A network growth model published in Scientific Reports concludes that Europe could expand its high-speed rail network from 10,951 km to 24,154 km for €269 billion in 2023 euros, generating a net present value of €563 billion and a benefit-cost ratio of 3.

The study, which covers 28 countries including EU members, Norway, Switzerland and the UK, uses an iterative investment model dubbed ENGINEER. It starts from the existing and planned 2023 network, applies a four-step transport demand model across major urban hubs, and adds the highest-scoring candidate link each year under a budget constraint calibrated to historical European HSR spending. Construction costs are estimated on a 3D hexagonal grid using terrain data, population density and national price levels, allowing surface and underground alignments within a 2.5% gradient limit.

The results point to a network roughly half the length of the TEN-T Comprehensive 2050 plan, but one the authors argue is economically coherent.

Where the model builds

Growth concentrates in Central Europe — Germany, Poland, Czech Republic, Austria and Switzerland — plus the UK. Spain, whose network is already extensive, receives little or no new investment. Denmark, Estonia, Latvia and Lithuania see no extensions at all. Peripheral links such as Athens–Bucharest, Stockholm–Bergen, Stockholm–Gothenburg and Lisbon–Porto still make the cut because of large travel-time savings.

The model sequences investment in three phases. Early projects close gaps between national sub-networks where demand already exists: Brussels–Antwerp, Düsseldorf–Ruhr and Mannheim–Frankfurt. A second phase extends long-distance connections such as Stockholm–Gothenburg and Katowice–Kraków. A final densification phase adds alternative routes and shortcuts around major hubs, including Genova–Florence and Ghent–Antwerp.

One outlier is Leeds–London, built as early as 2029 on the strength of demand and value of time.

The analysis shows strong path dependency. Warsaw–Łódź makes Łódź–Katowice viable the following year in 2027; Luxembourg–Saarbrücken (2039) unlocks Saarbrücken–Karlsruhe; Milan–Genoa (2048) unlocks Genoa–Florence. Prior travel-time reductions raise demand enough to justify the follow-on link.

Market share shift

By 2065 rail's share of inter-city long-distance trips rises from 13% to 27% in the model. The break-even distance at which rail and air hold equal market share moves from 380 km to 640 km, extending rail's competitive reach over longer journeys.

The largest jumps in rail share coincide with cross-border openings: Stuttgart–Munich and a Turin–Lyon corridor via Grenoble in 2043, multiple Central European cross-border links between Poland, Austria, Czech Republic and Germany in 2046–47, Italy's enhanced connections with Austria and Switzerland — Munich–Innsbruck, Trento–Verona, Milan–Zurich — in 2051 and 2058, and a Wroclaw–Prague–Nuremberg link in 2063.

The authors note that most key national links already exist in the base network, so cross-border projects dominate the gains. They argue such projects are difficult to justify within national appraisal frameworks because benefits accrue beyond one country's borders.

Winners and losers under central planning

A Netherlands experiment illustrates the case for wider appraisal. Constrained to Dutch borders, the model builds Amsterdam–Utrecht–Eindhoven. With a cross-border view, it builds Brussels–Antwerp — a link entirely outside the Netherlands that nonetheless yields more benefit to the country — and Amsterdam–Ruhr. A full European perspective reroutes Amsterdam–Ruhr via Liège and Aachen to serve Stuttgart, Zurich and Milan.

The centralised budget mechanism produces uneven national outcomes. The UK, France, Italy, Spain and the Netherlands contribute most relative to direct benefits. The UK still returns a positive NPV; others sit near break-even or negative. The authors flag this as a political acceptance risk and suggest future model versions could add break-even constraints for every participating country.

Divergence from TEN-T

Benchmarked against the TEN-T milestones, ENGINEER's 2030 network is more conservative than the Core Network, skipping Alpine projects and long extensions to the Baltics, Romania, Bulgaria and Greece, while identifying short high-return segments TEN-T does not list — Mannheim–Frankfurt, Aachen–Köln, Katowice–Kraków. It also builds in Sweden, Finland and the UK, only marginally covered by TEN-T's core layer.

By 2040 the model completes the Paris–Strasbourg–Stuttgart–Munich–Vienna–Budapest corridor ahead of TEN-T. By 2050 it omits the full Baltic mainline, Budapest–Bucharest and further Iberian extensions, while including corridors TEN-T leaves out, such as Milan–Genoa–Florence, Lyon–Bern–Zurich–Stuttgart, Oslo–Stockholm and Athens–Sofia–Bucharest.

The authors caution that even this expanded network would double rather than triple HSR ridership, the EU's stated 2050 goal — implying significantly higher investment and faster construction than historical trends deliver. The model omits intermodal hubs, mixed freight-passenger operation and strategic or military uses of rail infrastructure, and the authors acknowledge it may understate cross-border demand by not fully capturing latent and induced demand. Future work will examine service design variables — lines, frequencies, capacities and prices — alongside infrastructure expansion.