AI’s rapid growth is fuelling a race for energy. Dominic Howell speaks to Michael Rae, manager of the M&G Global Energy Opportunities Strategy, about the power demands of the AI revolution and the investment opportunities it could create along the energy value chain.
The number of technological changes in modern history that have left a visible mark on global energy demand can be counted on one hand. The steam engine and railways helped propel the rise of coal in the 19th century; the automobile transformed oil into the dominant fuel of the transport age; and mass electrification brought an expanding universe of appliances, motors and machines into homes and factories in the second half of the 20th century.
Artificial intelligence (AI) appears to be the latest technology which will drive meaningful changes in energy flows. The International Energy Agency (IEA) expects global data centre electricity consumption to more than double from around 415 terawatt-hours (TWh) in 2024 to approximately 945 TWh by 20301 − an addition roughly equivalent to the annual power demand of Germany.2
Reconfiguring the energy system to meet this anticipated increase will inevitably require major investment. It could also unlock diverse investment opportunities as the required capital works its way through the physical economy: into power generation, electricity grids, cooling systems and the equipment needed to deliver usable power to AI chips.
The most powerful driver of the changes in the energy sector has always been persistent growth in global energy demand. As Michael Rae, manager of the M&G Global Energy Opportunities Strategy, say, “rising wealth, population growth and human progress through innovation have all historically been associated with increased energy use.”
In the past decade, these forces have been masked by economic volatility and the COVID pandemic, but they have not gone away. The IEA forecasts global electricity consumption will increase by 3.6% a year between 2026-2030, fuelled by demand from industry, the take-up of electric vehicles (EVs), increased use of air-conditioning and the adoption of AI. This growth is 50% higher on average compared with the rate across the previous decade.3
As AI data centres become more pervasive, the focus of the debate is increasingly turning to power demand with the US set to add a record 86 gigawatts (GW) of new power capacity in 2026 to meet that demand.4 Rae notes that “the power needs of AI have been talked about for some time, but 2026 is the first year in which the phenomenon has been visible in system-level demand trends and, to an extent, power prices”.
The sheer scale of the power demand for AI projects is coinciding with robust economic growth in emerging markets, especially China, and a generational upgrade of the grid in many countries. Combined, these factors could mean we are at a defining moment for the global energy sector.
With the world also on a path to cleaner energy sources, it is plausible that the energy landscape is set to be transformed by massive capital investment in the coming years. The question for investors is where the opportunities will lie.
Since the launch of ChatGPT in November 2022, AI has dominated headlines and market narratives. However, the so-called AI trade has mostly been focused on technology and computing equipment, rather than the wider framework that makes it possible.
This is evident from the phenomenal increase in the market capitalisation of the global semiconductor and semiconductor equipment sectors in the past couple of years.
As the proverbial ‘picks and shovels’ companies that supply the chips used in AI models and data centres, firms like Nvidia and TSMC are expected to play an important role. However, as data centre construction activity accelerates in the physical economy, the focus is increasingly turning to power.
AI data centres require energy for computing and cooling. According to the IEA, they already accounted for around 1.5% of the world’s electricity consumption in 2024.5 Unsurprisingly, given US companies are at the forefront of AI, data centres tend to be concentrated in the US, but many are located in China, Japan and Europe, with notable clustering in certain areas. In fact, in Ireland, they consumed over 20% of the country’s electricity in 2025.6
With many more data centres being planned, the energy intensity of AI means that energy infrastructure is critical to its success. However, the market caps of the energy, utilities and materials sectors suggest that this is not being fully recognised by investors.
Artificial intelligence may appear weightless, but every calculation depends on an immense physical network. We follow the journey of an electron from a photovoltaic solar panel, through the electricity grid and into a semiconductor inside an AI data centre.
A modern utility-scale solar panel converts roughly 20% of incoming sunlight into electricity. A large solar farm can generate hundreds of megawatts (MW), enough to power hundreds of thousands of homes under favourable conditions, or a single hyperscale AI campus.
