Big tech’s energy crunch: Is nuclear energy making a comeback?

In a world increasingly driven by artificial intelligence (AI), cloud computing, and data-hungry applications, the question is no longer whether Big Tech will need more power, but how it plans to get it.  Training large language models like ChatGPT, DALL·E, and other AI assistants requires thousands of GPUs running for weeks or months, consuming huge amounts of electricity, not to mention the constant power needed for everyday use.  Data centres already consume more than 460 terawatt-hours (TWh) a year globally, and the International Energy Agency warns this could more than double by 2026, putting tech giants on par with medium-sized countries in terms of energy use.

Yet many national grids are already stretched thin. Additionally, the clean energy transition is electrifying transport, heating, and other sectors, tightening supply further.  The result is a worsening global power crunch that is forcing tech giants to rethink how and where they source energy.

A recent Economist article reports that Big Tech is deploying a mix of strategies to cope with AI’s exploding energy needs.  This includes shifting data centres to new geographies as prime locations hit capacity, leasing from “neo cloud” providers that repurpose facilities such as old crypto mines, striking long-term deals for renewable and hydro power, and increasingly building on-site generation to reduce grid dependence. Companies are also experimenting with grid-flexibility agreements while using batteries or backup generators to ease peak loads, in exchange for faster grid access, while also looking abroad to energy-rich regions like the Gulf states and Spain.  Alongside these moves, hyperscalers are investing in cutting-edge sources from geothermal to hydrogen fuel cells.  Moreover, companies are considering relocating to cooler or remote areas like the Nordics or even underwater to reduce cooling needs and spread demand, as Microsoft has tested in Project Natick.

Massive renewable investments are also already underway. Microsoft, Google, and Amazon have committed to matching their energy use with clean sources like solar and wind.  Google, for instance, has announced multiple power purchase agreements (PPAs) across North America and Europe to ensure 24/7 carbon-free electricity.  However, despite great green energy efforts, the pace of energy demand from AI and cloud computing is outstripping the rollout of new renewable capacity.

This is where the conversation turns to nuclear energy, which seems to be one of the most closely watched bets of Big Tech.  After decades of controversy, nuclear power is once again being viewed as a serious, scalable option with the potential for a full-blown comeback.  While it has long divided public opinion, plagued by safety concerns, high capital costs, and lengthy construction timelines, especially in regions like Europe, recent technological advances are shifting the narrative. A new wave of Generation IV nuclear reactors, featuring smaller, safer, and more flexible designs, promises lower waste, passive safety systems, and better integration with modern energy grids.  These small modular reactors (SMRs) are now catching the eye of technology giants seeking reliable, low-carbon baseload power to fuel their data-hungry operations.

Among the early movers, Google has struck a deal with Kairos Power, a startup developing SMRs, to supply nuclear power from 2030. Amazon has invested in X-energy, another SMR startup, and previously sought to partner with Talen Energy to develop a nuclear-powered data centre, which regulators, however, ultimately blocked over concerns about local electricity bills.  Microsoft has signed agreements to restart dormant nuclear plants, while Meta has committed to long-term nuclear supply contracts, signalling a broader industry shift toward atomic energy as a reliable, low-carbon power source.

Thus, it also does not come as a surprise that U.S. President Donald Trump’s recently published “AI Action Plan” places nuclear energy squarely at the heart of the country´s strategy to meet the surging electricity demands of AI.  Framing the U.S. power grid as both the lifeblood of the modern economy and a pillar of national security, the plan calls for preventing the premature closure of existing generation facilities and prioritising the rapid interconnection of “reliable, dispatchable power sources”, explicitly naming advanced nuclear fission and even nuclear fusion alongside enhanced geothermal.  This signals a political push not only to preserve legacy nuclear capacity but to accelerate deployment of next-generation designs, aligning with Big Tech’s interest in Generation IV reactors and SMRs as low-carbon, high-output solutions.

However, while Generation IV nuclear is technically promising, it is not yet scalable in the short term as SMRs are unlikely to make a meaningful dent before the mid-2030s at best.  The success of nuclear ventures ultimately depends on whether governments can accelerate demonstration projects, secure fuel supply chains, and streamline regulation without compromising safety.

To conclude, the moves of Big Tech and the U.S. approach reflect a broader recognition that nuclear, which has long been sidelined in energy debates, may be essential to securing the stable, large-scale power supply that the AI era demands.  Time will tell whether technologies and innovations can scale in step with demand, or whether the AI revolution will stall under its own energy weight.

Dr Lina Klesper is an International Legal Assistant at PKF Malta