In the past two years, billions have been earmarked to flow into data centres. In Europe, power-connected land near London, Frankfurt, and Dublin is now changing hands at prices that would have been unthinkable five years ago. In some cases, developers are securing land before planning permission is granted, betting that grid access or government policy will eventually catch up.

In the United States, the Stargate project - backed by OpenAI, SoftBank, and Oracle - has announced plans to invest up to $500 billion in AI infrastructure. That’s roughly equivalent to the entire GDP of Norway. Even if only a fraction materialises, the scale is immense.

For some investors, the logic feels bulletproof: AI needs data, data needs servers, servers need homes- incredibly clean. Yet, this carries echoes of another moment in history. One when Britain poured its hopes and capital into a world-changing technology. And got badly burned.

In the 1840s, Britain was already riding the first wave of railways. Steam trains had slashed journey times, transformed trade routes, and fired the imagination of a population in the midst of industrial change. Railways weren’t just infrastructure, they were modernity and they were going to change the world.

Investors saw money could be made. In 1844, Parliament authorised projects covering 2,800 miles. In 1845, that jumped to nearly 9,500 miles - that’s the distance required to drive around the Island of Great Britain TWICE. 

Shares could be bought with just a 10% deposit, meaning investors could control vast sums of railway stock on leverage. The rest would be “called” once construction began. But in a booming market, few imagined needing to pay more. They planned to sell the shares on for a profit long before then.

It wasn’t just industrialists. Shopkeepers, clergymen, widows, and clerks all piled in. Local newspapers advertised dozens of schemes a week. Some projects were grounded in reality: routes linking growing cities and ports. Others were harder to justify.

Take the Great North of Scotland Railway to build a full double-track line from Aberdeen to Inverness at a cost of £1.5 million. It never ran as planned. Funding dried up; construction was delayed until 1852; only reached Huntly by 1854; and extended to Keith (around half way) by 1856- with Inverness still years away.

Indeed, by the end of 1845 examples like the above were widespread. Construction costs were far higher than expected. Tunnels and bridges ate through budgets. Labour shortages pushed wages up. And crucially, Parliament stepped in - requiring that companies prove they had raised all their capital before beginning work.

Suddenly, investors who had paid just 10% were on the hook for the other 90% (plus interest). Many couldn’t pay. Share prices collapsed. By 1850, railway stock values had fallen by over 70% from their peaks (for reference, the fall of the Nasdaq during the dot com bust was 75%; the fall of land prices in Tokyo in the early 90s, peak to trough was circa 80%) .

The infrastructure did transform Britain - but not without leaving behind a trail of bankruptcies, ruined reputations, and abandoned dreams (as well as some new millionaires).

Fast forward to today. Data, AI, cloud computing, and the web are the steam railway of today. The need is real. But the capital pouring into them is reaching historic proportions.

Projects are being priced on expectations of uninterrupted, high growth for years to come. Some land is being bought speculatively, without planning permission - and gambling that power access, growth in power capacity, demand, and government policy (and voter acceptance) will align.

The Stargate initiative - at $500 billion -would have sounded like science fiction a few years ago. It could reshape computing. Or it could overbuild capacity that may become redundant.  But in the meantime, everyone is piling in.  Here is a stat, the yield on a data centre today in Germany is the lowest yield of all property sectors bar senior housing, according to CBRE.

And there’s another risk: technology leapfrogging the asset altogether.

Several companies are now working on orbital data centres - computing hubs that would sit in low Earth orbit, powered by solar energy, cooled by the vacuum of space, and connected via high-speed optical links. Axiom Space, in partnership with Red Hat, is launching a prototype to the International Space Station this year. Other ventures are exploring lunar and deep-space nodes.

If such facilities become cost-effective, they could bypass many constraints that terrestrial data centres face - high land costs, local planning battles, and grid bottlenecks. In that scenario, billions invested in today’s earthbound hyperscale facilities could find themselves competing with infrastructure that orbits the planet.

It’s a reminder that even “essential” infrastructure can be undermined by a change in the technological baseline. For some 19th-century Highland lines, the baseline changed when steamships, roads and motor cars offered faster or cheaper connections. In the 21st century, it could be satellites.

Today’s data centre boom is a real phenomenon. But history whispers: not all of it will make sense in hindsight. Some sites bought at a premium may never get power. Some multi-billion dollar campuses may serve less traffic than expected. And some may face competition from above - quite literally.

Technology changes. Human behaviour doesn’t.