Artificial intelligence is advancing at extraordinary speed. But while much of the focus remains on algorithms, chips and computing power, a more fundamental constraint is coming into view: electricity.
Across the US and Europe, the rapid expansion of data centres is placing growing strain on energy systems. According to the International Energy Agency (IEA), data centres already account for around 1.5% of global electricity consumption, a figure projected to more than double by 2030 as AI workloads scale. In the United States alone, data centres are expected to drive nearly half of all electricity demand growth this decade.
In some US states, households have already seen sharp increases in electricity bills, as utilities pass on the cost of new generation and grid upgrades driven largely by demand from large industrial users, including AI operators.
This is a small but telling signal of a much larger shift underway. AI is no longer purely constrained by digital innovation. It is constrained by power. And that reality is forcing a reckoning across global energy systems.
AI is pushing the energy system past its limits
The scale and pacing of AI workloads are testing conventional energy planning. Unlike traditional IT infrastructure, AI systems require continuous, power-dense electricity. Pauses for overnight cooling or low-use periods are rare. These usage profiles resemble industrial load far more than conventional data centres of the past.
The scale is material. The IEA estimates global data-centre electricity consumption at over 415 terawatt-hours (TWh) per year today, with projections of approaching 1,000 TWh by 2030 – roughly equivalent to the annual electricity consumption of a major industrialised economy.
Meeting that growth will require enormous additions of generation capacity. Industry estimates suggest that in the US alone, AI-driven data-centre expansion could drive demand for 100-300 GW of new solar capacity by 2030.
Yet energy systems have not evolved on the same timeline. Whilst data centres can be planned and brought online in months, new power plants, transmission upgrades, and permitting processes typically unfold over years. This mismatch is increasingly becoming visible, putting the grid system under strain with rising costs, congestion in key regions, and growing reliance on short-term solutions.
Energy needs to become a competitive advantage
As AI technologies mature, access to energy is emerging as a core determinant of competitive advantage.
What differentiates regions today is not only intellectual capital – which has become more globally distributed – and presence of high performance computing facilities (GPUs/TPUs), but also the availability of reliable, scalable power. Regions with abundant, scalable power can develop and deploy AI faster, cheaper, and at larger scale. Those without it face bottlenecks.
China’s approach illustrates this clearly – in 2024 alone, China accounted for over 60% of global renewable capacity additions, driven primarily by solar deployment. Under its ‘East Data, West Computing’ initiative, newly built National Hub data centres are required to source at least 80% of their electricity from renewables by 2030. This large-scale investment into renewables has expanded electricity capacity at a scale that supports data centre growth, while limiting long-term dependence on coal – strengthening both energy security alongside industrial competitiveness.
India is pursuing a similar strategic alignment. The Adani Group has announced plans to invest $100 billion in renewable-powered hyperscale AI data centres, expanding capacity to 5GW by 2035. This could catalyse a further $150 billion in related industrial development.
In contrast, many advanced economies in the West are limited by legacy grids and slow clean-energy deployment – despite strengths in technology and research. Without sufficient investment in new capacity, energy systems will remain exposed.
Fossil fuels are the wrong response
Faced with this rising demand, some regions have turned back to fossil fuels. In the US, recent preliminary emissions data from the IEA shows an increase in carbon output in 2025, reversing the previous decline. This is linked in part to higher coal and gas generation as electricity demand has grown.
It’s an understandable response, but it’s flawed. Coal and gas are used because they exist, not because they make sense – coal remains the most carbon-intensive source of electricity, responsible for over 40% of global power-sector CO₂ emissions, despite generating a smaller share of total electricity. Gas, while less carbon-intensive than coal, is no longer cheap in many markets and still locks in emissions for decades. Continuing to use fossil fuels undermines climate commitments, locking in long-term emissions without the power payoff required.
Crucially, neither can scale fast or cleanly enough to meet demand. New fossil fuel generation can take 10-15 years to deploy – far slower than the pace at which AI infrastructure is being deployed.
AI is already here. But we’re trying to power 21st-century technology with a 20th-century energy playbook.
Why solar fits
Solar power offers a more credible response, aligning with AI infrastructure in ways that fossil fuels simply cannot. Solar is already one of the lowest-cost sources of new electricity in most major markets, according to the IEA. It can be deployed quickly, modularly, and scaled incrementally alongside demand – a critical advantage when electricity needs are rising fast and unpredictably.
Recent forecasts from the US Energy Information Administration suggest solar capacity additions will accelerate sharply, with around 70 GW of new solar capacity expected by 2027, representing close to a 50% increase in installed capacity in just a few years. Globally, solar now accounts for around three-quarters of all new renewable capacity additions, making it the fastest-growing source of electricity worldwide.
Importantly, solar can be co-located with data centres clusters, reducing grid strain and price volatility. Combined with storage it offers a pathway to firm, reliable capacity without the emissions of fossil fuels.
So, solar isn’t an idealistic solution – it’s a practical one. This perspective is reflected in industry sentiment: according to a recent PV Magazine USA survey of US data centre developers and energy planners, 55% of respondents identified a largely renewable energy mix – predominantly solar paired with battery energy storage – as the ‘ideal’ way to meet growing power demand, ahead of natural gas, nuclear, and other low-carbon sources. More than half (54%) cited ‘energy availability and redundancy’ as the greatest obstacle to data centre deployment through 2030, and nearly three quarters (74%) saw advanced battery energy storage systems (BESS) as key to ensuring resilience.
This alignment between strategic planning and technology trends proves that renewables paired with storage are now viewed as the most credible long-term foundation for powering the AI era.
The efficiency multiplier effect
Not all solar is equal, and as AI data centres push electricity demand to new heights. The efficiency of individual panels is becoming as strategically important as the scale of deployment. CPT’s Photon Multiplier (PM) technology addresses this directly, enabling solar panels to convert a greater proportion of the solar spectrum into usable electricity – including wavelengths that conventional panels waste as heat. The result is meaningfully higher energy output from the same physical footprint; a critical advantage when land, grid connection points, and capital are all constrained.
This effect is most powerful when solar is paired with battery storage – the configuration that the majority of data centre developers now view as the long-term solution. In PV+BESS configurations, CPT’s high-efficiency panels deliver compounding benefits across the whole system: reduced Balance-of-System costs, higher generation per area with extended daily charging window and less seasonal and daily variation, faster battery charging, better battery utilisation and more energy available for time-shifting to cover night-time or peak demand periods. Critically, greater panel efficiency also reduces the BESS capacity needed to guarantee reliable, round-the-clock power – lowering capital costs and reducing residual dependence of fossil fuel backup.
In an era where the race to build AI infrastructure is also a race to secure clean energy, efficiency gains of this kind are not marginal improvements – they’re multipliers on the entire system.
AI, growth, and climate don’t have to clash
The AI boom forces a choice – but not between economic growth and climate action.
AI can either drive renewed fossil-fuel dependence, pushing emissions higher, or it can accelerate the deployment of clean energy at scale. Clean power abundance enables both competitiveness and decarbonisation. Energy scarcity, by contrast, will eventually slow AI itself.
What AI is revealing, more quickly than any previous technology, are the limits of our energy systems. It is stress-testing planning assumptions, grid resilience and policy frameworks in real time.
Solar is no longer an alternative. It is becoming foundational to the digital economy.
The story of this decade will not be defined by whether AI transforms society – that is already happening. It will be defined by whether our energy systems evolve quickly enough to support that transformation responsibly. Cambridge Photon Technology’s Photon Multiplier technology enables solar energy to play an even more pivotal role in powering the AI boom.