Solar panels generate direct current, but grids operate on alternating current. Inverters synchronise every solar installation with grid frequency and voltage. Modern power electronics have become one of the most critical components in the global energy transition.
Voltage is increased before electricity enters the transmission network. Higher voltage allows power to travel efficiently over long distances with lower resistive losses.
Electricity enters a vast interconnected system in which generation and demand must remain continuously balanced.
High-voltage transmission networks move electricity across regions and countries. In many AI projects, transmission availability has become a larger constraint than generation availability.
Electricity must be stepped down as it approaches end users. For new data centres, access to local substations is often the binding constraint. Grid connection delays can exceed the construction time of the data centre itself.
Electricity is distributed to homes, businesses, industry and critical infrastructure. Data centres require high-capacity, reliable connections and can place significant new demands on local networks. Hyperscale campuses increasingly seek 500 MW to 1 GW (gigawatt) connections, equivalent to a large industrial complex or a small city.
Data centres consume vast amounts of electricity, not only for computation, but also for cooling, networking, data storage, lighting and operational resilience.
Before electricity reaches a processor, it passes through multiple conversion and regulation stages, with energy lost at each step. At hyperscale, efficiency is valuable.
At server level, electricity is converted again and delivered at the exact voltages required by processors, memory and networking equipment.
Inside the semiconductor, billions of transistors manipulate electrical signals to perform the calculations behind AI models.
The importance of sufficient electricity supply for AI’s future needs has not escaped the CEOs of large technology firms, including Elon Musk. Although there has been much discussion about a shortage of memory chips, the real AI bottleneck is power, with demand racing ahead of capacity deployment.
The scale of the power constraint can be seen from the current grid connection queue. This represents all the power projects that are being planned to supply data centres and need to be connected to the electricity grid. Currently, the amount of proposed capacity is massive and very unlikely to be fully realised. In Texas, for example, where many US data centres are being built, the requests add up to 474 GW, which is more than five times Texas’s record peak demand.7
Public pushback and concerns around grid capacity are forcing large technology firms, or hyperscalers, to adopt ‘behind the meter power’, which entails building power generation, typically gas-fired, specifically for their data centres. This means that they can proceed without a grid connection and also avoid some of the concerns about their data centres putting strain on the grid and driving up energy prices.
Firms that are able to resolve this huge energy supply-demand mismatch could be in a promising position. For example, NextEra Energy, the largest utility firm in the US, is one of the companies that has been identified as enabling the hyperscalers to secure reliable, long-term power supply.
Beyond the utility companies, another industry that could benefit from the power crunch is gas turbine manufacturing. Leading firms in this market such as Siemens Energy and GE Vernova are seeing demand increase dramatically, as more gas-fired power plants are being constructed to supply data centres.
With an imbalance between orders and global manufacturing capacity, turbine prices have also rocketed. Wood Mackenzie, an energy industry research firm, estimates that by 2027 the cost of turbines will have risen by 195% since 2019.8
Although data centre projects are currently the main driver of demand for turbines, the reliability of gas-fired power stations means that they are also being introduced to complement intermittent renewable power sources, such as wind and solar.
As a result, the gas turbine sector is booming, with companies reporting multi-year order backlogs. While the sector’s integral role in the AI buildout and the energy transition makes it a potentially attractive investment proposition, it remains a cyclical industry and it is important for investors to think about the durability of demand, as well as any potential risks with contracts.
However, in the near term, as the race for power continues and bottlenecks in the energy supply chain are increasing, gas turbines appear to be attracting wider attention as an essential part of solving the current electricity capacity constraint.
Besides the bottleneck in power generation equipment, another challenge in meeting AI’s rising electricity demand is the state of the grid. In many developed markets, grids have not been upgraded for several decades.
To prepare for what the IEA has called the ‘Age of Electricity’, grid infrastructure around the world will have to be expanded and modernised to meet increased demand and integrate renewable energy sources.
