The head of Amazon Web Services (AWS), Matt Garman, is advocating for the expansion of nuclear energy in the UK to meet the increasing power demands of artificial intelligence (AI) and the infrastructure supporting it. This call to action comes as AWS plans a substantial £8 billion investment in new data centers across the UK over the next four years. This investment is part of a broader strategy that AWS projects will contribute £14 billion (US$18.63bn) to the UK's GDP over the next five years.
Data centers, which are vital for AI, data processing, and streaming services, consume vast amounts of energy, on par with small towns. Garman emphasized that nuclear power is a "great solution" to fulfill these energy needs, describing it as "an excellent source of zero carbon, 24/7 power”. Currently, the UK's 500 data centers account for 2.5% of the nation's electricity consumption, a figure projected to rise to 6% by 2030. To provide context, Ireland's 80 data centers consume 21% of its electricity, a figure expected to increase to 30% within the same timeframe. The UK's National Grid anticipates that by 2050, data centers alone will consume nearly as much energy as all current industrial users combined.
Garman noted that AWS's planning process considers future energy needs extensively, stating, "It's something we plan many years out. We invest ahead. I think the world is going to have to build new technologies. I believe nuclear is a big part of that, particularly as we look 10 years out.”
Several factors contribute to the growing energy demands of AI. AI's acceleration has caused demand for computing power to reach new heights, putting energy supply at the centre of AI infrastructure planning. Data centres, as the physical backbone of cloud computing and AI systems, now consume more than 1% of global electricity, with projections suggesting this figure could reach 3-4% by 2030 as AI deployments accelerate. This surging energy requirement presents a complicated challenge for technology companies balancing ambitious AI development roadmaps with climate commitments.
While renewable sources like wind and solar offer zero-carbon electricity, their intermittent nature complicates reliability for data centres that require constant power. As a result, this tension has renewed interest in nuclear energy as a potential solution for AI's growing appetite for electricity. As the conversation around sustainable energy solutions continues, nuclear power emerges as a viable option, primarily due to its low carbon footprint and ability to supply consistent energy.
The UK, recognizing its lag in AI technology adoption, is advancing with initiatives like AI Growth Zones (AIGZs). These zones aim to attract investment for AI data centers while enhancing access to power and offering planning support, thereby addressing sustainability concerns. The UK government is also streamlining planning processes to expedite the building of more data facilities.
French energy company EDF is constructing a significant new nuclear plant at Hinkley Point in Somerset, with a decision pending on another at Sizewell in Suffolk. EDF's UK Chair, Alex Chisholm, concurred with Garman's perspective, noting that data center providers are turning to nuclear because "They will need a lot of energy, reliably. Replication of Hinkley Point C, alongside the roll out of SMRs, can power Britain's digital economy.”
Small Modular Reactors (SMRs) are also gaining attention. SMRs offer a scaled-down version of traditional nuclear power plants, with the potential for modular construction facilitating quicker deployment. This feature boosts their appeal, especially in data centre sectors keen on integrating sustainable energy sources. Amazon is collaborating with SMR firms and could potentially partner with Rolls Royce, which is developing its own SMR designs in the UK. A spokesperson for the Department of Energy Security and Net Zero stated that modular reactors “will play particularly important roles in growing energy-hungry sectors like AI and we're shaking up the planning rules to make it easier to build nuclear power stations across the country.”
However, investing heavily in nuclear energy doesn't address the core issue, as nuclear projects have long lead times while AI's energy demands are coming now. While some hyperscale tech firms are investing in technology such as small modular reactors (SMRs) to either feed the grid or power their datacenters directly, this is immature technology that isn't expected to be ready before 2030.
Despite concerns about radioactive waste, the minimal carbon emissions from nuclear power make it a strong candidate for reducing the environmental impact of data centres. As SMR technology becomes more widespread, the feasibility of nuclear-powered data centres is expected to increase significantly over the next decade, supported by favourable policies and investments in the nuclear sector.
