The rapid modernization and expansion of digital infrastructure is creating a long-term opportunity for the construction industry, with McKinsey estimating that investment could reach approximately $19 trillion by 2040.
Much of the attention has focused on the billions of dollars being committed to data centre construction. However, the physical buildings themselves account for only about 25 per cent of the massive investment figures frequently highlighted in the media.
The remaining 75 per cent is linked to the technology housed inside these facilities. But as artificial intelligence expands — particularly through data centre connectivity involving fibre networks, satellites and cables — another critical requirement is becoming increasingly difficult to overlook: the infrastructure needed to generate and distribute enough electricity to power it all.
This could represent one of the construction industry’s most significant long-term opportunities.
“We are not paying enough attention to the less glamorous reality underneath it: the physical infrastructure required to make AI work,” writes George Sakellaris, founder and CEO of energy infrastructure solutions provider Ameresco.
The technology may be advancing at extraordinary speed, but its success ultimately depends on a reliable supply of electricity. As Sakellaris points out, even the most advanced AI models, software and chips cannot operate without sufficient power.
AI is effectively creating a new, electricity-intensive industry at a time when much of the existing power grid is still based on assumptions from an earlier era. Meeting the new demand will require expanded energy infrastructure, a modernized grid, additional generation capacity and greater energy storage.
Data centres cannot afford to wait
For North America, expanding power generation capacity to support the next generation of hyperscale data centres has become a time-critical issue.
In the United States, traditional grid expansion can take between five and 15 years, particularly when new high-voltage interconnections require approval. For many hyperscale developers, those timelines are simply too long.
As a result, some are taking power generation into their own hands by installing “behind-the-meter” generation directly at or next to their facilities.
Behind-the-meter generation means electricity is produced either on a data centre site or immediately adjacent to it and consumed directly by the facility. This allows operators to avoid relying entirely on the public utility grid and its retail electricity tariff structure.
International power generation designer, builder and operator US Power & Environment (USP&E) says the model can eliminate grid dependency, avoid utility interconnection delays, give operators greater control over fuel supply and generation costs, and allow facilities to continue operating during public grid outages or curtailment events.
The approach has already become a major power strategy among hyperscalers, bitcoin mining operations and AI infrastructure developers that cannot wait for grid expansion or require reliability levels beyond what utilities can contractually guarantee.
A submitted photo notes that Elon Musk’s behind-the-meter natural gas generation for the Colossus xAI data centre southwest of Memphis was completed in 122 days — a striking example of the speed required by the rapidly developing AI sector.
Natural gas remains the fastest option
The capital required to build behind-the-meter generation varies significantly depending on the technology selected. A modern 100 MW AI facility and a next-generation 1 GW supercluster can require dramatically different levels of investment.
According to USP&E, simple-cycle or reciprocating natural gas generation currently represents the least expensive option. Estimated construction costs range from $1,200 to $1,500 per kW.
For a 100 MW data centre, that translates into an investment of approximately $120 million to $150 million. A 1 GW facility could require between $1.2 billion and $1.5 billion.
Natural gas generation remains subject to debate over its environmental impact. Nevertheless, USP&E identifies it as the preferred fuel for behind-the-meter AI data centre generation.
The company points to several advantages, including high power density, lower emissions compared with diesel or heavy fuel oil, greater fuel-price stability through long-term supply agreements and the ability to combine gas generation with solar photovoltaic systems and battery storage.
Fast-track gas turbine installations for AI data centres generally take between 12 and 24 months from contract signing to commercial operation, depending on whether the project uses new equipment or verified surplus equipment.
In some cases, the timeline can be substantially shorter.
Natural gas generators supporting Elon Musk’s Colossus xAI data centre were reportedly operational in just 122 days.
Alternative technologies come with higher costs
The price of power generation increases when other technologies are considered.
A solar photovoltaic and battery storage system capable of providing 24/7 firming — combining photovoltaic arrays, large-scale batteries and additional generation capacity to ensure continuous electricity around the clock — can cost three times or more than a natural gas-based system.
The scale of the opportunity is only now beginning to emerge.
According to private market research and business consultancy S&S Insider, the global behind-the-meter market was valued at US$105 billion in 2025 and is expected to reach US$520 billion by 2035. That represents a compound annual growth rate of 18 per cent between 2026 and 2035.
USP&E describes understanding the power requirements of AI data centres as one of the most consequential planning questions facing the energy sector today.
For construction companies, that means the opportunity extends far beyond erecting data centre buildings. Power plants, generation equipment, energy storage, electrical infrastructure and associated systems could become an increasingly important part of the data centre construction ecosystem.
Canada faces the same growing demand
Electricity generation requirements are also expected to rise rapidly in Canada. However, Natural Resources Canada has so far placed greater emphasis on expanding the country’s traditional electricity grid.
The department says Canada enters this period with a significant competitive advantage: a reliable, low-cost and low-emission electricity system that ranks among the most affordable globally, supported by abundant and affordable natural resources that can help generate additional power.
Natural Resources Canada also notes that Canada’s electricity generation and distribution system has been developed over decades through sustained investment and leadership from provinces and territories, utilities, generators, system operators and ratepayers.
That existing foundation, the department says, must now be protected and strengthened as electricity demand rises and the global environment becomes increasingly complex.
Even if Canada’s approach means behind-the-meter generation develops more slowly than it has in the United States, the growing need for reliable power for AI and hyperscale data centres still represents a substantial and long-term opportunity for the construction sector.
The expansion of digital infrastructure is therefore creating a construction market that goes well beyond concrete, steel and data centre buildings. As AI continues to increase electricity consumption, the infrastructure required to generate, store and deliver that power could become one of the industry’s most important areas of growth.
