Rewiring intelligence: The race to power AI’s digital backbone

The world’s need for computing capacity is far outpacing the infrastructure built to support it and we can’t build what we need quick enough. Artificial intelligence has unleashed the biggest deployment of capital in human history.

Global AI power demand is projected to surge from 82 GW in 2025 to 219 GW by 2030, a 21.7 percent annual growth rate. The race to keep up is transforming both the scale and geography of the world’s digital backbone.

This explosive growth is already testing the limits of the systems it relies on.

As Tai Hollingsbee point out on GHD’s Transform podcast series on AI data centres, in some markets, power connections can face up to seven-year delays, while land constraints and community pushback in urban areas often slow planning approvals. The evidence suggests the digital economy is outgrowing its physical foundations and the energy needed to keep it running.

When resources lag, innovation accelerates. Developers are now finding new ways to build faster, smarter and closer to where power is available.

Decentralised, off-grid and prefabricated data centre models are gaining momentum, with offshore and remote facilities emerging to bypass the congestion of urban power networks. Prefabricated and scalable designs are also shortening delivery timelines and improving schedule certainty.

One of the most significant shifts is happening behind the meter which describes how data centres generate and store their own energy, often through renewables and large-scale battery systems, instead of relying on overstretched grids. Energy now accounts for around 60 per cent of total operating costs for AI servers, which consume five to ten times more power than conventional systems.

As resources tighten in major cities, we’re starting to see that developers are turning to regional and “edge” locations which are smaller, decentralised facilities built closer to where data is generated and used. These areas can offer available land and low-cost power faster approvals and improved fibre connections.

Globally, the edge data-centre market is projected to reach USD 45.1 billion by 2029, growing at a 32.8 per cent annual rate. North America is expected to account for nearly half of that expansion.

Speed and scale are now essential in the AI race. Prefabricated data centres, built off-site using scalable components, can dramatically shorten the journey from concept to operation.

Instead of constructing facilities from the ground up, developers are assembling them like engineered systems of parts: factory-built blocks shipped to site, connected and energised in record time.

Prefabrication allows around 80 percent of construction to be completed off-site, reducing schedule risk and labour constraints.
By industrialising construction, we can deliver data centres 20 to 50 percent faster and with greater precision. It’s the only way to match the pace of AI demand.

The approach also strengthens sustainability outcomes. Controlled factory environments often reduce waste, optimise materials and allow blocks to be reused or upgraded throughout their lifecycle. Prefabricated systems can be scaled, relocated or decommissioned with minimal disruption, giving operators flexibility as technology and capacity requirements change.

These prefabricated models are particularly effective in regional and edge locations, where rapid deployment and standardised performance are essential to meeting growing compute demand.

This decentralisation is unfolding against a backdrop of extraordinary financial momentum. While McKinsey projects USD 6.7 trillion in global data centre investment by 2030, Citi estimates a further USD 2.8 trillion in AI-specific infrastructure spending by 2029, while Goldman Sachs forecasts a 165 per cent surge in power demand over the same period. Together, the numbers tell a clear story: capital is flooding into digital infrastructure at unprecedented speed.

For investors, this is both a race and a recalibration. This is the biggest deployment of capital in human history. The early investors who pick the right bets will do very well. The internet took 20 years to develop; AI will do it in less than five and the rewards will likely follow that curve.

A parallel shift is occurring in the source of capital itself. Estimates suggest that between USD 68 trillion and USD 84 trillion in wealth will transfer from Baby Boomers to younger generations in the coming decades.

Much of it is likely to flow into ESG-aligned funds, favouring projects that combine digital performance with social and environmental value.

Despite their technical complexity, the future of data centres is as much about people as it is about machines. The industry’s social licence will depend on how it engages with the communities around it.

Data centres consume land, water and energy. Yet they can also give back through renewable energy partnerships, education programs and skills pipelines that prepare people for the digital economy.

AI’s expansion is not just a technological shift; it is a once-in-a-generation transformation of global infrastructure and society. The next five years will define the future of digital infrastructure. Whether the industry can build smarter, cleaner and more inclusive systems will determine not only the pace of AI innovation but also the shape of the global economy it powers.