economists, policy shops, think-tanks and statistics agencies. The model tracks various economic metrics of the EU economy year by year, including output, jobs, incomes, investment, energy, AI computing capacity and public finances.

It describes one scenario, in which the reforms in the essay are carried through and are largely effective.

The model covers the 27 EU member states. Countries that join Europa+ are not added to EU figures. Growth, incomes and electricity prices are adjusted for inflation and given in 2026 prices. Amounts invested and held, such as Esia’s annual spending and the EUSWF’s assets, are in the euro value of each year.

Starting data

Measure2025Source
GDP€18.9trn, growth of 1.5%Eurostat
Population452m (January 1st 2026)Eurostat
Employment rate, ages 20 to 6476.1%Eurostat
Business electricity price€160 an MWhEurostat
Share of energy imported57% (2024)Eurostat
Renewable and nuclear share of electricity70%Eurostat
Investment21.4% of GDPEurostat
Income inequality (Gini)29.2Eurostat (EU-SILC)
Greenhouse gas emissions2,786m tonnes (2024)European Environment Agency
AI computing capacityAbout 30GW worldwide, about 2GW in the EUEpoch AI
EU debt later taken over by Esia€739bnEU annual accounts

The 2026 estimate encompasses the start of the essay’s crisis period in the winter of that year: growth of 1%, an electricity price of €190 an MWh, and inflation of 3.4%.

Key assumptions

Economic growth

Output per hour grows by an assumed 1% a year from underlying progress. The programme in the essay adds about one point a year on top of that, from more investment, cheaper energy, the Special European Zones, new transport and grid links, and reform of the single market, planning rules and capital markets. Reform is the largest of these. It raises output by 8% over time. For the single market we use the ECB’s estimates of the long-run welfare gains from lowering barriers to trade inside the EU: 1.3% for goods and 1.8% for services. The ECB models the two separately; adding them together and counting the 3.1% as extra output is our simplification. The remaining 4.9%, from planning and capital-market reform, is our own assumption. Investment rises from 21% to 28% of GDP, more than the extra 4.4 to 4.7% of GDP a year called for in the Draghi report, with much of the difference going on AI hardware.

GDP growth

A recession in 2027–28, then growth builds to a 4.0% peak in 2035. After a downturn in 2040, it settles near 2.7% by 2045.

% a year
2045
  • Real GDP growth +2.7%

Where growth comes from

AI adds most in 2035, 2.3 points. From 2039, investment, the SEZs, cheap energy, networks and reform add more between them: 1.2 points in 2045, against AI’s 0.6.

Percentage points of real GDP growth
2045
  • Background productivity +1.0
  • Investment +0.4
  • SEZ innovation and new firms +0.2
  • Cheap energy and industry +0.1
  • Rail, transport and grid networks +0.0
  • Single market and reform +0.5
  • AI, including ModelEU +0.6
  • Jobs and hours −0.1
  • Real growth +2.7

GDP per person and typical income

GDP per person ends 45% higher than in 2026, the middle fifth’s income 34% higher.

Index, 2026 = 100
2045
  • Real GDP per person 145
  • Typical household income 134

Income by fifth of households

Every fifth is poorer after the crisis and back above 2026 by 2033–35. By 2045 the poorest fifth is up 43%, the richest 20%.

Real disposable income per person, 2026 = 100
2045
  • Poorest fifth 143
  • Second 140
  • Middle 134
  • Fourth 127
  • Richest fifth 120

AI and productivity

The effect of AI is calibrated to Anthropic’s Economic Scenarios for Transformative AI (Korinek, Jones, Sacher, Cotter and McCrory, 2026). In its “substantial” scenario, American output is 8.3% higher by 2030 than it would be without AI, which works out to about 9% more output per worker. At that point about 12% of tasks use AI. We take 65% of that effect for each task that uses AI.

After that, the model goes further than the source. The share of tasks using AI keeps rising to 2045, following our own assumptions about AI capability, European adoption and ModelEU, and each further increase adds less than the one before. By 2045 AI has raised output per hour by about 23% compared with 2026. That extrapolation is our own assumption.

In the model, AI adds 1.2 percentage points a year to productivity growth between 2030 and 2045. That is within the published range, at the optimistic end of the European estimates. The estimates below measure different things over different periods, so they give a sense of scale only:

EstimateMetricAdditional growth, points a year
Acemoglu (2024)Total factor productivity, America, next ten years0.07
Bergeaud (2024), ECB Forum on Central BankingTotal factor productivity, euro area, next ten years0.3
Aghion and Bunel (2024)Productivity, next ten years, by analogy with electricity and computing0.8 to 1.3
OECD (2025), Germany, France and ItalyLabour productivity, next ten years, rapid-adoption scenario (0.2 to 1.2 across all scenarios)0.9 to 1.2
Goldman Sachs (2023)Labour productivity, America, ten years after widespread adoption1.5
Baily, Brynjolfsson and Korinek (2023)Labour productivity, America, ten years, illustrative scenarios1.7 to 2.7
Our modelOutput per hour, EU, 2030 to 20451.2

AI’s lift to output per hour

23% by 2045, rising more slowly once past the range the calibration covers. ModelEU’s own models add about 8% on top of commercial AI.

