As the AI race drives up energy demand, US policymakers and Big Tech companies need to balance climate targets, tensions with China, and rising costs. On his second day in office, President Donald Trump announced a massive $500 billion in funding for “Project Stargate.” Should this AI boom be fueled by fossil fuels, nuclear, or renewables?
The US should focus on renewables and innovative low-power alternatives to reduce AI energy demands. Compared to nuclear power, solar and wind offer cheaper and quicker sources of energy — but they must be accompanied by investment in energy storage and decoupled from Chinese supply chains.
Chinese startup DeepSeek’s launch of a low-cost, energy-efficient model temporarily threw some doubt into predictions of massive AI energy consumption and caused US energy stocks to fall.
But these doubts look premature. Over the past week, leading US companies have repudiated the idea that increasing computing power no longer produces better models. They are also spending more than ever to fuel AI. Alphabet CEO Sundar Pichai just announced capital expenditure plans of $75 billion for 2025. All told, Alphabet, Amazon, Microsoft, and Meta forecast $320 billion of spending in 2025, up from $246 billion in 2024 and $151 billion in 2023.
Data centers already gobble up 4% of the total US energy supply – a number set to triple by 2030. This has made certain AI companies look to fossil fuels, sending gas-fired energy generation soaring. In the tradeoff between climate pledges and powering the AI boom, many US actors would choose the latter. But if we accept that we cannot burn more fossil fuels and accelerate climate change to power AI, two choices emerge: nuclear power or renewables like wind and solar.
Leading tech companies currently have high hopes for nuclear energy. Microsoft has entered into a power-purchase agreement that would reopen the Three Mile Island nuclear plant in Pennsylvania. Much hope is placed on Small Modular nuclear Reactors (SMRs). These mini nuclear plants can be built on a wider range of sites than a conventional reactor, allowing them to be co-located with a datacenter. Google and Amazon have signed deals with providers of small reactors.
But nuclear – small or big – is not a panacea. Talk of small reactors has been around for many years – but not a single one is in commercial operation. The things that have always held back the nuclear industry apply: public concerns, complexity, the necessity of underwriting by governments, nuclear waste management, long build-out timescales, and above all: high costs. The price of electricity from new nuclear plants will be $131–204 per megawatt-hour (MWh). Newly constructed utility-scale solar and wind plants produce electricity at somewhere between $26–50 MWh.
Renewables such as wind and solar offer a better alternative. There are some promising signs: for instance, Amazon has entered into a deal that will provide wind and solar energy to power Amazon’s European cloud operations.
But an AI boom powered by solar and wind must not rely on China. The country dominates renewable supply chains. It is unlikely that the current President will mandate Chinese solar panels and Chinese lithium batteries surrounding US data centers.
In addition – as the sun does not always shine and the wind does not always blow – it is crucial to boost domestic energy storage. The US should invest in low-cost energy storage technologies such as pumped storage hydropower (PSH) and sodium batteries – to decouple from the Chinese stranglehold on lithium batteries.
Another important step is reducing AI’s energy demand. As history has shown, new low-power alternatives facilitate technological adoption. It was an Arm low-power breakthrough that made smartphones viable. Now, the energy efficiency of data centers can be improved by investing in next-generation chips. New AI chips offer huge power reductions over current Nvidia chips. Many of these innovations are being developed outside the US – for example by Canada’s Blumind and the UK’s VyperCore.
At the same time, it is important to improve AI’s algorithmic efficiency. DeepSeek demonstrated how it is possible to produce a leading AI model with much less computing power than previously assumed. Yet “even with more energy-efficient models, power demand could surge everywhere,” says Ed Hirs, an energy economist at the University of Houston.
The AI revolution is inevitable, and so is the increased demand for energy to power it. The US should invest in renewables and energy storage technologies and embrace innovations – often produced by its allies – to reduce energy demand. DeepSeek has signposted the way. Now, the US must get there before China.
Christopher Cytera CEng MIET is a non-resident senior fellow with the Tech Policy Program at the Center for European Policy Analysis and a technology business executive with over 30 years’ experience in semiconductors, electronics, communications, video, and imaging.
Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.
Tech 2030
A Roadmap for Europe-US Tech Cooperation