China has been preparing for this new era by investing heavily in renewable power generation in recent years. It is planning to spend a further ¥5 trillion (US$722 billion) on its power grid over 2026-2030, up from ¥2.85 trillion during 2021-2025.9
“As is often the case, the scale of China’s ambition is in a category of its own – it will add the equivalent of the entire US power base in five years”, Rae observes. If the AI race eventually becomes more of an energy cost battle than a compute battle, China could find itself in a very strong position. It will also help with the country’s path to decarbonisation.
The European Union has also recognised the importance of a digitalised and flexible power network and is planning to spend €584 billion on its grids by 2030.10 Meanwhile, the UK is replacing old components and connecting renewables as part of its £30 billion ‘Great Grid Upgrade’ by 2029.11 These long-term and major capital investment projects may provide potential opportunities for businesses that can provide the cables, interconnectors and transformers required for this huge modernisation effort. For example, cable suppliers like Prysmian and Nexans are currently involved in many projects around the world, including sub-sea and offshore solutions, providing power connections for the infrastructure of the future. Building tomorrow’s world is not simply dependent on silicon and AI models. The energy sector has an essential role to play and the potential investment opportunities exist along the entire energy value chain.
Energy has been described as ‘the only universal currency’,12 given it is fundamental to the functioning of our world. This concept is apt when considering how AI is expected to transform the economy and society. The vast amount of capital expenditure involved in deploying AI has been widely discussed. However, this typically focuses on the cost of chips rather than the amount of money needed for energy infrastructure to power them. At some point, the true value of AI’s power demand will have to be recognised. AI is not the only factor expected to reshape the global energy sector in the next few years. Rising wealth in emerging markets may also fuel increased power demand, and the pending transformation of the entire energy system from generation to transmission to adapt to the Age of Electricity is seismic in nature. Rae believes that “the question for investors is: Are these forces adequately reflected in current equity market valuation?”. He points to the fact that the combined weighting in the MSCI World Index of the energy and utilities sectors is only 6.5%, approximately half its level of a decade ago.13
Given the increasing reliance of innovative growth sectors such as AI, robotics and EVs on energy, it is questionable how long this apparent market ambivalence to the currency that makes it all possible can last.
1 IEA, Energy and AI, (www.iea.org) April 2025.
2 According to the Bundesnetzagentur, Germany’s generation in 2025. amounted to 437.6 terawatt hours. (Bundesnetzagentur, January 2026).
3 IEA, ‘Electricity 2026’, (iea.org), February 2026.
4 US Energy Information Agency, ‘New U.S. electric generating capacity expected to reach a record high in 2026’, (eia.org), February 2026.
5 IEA, Energy and AI, (www.iea.org) April 2025.
6 RTE, ‘Data centres accounted for 23% of electricity usage last year’, (rte.ie), July 2026.
7 Office of the Texas Governor, ‘Governor Abbott Directs Comprehensive Data Center Audit’, (gov.texas.gov), August 2026.
8 Wood Mackenzie, ‘Gas turbine prices soar 195% as market faces supply-demand crisis’, (woodmac.com), April 2026.
9 China Daily, ‘China to invest 5 trillion yuan in power grid over next 5 years’, (global.chinadaily.com.cn), February 2026.
10 European Commission, ‘European Grids’, (energy.ec.europa.eu), accessed September 2026.
11 National Grid, ‘The Great Grid Upgrade: what's happening and why?’, (nationalgrid.com), accessed September 2026.
12 Vaclav Smil, ‘Energies: An Illustrated Guide to the Biosphere and Civilization’, 1998.
13 Source: MSCI.com and M&G Investments.
The views expressed on this webpage should not be taken as a recommendation, advice or forecast, nor a recommendation to purchase or sell any specific security.
The value of investments will fluctuate, which will cause prices to fall as well as rise and you may not get back the original amount you invested. Past performance is not a guide to future performance.