% above 2026
2045
  • All AI 22.7%
  • ModelEU’s own addition 7.8%

Workers in transition

Displaced workers waiting for new jobs peak at 0.3 million in 2035. From 2035, 0.7–1.6 million a year change occupation because of AI.

Million people
2045
  • Displaced, awaiting work 0.15
  • Changing occupation, a year 0.75
  • In public retraining 0.58

AI and compute capacity

Compute capacity is measured in gigawatts of data-centre power, the unit used by Epoch AI, the International Energy Agency and the Europe 2031 compute forecast. The starting point is Epoch AI’s estimate of about 30GW worldwide at the end of 2025, about 2GW of it in the EU.

Europe’s capacity grows with what is spent on it each year: public investment through Esia, plus private investment that follows announced projects. A gigawatt costs about €35bn in 2026. The energy crisis slows the early build-out, so the EU has 6.8GW in 2031, against 21GW in Europe 2031’s central case. It reaches 79GW by 2045, about a tenth of the world total, after Europe re-equips early for the 2037 shift to new AI hardware.

AI computing capacity in the EU

79 GW by 2045, about 10% of the world’s. A physical measure, not of performance or ownership.

GW of facility power
2045
  • EU capacity 78.8

Energy

The energy system loosely follows the strategy in Ember’s New Generation (2022), which sets out a least-cost route to clean European electricity. As in Ember’s pathways, wind and solar supply most of the power. Surplus renewable electricity is used to make green hydrogen, and hydrogen-ready turbines provide back-up for calm, dark winter weeks. Existing nuclear plants are refurbished but no new reactors are built.

Ember’s pathways reach about 95% clean electricity by 2035. The model gets there more slowly because of the crisis: renewables rise from 47% of electricity in 2026 to 82% in 2045, and by 2045 all electricity is low-carbon. Prices are built up from the cost of generation, grids and back-up, and the industrial price, in 2026 prices, falls from €190 an MWh in 2026 to €116 in 2045. Emissions end 93% below their 1990 level.

Electricity by source

82% renewables, 12% nuclear, 4% gas with carbon capture and 2% hydrogen in 2045.

% of delivered electricity
2045
  • Renewables 82.5%
  • Nuclear 11.5%
  • Gas with carbon capture 4.5%
  • Hydrogen 1.5%
  • Unabated fossil 0.0%

Net emissions by source

By 2045 power emissions fall 98% and fuel burned outside power 97%. Farming, industrial processes and waste are what remains.

Million tonnes CO₂ equivalent
2045
  • Power 10
  • Fuel burned outside power 45
  • Farming, processes and land, net 266
  • Climate Law path 230
  • Total 321

What the business electricity price is made of

Generation costs fall from €117 to €44 as gas stops setting the price. Network costs edge up only from €35 to €37, spread over more electricity.

€/MWh, 2026 prices
2045
  • Generation 44
  • Network 37
  • Backup capacity 10
  • Taxes, supply and crisis premium 25
  • Total 116

Finance

Borrowing costs start at about 3.2%, close to the EU’s own funding cost in 2025, and then move with the scenario. They rise with inflation above 2%, with the investment boom and with higher world interest rates in the 2030s, and ease once reserve managers start buying Esia bonds as a safe asset after the EU–China agreement. Esia pays a margin on top, which narrows from 0.75 to 0.25 points as its bonds become established, so it borrows at between about 3.3% and 3.7% from 2029 onwards. The fiscal rules are suspended during the crisis and rewritten from 2035 so that public investment does not count against the deficit limit. New taxes fall on economic rents, land, large inheritances and fortunes above €100m, following the Mirrlees Review’s design for taxing rents and the EU Tax Observatory’s proposal for a 2% minimum tax on the very rich. Together they raise about 4% of GDP by 2045. The sovereign wealth fund earns about 4% a year after inflation, in line with the long-run record of Norway’s and Singapore’s funds. Defence spending follows Nato’s 2025 target of 3.5% of GDP.

Public debt, including Esia’s, peaks at 111% of GDP in 2040 and falls to 100% by 2045.

People and incomes

Population follows Eurostat’s EUROPOP2025 projections, plus extra migration from 2032 through skilled worker routes built on the EU’s existing Talent Partnerships with Tunisia, Morocco and Egypt. Retraining, migration and the Erasmus Service Programme fill the skills gap, which peaks at 1.4m workers in 2033. Incomes are tracked for five income groups, calibrated to Eurostat’s measure of inequality, so the model can show how the gains are shared. We use the real disposable income per person of the middle fifth as our measure of the typical household. By 2045 it is about a third higher than in 2026.

Sources

A non-exhaustive bibliography and list of data sources that informed the piece.

AI and compute

Energy

Growth, public money and data

Back to top ↑