Ask investors what investing in artificial intelligence (AI) means and until recently, the same familiar names would have been mentioned: companies such as Nvidia, Amazon, Microsoft and Alphabet, which have dominated both headlines and portfolios as investors chase the technology driving the AI revolution.
But as AI moves from the experimental and developmental phase towards widespread commercial adoption, a new investment story is emerging. The next phase of AI is not simply about creating smarter models. It is about building the enormous physical infrastructure needed to run them.
Just as the industrial revolution required railways, canals, factories and steam power, today’s AI revolution needs data centres, semiconductor manufacturers, upgraded power networks, cooling systems, cybersecurity and water infrastructure. For investors wondering how to gain exposure to AI without concentrating purely on the mega-cap tech stocks, a much broader range of opportunities is beginning to emerge.
What exactly is a hyperscaler?
The companies dominating today's AI race are often referred to as ‘hyperscalers’ - these are technology firms that operate cloud computing platforms on a massive global scale.
The largest include Amazon Web Services (AWS), Microsoft Azure, Google Cloud and Meta. Their enormous computing capacity allows businesses to access AI models through the cloud rather than build their own infrastructure.
The scale of investment due to be unleashed by the hyperscalers is unprecedented. The largest are expected to spend around US$700 billion on AI infrastructure and data centres in 2026 alone. This is almost six times what they were spending when ChatGPT first appeared in 20221. This spending is expected to ripple through the global economy.
AI’s biggest constraint may now be physical
Ironically, AI's greatest challenge may no longer be developing better algorithms. Instead, the bottlenecks are becoming physical – real estate, hardware, electricity sources and water.
The International Energy Agency estimates that electricity demand from data centres rose by 17% during 2025, with AI-focused facilities growing even faster. It expects electricity consumption from AI data centres to continue climbing sharply over the rest of the decade, placing increasing pressure on power grids and transmission networks and creating opportunities for infrastructure builders.
Utilities companies, electrical equipment manufacturers and companies specialising in high-voltage transformers, transmission cables and grid management software could all benefit from the years of investment ahead required to modernise ageing electricity infrastructure.
In most developed economies, the electricity grid is completely unprepared for clusters of AI data centres consuming hundreds of megawatts around the clock, so major upgrades will be needed that will hopefully also benefit communities and other businesses, as AI accelerates the need for new energy sources and technologies.
Cooling: the essential AI necessity
Powerful AI processors generate significant amounts of heat. Keeping thousands of advanced chips operating safely requires sophisticated cooling technologies, making thermal management one of the fastest-growing segments of the AI supply chain.
Traditional air cooling is increasingly being replaced by more efficient cooling technologies, such as liquid cooling systems and specialist heat exchangers. Liquid cooling means water infrastructure will become increasingly important. Many data centres rely on significant quantities of water for cooling, creating demand for water treatment and recycling plants that can meet stringent environmental requirements.
The demand for semiconductors
Memory chip manufacturers, networking specialists and chip equipment suppliers all play essential roles in the AI supply chain. Years of cyclical underinvestment following the post-pandemic semiconductor downturn have left parts of the industry racing to expand capacity just as AI demand accelerates. Building new fabrication plants takes years rather than months, meaning supply constraints may persist for some time.
It is worth noting that many of these businesses are based beyond the US, so investors can gain broader geographic exposure to the AI theme.
AI is becoming more autonomous
Another reason that the demand for infrastructure continues to grow is the ongoing evolution of AI itself.
Early generative AI typically answered one question at a time. Increasingly, AI is moving towards so-called agentic AI: systems capable of planning, reasoning and carrying out multi-step tasks with limited human supervision.
Products such as OpenAI’s Codex and Anthropic’s Claude Code illustrate the shift from simple chatbots towards digital assistants that can complete increasingly complex work.
These systems require significantly more computing power and cloud infrastructure while creating opportunities for every kind of company that can integrate AI into its business processes.
The potential brakes on the buildout
Building data centres at the scale required involves land, electricity, water and planning approval. Increasingly, there is significant pushback from local communities.
A recent Gallup survey found that seven in ten Americans oppose AI data centres being built in their area because of concerns over the environmental impact and water and electricity consumption.
In addition, delays in renewing grids, plus shortages of materials and skilled labour, are all likely to slow the buildout, proving that even the strongest structural investment themes will not progress in a straight line.
Looking beyond the biggest names
AI remains one of the defining structural growth stories of the coming decade. And the investment opportunity is becoming much broader that owning the companies creating the AI models.
History suggests that during technological revolutions, some of the most durable returns come from the businesses providing the ‘shovels’ – the essential infrastructure – rather than the consumer products.
As AI scales from millions of users to billions, investors may increasingly find opportunities in the companies building the roads, power networks and tools that make the next revolution possible.