The excitement around artificial intelligence can feel like a once-in-a-generation investment opportunity. Companies are pouring enormous sums into data centres, specialised chips and electricity infrastructure to power AI's rapid growth. More than 180 years ago, something remarkably similar happened with railways — and the parallels are worth heeding.
The 19th-century railway boom transformed transportation and reshaped the global economy. But it was also accompanied by rampant speculation, excessive investment and steep losses for investors.
Building railways in the 1840s was extraordinarily expensive. Investors financed track, bridges, tunnels, stations, locomotives and land, often well before passenger or freight demand justified the outlay. Today's AI infrastructure is similarly capital intensive: the modern equivalent of railway tracks is the vast network of data centres, chips, power generation and communications infrastructure needed to run increasingly sophisticated AI systems.
The numbers are staggering. Global investment in AI infrastructure is expected to reach into the trillions of dollars over the coming years, as technology companies race to build ever-larger computing networks — much as railway companies once raced to lay track across the country. The key question for investors is whether demand will grow fast enough to justify this spending.
Competition defined the railway boom. Companies rushed to build lines, sometimes duplicating routes between the same towns, drawn in by optimistic forecasts of passenger and freight volumes. Eventually, capacity outstripped demand. AI could face a similar problem. The technology's potential is enormous, but the infrastructure is being built at extraordinary speed — previous technological transformations typically unfolded over decades, while today's AI build-out is attempting to establish global computing infrastructure on a massive scale in roughly a decade.
Companies may spend billions on data centres and chips today, only to find that AI revenue doesn't arrive quickly enough to deliver attractive returns. That doesn't mean AI will fail — it means some of today's investments may not generate the returns investors currently expect. That distinction, between the promise of a technology and the returns on investing in it, matters enormously.
Railways were a transformative invention: they cut transportation costs, opened new markets and reshaped economies. But that didn't make every railway company a good investment. Many failed when revenues fell short, even as the underlying technology succeeded. The same could happen with AI. It may prove one of the most important technologies of the 21st century, capable of lifting productivity and creating new industries — but that doesn't guarantee today's AI stocks will all deliver exceptional long-term returns. When expectations run high, share prices can reflect years of anticipated growth already priced in. If profits fall short of those expectations, even amid genuine technological success, investors can still lose money.
This is one of the enduring lessons of past investment booms. Railway share prices eventually collapsed as enthusiasm met financial reality. Yet the tracks didn't disappear — the bridges, stations and networks remained, underpinning decades of subsequent growth. The boom was wrong about near-term financial returns but right about the technology's long-term value. The same distinction may apply to AI: even if some companies fail or share prices fall, the data centres, computing networks and electricity infrastructure built today will continue supporting economic activity long after today's enthusiasm fades.
What this means for investors
This isn't a prediction that the AI boom is about to collapse. It's a reminder that great technologies can still produce poor investments when expectations and prices outrun reality. Many early railway investors didn't share in the economic gains their investments helped create, simply because they paid too much.
AI may follow a similar path — delivering real long-term economic benefits while some of today's most highly valued companies disappoint investors. The lesson isn't to avoid AI, nor to chase the excitement by concentrating heavily in it. It's to stay disciplined, diversified, and focused on long-term goals.
Dr Steve Garth
23-Sep-2026