A moment ago, I e-mailed a weighty report to a colleague. I also used an AI tool to tidy up the note in the e-mail and then used Gemini to answer a question for a story. In between, I looked at my iPhone to send a photo to a friend. These simple daily tasks have data centres – and increasingly AI – in common. And energy. That’s a problem. AI and data centres might be indispensable, but their vast and growing appetite for energy has
reached a crunch point. This threatens not only the stratospheric growth projections of AI and data centre expansion but also Big Tech’s reputation and social licence to operate. There simply are not enough electricity or grid connections right now to meet surging power demands. Or enough water in some cases. There’s an urgent need for more renewable energy, battery storage, nuclear power (and eventually fusion energy) but AI demand is growing faster than energy sources can keep up with. Data centres can be constructed in
two to three years – faster in many cases than utilities can build the grid connections. That is forcing tech giants such as Alphabet, Amazon, Meta and Microsoft to invest in renewables, on-site power generation, as well as nuclear power and next-gen energy technology such as small modular reactors and fusion energy. There’s a lot at stake as Big Tech firms race for AI and data centre dominance. Globally, major tech firms are expected to spend more than US$650 billion (S$840 billion) in 2026 to
expand AI capacity. Yet, nearly half of the planned US data centres in 2026 are projected to be delayed or cancelled, according to Bloomberg. Stressed grids There simply aren’t enough key electrical components – such as batteries, switchgear, transformers – or grid connections to new or planned mega data centres that in some cases are hundreds of kilometres away from large power generators. Grids are already stressed by the growing numbers of electric vehicles and home heating, such as heat pumps. This has slowed, but
not stopped, expansion. US data centre power demand is still forecast to more than double to 66GW in 2027 from 31GW in 2025, according to Goldman Sachs Commodities Research. The power crunch is being felt in China, too, even as it builds record amounts of solar, wind and battery storage. The nation is on course to nearly double its data centre capacity within five years to 60GW, according to new analysis from Rystad Energy. In South-east Asia, underfunded power grids are failing to keep pace
with surging energy demand, a report by Bain & Company and Standard Chartered bank said on May 18. Investors are lining up to build data centres, renewable energy and other green projects. But most grids in the region are too slow to adapt to the rapid electrification of economies. So, Big Tech is increasingly in the energy business, building power plants, funding grid upgrades and connections. That is making them a change maker. The International Energy Agency (IEA) said the tech sector remains a major
driver of renewables procurement, accounting for around 40 per cent of all corporate renewables power purchase agreements signed globally in 2025, according to the IEA. These agreements underpin investments in renewables by locking in long-term revenues. Nuclear energy has got a boost, too. In January 2026, Meta announced it struck deals with Oklo, Vistra and TerraPower in the US to supply up to 6.6GW of nuclear power by 2035. In June 2025, power firm Talen Energy Corporation said it would provide Amazon nearly 2GW of
power annually until 2042 from Talen’s Susquehanna nuclear plant in Pennsylvania. Long-term battery storage is also a growing solution for hyperscalers fretting over long grid connection delays. In February 2026, Google said it would pay about US$1 billion for a 100-hour iron-air battery from startup Form Energy for a new Minnesota data centre powered by renewables. In 2024, Microsoft signed a deal with Brookfield Asset Management to invest more than US$10 billion to develop renewable energy capacity to help power data centres. Amazon said it
has committed to more than 700 green energy projects globally to power cloud and AI data centre growth. They include investments in more than two dozen nations in large-scale and on-site solar, offshore wind, battery storage and nuclear. But more recently, the rush to build more data centres and secure more power has put Big Tech’s sustainability goals under pressure. In May 2026, Bloomberg reported that Microsoft is weighing whether to delay – or even abandon altogether – its 2030 target of matching 100 per
cent of its hourly electricity use with renewable energy purchases. Data centre owners are also investing heavily in gas-fired generation, off-site and on-site, leading to criticism for their reliance on fossil fuels. This underscores that while renewables are still preferable, the need now is for power. The United States has the most gas-fired power capacity in development globally, over a third of which is slated for data centres, Global Energy Monitor, which tracks fossil fuel projects, said in a January 2026 report. In Texas, about
40GW of new gas power is planned for data centres. The giant (and still growing) Stargate data centre in Texas by OpenAI, Softbank, Oracle and others uses on-site gas-fired turbines for power. In Louisiana, Meta is developing its Richland Parish Data Centre, among the largest in the country. In a deal with the local utility to speed things along, Meta will pay for the new energy infrastructure, including 10 gas-fired power plants (up from three originally), plus high-voltage power lines and battery storage. It also
said it is committed to supporting new renewable energy capacity. Power challenge The IEA spells out the sector’s power challenge and why Big Tech is upping its power investments. By 2030, global electricity consumption from data centres will roughly double from 2025 levels, accounting for around 3 per cent of global electricity demand by then. Electricity consumption from AI-focused data centres will triple over this five-year period, the IEA said in a recent report. In the US, data centres already consume about 6 per cent
of electricity, according to the International Data Centre Authority. Estimates vary, but this could rise to 9 to 12 per cent by 2030. In Singapore, data centres consume about 7 per cent of electricity, rising to 12 per cent by 2030, according to the Infocomm Media Development Authority. Data centres are set to be the fastest-growing source of power demand in China, according to Rystad’s research. Data centres are warehouse-sized buildings containing rows of stacked servers. Specialist AI data centres are more like giant supercomputers
with racks of powerful processors, vast amounts of memory storage and liquid cooling systems direct to the computer chip or even full immersion of each server. Air-conditioning cannot handle all the heat from AI server racks. AI’s immense and growing energy demands are also colliding with worries about energy security and competing demands for electricity for other uses. Plus, there are increasing concerns about environmental and social issues, such as ensuring equitable access to electricity and water for local communities. A UN report published on
June 3 starkly lays out the risks from unbridled AI and data centre growth. If data centres were a country, their 2025 electricity consumption would rank them 11th globally, roughly on a par with France, said the report by the UN University Institute for Water, Environment and Health. AI-related workloads accounted for roughly 20 per cent of total data centre electricity use in 2025. That share could double to 40 per cent by 2030, generating more heat and water use as more data centres are
built, the report said. There are already more than 4,000 data centres in the US and in some areas, local power bills have shot up. It notes that Google processes an estimated five trillion searches annually and a conventional search uses about 0.3 watt-hour (Wh) of electricity. An AI-enhanced generative search uses up to 3Wh, a 10-fold increase. By 2030, data centres could consume 9.32 trillion litres of water – enough to meet the annual basic water needs of the entire population of sub-Saharan Africa,
the report said. And by 2030, AI infrastructure could generate e-waste each year that is roughly equivalent to discarding 250 Eiffel Towers. What’s needed is greater transparency on AI’s carbon, water and land footprints to improve land and resource use planning and avoiding shifting environmental costs onto vulnerable communities, the report’s authors said. Ultimately, Big Tech is all about innovation. To harness AI and data centre riches, tech firms need to focus on solutions, from more energy- and water-efficient data centre design, more efficient AI
models and apps to building a better rapport with communities. Data centres can drive the vital grid upgrades, and their investments can, if targeted wisely, enhance energy security by ramping up green energy investment and the electrification of everything. And doing so will also boost public trust, a force more powerful than electricity itself.
AI energy crunch, data centre expansion, grid constraints, renewables, nuclear power, gas-fired generation, Microsoft, Meta, Amazon, Alphabet, water use, e-waste, IEA