AI data centres spark trillion-dollar race to power the AI era

Global spending on data centres that support AI is projected to reach about $3 trillion worldwide from now through 2029, according to Morgan Stanley. Roughly half of that sum will go toward construction, and the other half on the specialized hardware powering AI systems. The forecast underscores the scale of the AI infrastructure push and compares the potential economic footprint to major national economies, with the French economy cited as a rough benchmark for 2024. In the United Kingdom, analysts anticipate around 100 additional data centres to meet surging AI processing demand, and Microsoft announced a $30 billion investment in the UK's AI sector.

AI data centres differ from traditional server farms because they are built for density and ultra-low latency. Most AI models rely on Nvidia GPUs, which are deployed in large cabinets that can cost around $4 million each. To train large language models, thousands of machines must operate in close proximity, and every metre of distance between chips adds nanoseconds of delay. In AI workloads those tiny delays accumulate, so operators pack cabinets tightly to enable parallel processing and treat the system as a single, massive computer.

![Dense AI data centre cabinets]https://ichef.bbci.co.uk/news/480/cpsprodpb/3527/live/76c28bf0-930b-11f0-a647-dff301f4a439.jpg.webp

That density comes with energy costs. The data-centre footprint can draw gigawatts of power, with training workloads triggering spikes in electricity demand comparable to thousands of kettles being boiled at once. Daniel Bizo of The Uptime Institute says the scale of AI workloads creates an electricity challenge unlike traditional data centres, describing the load as a singular demand on the grid. He notes that normal data centres produce a steady hum, while AI workloads create sharper, harder-to-manage fluctuations.

Operators are pursuing multiple energy strategies to manage the load. Nvidia chief executive Jensen Huang has suggested that in the United Kingdom, short-term measures could include off-grid gas turbines to avoid burdening the public grid, while AI itself could drive advances in energy technology such as more efficient turbines, solar, wind and even fusion energy. Microsoft, which has invested heavily in energy projects, has a deal with Constellation Energy aimed at bringing nuclear power back to Three Mile Island. Google has outlined plans to pursue carbon-free energy by 2030, and AWS has described itself as the largest corporate buyer of renewable energy.

![Cooling and energy infrastructure at AI data centre]https://ichef.bbci.co.uk/news/480/cpsprodpb/711b/live/0e174f10-930a-11f0-a647-dff301f4a439.jpg.webp

Water use for cooling is another critical issue as data centres expand. In Virginia, a state seeing rapid data-centre growth, lawmakers are weighing a bill that would tie new site approvals to water-use figures. In the United Kingdom, a proposed AI factory in northern Lincolnshire has faced objections from Anglian Water, which has argued that it is not obliged to supply water for non-domestic use and has proposed using recycled water from the final stage of effluent treatment as a coolant instead of drinking water.

Industry observers acknowledge the debate around the scale of AI data-centre development. A data-centre specialist described the industry as talking up capacity with a term that has been described as bragging about watts. Nonetheless, some analysts argue that AI will have a larger impact than the internet and that the real estate value of AI data centres makes the sector a durable bet. One adviser says the current path may not be sustainable forever, but AI’s potential justifies continued investment in density and reliability.

![Industry analysis on AI data centre economics]https://ichef.bbci.co.uk/news/480/cpsprodpb/45ea/live/1ad571b0-9309-11f0-a647-dff301f4a439.jpg.webp

Even with caution about the pace of spending, proponents say AI data centres will be required for the foreseeable future because end-user demand for AI services continues to grow, and manufacturers and cloud providers will need dense, high-performance computing capacities to support the AI era. In short, the data-centre sheen around AI is matched by real engineering, energy, and environmental challenges that will shape how quickly, and where, the next wave of infrastructure unfolds.