The future of AI depends on how we design the energy infrastructure that powers it

Every transformative technology eventually runs into the same contradiction: its potential outpaces the infrastructure built to support it. Artificial intelligence (AI) is now at that point. Its ability to transform how we work, discover, create, and solve problems could benefit virtually every part of the world, yet the physical systems required to sustain its extraordinary computational demands are still catching up. The gap is particularly visible in energy. More than 666 million people worldwide still lack access to electricity, while the energy infrastructure supporting today’s technology ecosystem is under pressure. Data centres consumed around 415 TWh of electricity globally in 2024, and the International Energy Agency (IEA) expects that demand to more than double by 2030, reaching roughly 945 TWh. The infrastructure that was sufficient for today’s computing is, therefore, being asked to support a radically different scale of computation.

AI’s ambitions demand a corresponding leap in the energy infrastructure that powers it. Expanding data centre capacity without strengthening energy generation, transmission, storage, and grid resilience risks further pressuring systems that businesses and communities already depend on. Energy capacity and efficiency must, therefore, be integral to AI’s growth, not considerations addressed after the computing infrastructure is built.

This calls for a more integrated approach to data centre design, where location, power, cooling, and the surrounding energy network are considered together. The objective is not simply to build more capacity but to create facilities that can scale efficiently and operate as part of a wider energy ecosystem.


The physical cost of an intelligent world

Much of the conversation around AI has understandably centred on chips, models, and computing power. Yet behind every model, every inference, and every AI application sits a physical environment that must supply enormous amounts of electricity, remove heat, maintain uptime, and accommodate increasingly dense computing.

The scale of that infrastructure is changing quickly. The IEA expects electricity use in data centres to grow more than four times faster than electricity demand from other sectors between 2024 and 2030. In the US, data centres could account for nearly half of electricity demand growth over the same period. These facilities are, therefore, becoming some of the most concentrated new sources of electricity demand in the markets where they operate.

That concentration matters because electricity systems are not infinitely elastic. When large computational loads arrive in regions where generation, transmission, or distribution capacity is already constrained, the pressure must be absorbed somewhere through new infrastructure, investment, and, in some markets, higher electricity costs. The challenge is, therefore, larger than finding enough power for individual facilities. It is about creating an energy ecosystem capable of accommodating a new category of demand without making existing access more difficult or less affordable.

The data centre is an energy asset

The industry is racing to build AI computing capacity, but the real constraint may lie in how data centres are designed to use and access energy. A facility can house powerful computing infrastructure and still struggle to scale if its architecture fails to account for power availability, rising heat loads, and cooling demands. Energy, therefore, cannot remain a utility decision made around the building. It must shape the architecture itself.

This means bringing site selection, spatial planning, electrical infrastructure, cooling systems, and renewable energy into a single design approach. The placement of computing equipment, the density of racks, the movement of heat, and the scope for future expansion all influence how efficiently a facility consumes energy and how readily it can adapt to changing demands.

The industry is increasingly investing in renewable power purchase agreements, battery storage, and smarter energy management. The IEA expects renewables to account for nearly half of the additional electricity generation required by data centres through 2030. But the energy transition must extend inside the facility, through architecture and engineering that reduce cooling loads, enable heat recovery, and make better use of every unit of power.

In the AI era, data centre architecture must do more than accommodate computing. It must determine how efficiently computing is powered, cooled, and scaled. Energy is no longer separate from the design brief. It is one of the forces defining it.

Efficiency begins with design

Data centres cannot solve the energy challenge by increasing supply alone. It also depends on how intelligently facilities are designed to use that power. AI is driving higher computing densities, concentrating more heat into smaller spaces, and placing greater demands on cooling and electrical systems. Technologies such as direct liquid cooling can help, but their effectiveness depends on the surrounding architecture.

Rack layouts, cooling routes, power distribution, floor loading, and maintenance access all influence one another. When planned in isolation, they can lock inefficiencies into the building and make future upgrades costly.

The design challenge is to make the building, its engineering systems, and its computing infrastructure work as one, accommodating rising densities and new technologies without repeated redesign. Efficiency is not simply installed through better equipment. It is designed into how the entire facility works together.

Designing for the energy beyond computation

Data centres have traditionally treated the heat they generate as waste to be removed. But as facilities grow, that heat could become a resource. Data centre heat is already being explored for district heating, greenhouses, and commercial use. In Denmark, Microsoft has incorporated surplus heat recovery into projects that can feed local district heating networks.

This opens up a broader way to think about site selection. Beyond land, connectivity, and electricity, a data centre’s value could also depend on its relationship with the surrounding energy and urban ecosystems. By combining efficient power use, energy storage, and heat recovery, facilities can function not as isolated buildings, but as active participants in the surrounding infrastructure.

The opportunity is to make energy efficiency a property of the entire asset, from the wider energy network to the placement of a server rack. As AI pushes the limits of computing, data centre design must move beyond accommodating more capacity to making that capacity viable, adaptable and efficient over time. The next generation of data centres will be defined not simply by how much computing power they house, but by how intelligently their physical infrastructure turns energy into computing performance.

The author is Founder of Flipspaces. Views are personal