For much of the digital age, power appeared almost weightless. Data crossed borders, the cloud seemed to exist everywhere and nowhere, and artificial intelligence was discussed as though it lived entirely inside algorithms. That illusion is becoming harder to sustain.

AI has a geography, and an increasingly demanding one. It needs land, vast quantities of electricity, access to water, advanced semiconductors, and enormous data centres whose locations have become matters of national security. The race for AI supremacy is therefore becoming a contest over physical infrastructure and finite resources, with echoes of the geopolitical struggles that shaped earlier industrial eras.

Data centres are emerging as the factories of the twenty-first century. Countries with ample computing capacity, or “compute,” enjoy advantages extending well beyond commercial AI. That capacity supports scientific research, intelligence collection, cyber operations, defence innovation, and productivity across the economy. What once looked like a corporate infrastructure decision is now bound up with national sovereignty. The energy implications are already striking.

According to the International Energy Agency, global electricity consumption by data centres rose by approximately 17 percent in 2025. In the United States, data centres accounted for nearly half of the annual growth in electricity demand, and the IEA projects that their expansion will drive roughly half of the increase in American demand through 2030.

The scale of the shift becomes clearer in projections for the United States. Lawrence Berkeley National Laboratory estimates that data centres could consume approximately 11.8 percent of the country’s electricity by 2030, with scenarios ranging from 9.5 to 15.3 percent. This is changing what energy security means. Governments must still keep homes warm, factories running, and transportation networks moving.

Increasingly, they must also provide uninterrupted power to the machinery on which their digital ambitions depend. Grid capacity, transmission lines, gas turbines, nuclear reactors, and renewable-energy projects are no longer peripheral to AI strategy. They are part of its foundation.

Japan offers a revealing example. A private-sector consortium involving JERA, Dell Technologies, and RHAELM is planning an expansion of AI infrastructure that could entail as much as $140 billion in investment. Its initial project in Chiba, near Tokyo, is expected to cost approximately $15 billion and provide around 400 megawatts of data-centre capacity.

The ambition is unmistakably technological, but its success will depend on something more prosaic: securing a substantial, reliable electricity supply. Japan’s plans illustrate how the ability to build computing infrastructure is increasingly inseparable from the ability to power it.

Electricity, however, is only one constraint. Data centres produce enormous amounts of heat, and keeping their equipment cool can require significant water resources. Their precise footprint depends on cooling systems, local climate, computing workloads, and how the electricity they consume is generated. Some facilities place greater demands on water than others, and much of the pressure can be indirect. Yet in water-stressed regions, the political question is unavoidable: how should governments weigh the needs of an expanding digital economy against those of households, farmers, and the environment?

Cape Town shows how quickly that question can become contentious. Proposals for two major Equinix data centres provoked objections over electricity demand, water consumption, and environmental disclosure in a city still marked by the 2017–18 “Day Zero” water crisis. Municipal authorities nevertheless approved the projects in July 2026. Their decision captured a dilemma facing governments that want to attract digital investment without disregarding local resource constraints.

Similar disputes have emerged in the United States, where communities have questioned the effects of rapid data-centre construction on electricity prices, land use, and water supplies. In October 2026, Amazon announced a $1 billion, five-year initiative intended to support U.S. communities hosting its data centres. The announcement reflected a broader reality: building these facilities now requires political negotiation, not merely engineering expertise and financing.

The stakes become higher still when computing capacity is treated as a strategic asset that must be defended. The United Arab Emirates is a case in point. Its pursuit of multi-gigawatt AI infrastructure is more than a bid to attract technology investment. Following regional attacks, the UAE reportedly reconsidered the physical design and geographic distribution of major AI facilities, exploring underground components, dispersed sites, and more resilient infrastructure. Such planning resembles the precautions associated with military installations and critical energy assets. A data centre may house servers rather than weapons, but governments are beginning to think about its vulnerability in similarly strategic terms.

That shift is not difficult to understand. Like ports, telecommunications networks, and energy terminals, large data centres can create dependencies far beyond their physical boundaries. As economies rely more heavily on cloud services, AI systems, and interconnected digital platforms, a disruption can radiate through businesses and public institutions. Protecting computing infrastructure is therefore becoming as much a question of economic resilience as of technological competitiveness.

India illustrates both the opportunities and the vulnerabilities involved. The country generates approximately one-fifth of the world’s data but has less than 6 percent of global data-centre capacity. That imbalance points to significant room for domestic expansion, while also exposing a potential dependence on infrastructure located elsewhere. India has already begun closing part of the gap: its installed data-centre capacity grew from approximately 375 megawatts in 2020 to around 1,500 megawatts by 2025.

Public computing resources are expanding, too. By June 2026, the IndiaAI Mission had facilitated access to more than 45,000 graphics processing units, or GPUs, to make computing capacity more affordable for researchers, start-ups, and other users. The government increasingly describes such investments, alongside domestic foundation models and semiconductor capabilities, as elements of “technological sovereignty.”

The phrase is revealing. Dependence on foreign cloud providers, advanced chips, and computing capacity overseas can leave a country vulnerable in ways familiar from its dependence on imported energy or critical minerals. Control over data alone is not the same as control over the infrastructure needed to process it.

Still, this contest cannot be measured simply by counting server farms. Governments that treat data centres chiefly as real-estate developments risk missing the larger strategic transformation. A competitive AI ecosystem requires advanced semiconductors, affordable compute, dependable electricity, high-quality connectivity, and cooling systems that can operate within environmental limits. Few countries possess all those advantages.

Semiconductor powerhouses can face energy shortages. Energy-rich states may lack sophisticated technology industries. Places with abundant water may struggle to attract capital or establish strong digital connections. Such mismatches will shape investment, alliances, dependencies, and rivalries.

In the twentieth century, the map of strategic power was drawn around oilfields, pipelines, factories, and shipping lanes. Those assets have not lost their importance. But another layer is taking shape: hyperscale data centres linked to electrical grids, fibre-optic networks, and semiconductor supply chains. Their location, resilience, and ownership will influence which countries can develop and deploy the most powerful technologies, and on whose terms.

National power has always rested partly on the ability to turn resources into strategic capabilities. What is changing is the list of resources that count.

In the AI era, computing power is not just a technical input or an economic commodity. It is becoming an instrument of geopolitical power, anchored in the very physical world the digital revolution once seemed to transcend.