Further, Artificial Intelligence (AI) services need huge computing power, fueling the blistering growth of data centres around the world. One, they’re going up faster than any sort of collective rational thought. Their huge electricity use is in many ways triggering long overdue conversations around sustainability, infrastructure and equity. Big tech companies like Google are competing to develop the infrastructure under which those AI technologies can thrive. Yet, this pursuit comes with huge environmental impacts and logistical realities.
It’s been a successful formula for the data centre business, which has been rapidly expanding to meet exploding demand for compute. With the extremely rapid growth and demands of AI workloads, the energy usage of these facilities has increased exponentially. Training each subsequent large language model (LLM) has resulted in an exponential growth of electricity use. It’s the equivalent of thousands of homes switching their kettle on and off every few seconds! This new reality presents significant challenges not only exacerbating the burden on current electrical grids but in the future management of energy resources.
As stated by experts in the space, the scale of demand created by AI workloads is unlike anything we have seen before. Daniel Bizo of The Uptime Institute stressed how markedly different the data centres running AI are, compared to run-of-the-mill data centres. He stated, “Normal data centres are a steady hum in the background compared to the demand an AI workload makes on the grid.” This surging demand for energy is prompting executives from all sectors—from manufacturing to healthcare—to reconsider how they purchase and use energy.
The financial implications are enormous as well. It is estimated that around $3 trillion (£2.2 trillion) will be spent on data centres that support AI technologies between now and 2029. For context, this combined sum is greater than the total value of the French economy in 2024. Projections suggest that in the UK alone, the next few years will bring the opening of an additional 100 data centres. This expansion is intended to address the increasing need for high-performance AI processing.
The race for AI infrastructure poses equally daunting challenges when it comes to safe, reliable water supply. Data centres, after all, need huge quantities of water in order to keep the chips that run their operations cool. Companies such as Anglian Water stress the point that they are under no obligation to supply water for non-domestic use. In addition, this position creates an additional layer of complication. Recycled water, produced from the final stages of effluent treatment, is now a practical substitute for cooling. This change lessens the burden of portable water sources.
Legislation, too, is catching up to these challenges. That’s exactly what lawmakers in Virginia have done in their lead bill. It would require new data center sites to be approved based on the level of water consumption they expect. This recent shift is a testament to the growing awareness among legislators on the climate consequences of allowing data center sprawl.
The hunger for AI is apparently insatiable. For firms like Google, saving energy, resources, and emissions in data centers will be key to meeting escalating pressure from stakeholders to increase sustainability. Jensen Huang, a prominent figure in the industry, remarked on the need for innovative solutions: “off the grid so we don’t burden people on the grid.” Now it is data centres that need to operate off-grid from fossil fuel energy sources. They must all do so while reducing harm to local communities.
Experts like Zahl Limbuwala acknowledge the challenges ahead: “The current trajectory is very difficult to believe. There has certainly been a lot of bragging going on. Investment has to deliver a return or the market will correct itself.” These companies are the frontrunners, as tech companies around the world are fiercely competing for dominance in AI technology. With great power comes great responsibility and sustainability.
Our experts are engineering solutions for the extreme engineering challenges from AI workloads. They say they are drawing parallels with equally extraordinary accomplishments in engineering history. Bizo noted, “The singular workload at this scale is unheard of. It’s such an extreme engineering challenge, it’s like the Apollo programme.” As with many such comparisons, it emphasizes the great innovation that will be needed and the potential pitfalls of rapidly scaling up such complex, unproven infrastructure.