AI is reshaping the power investment landscape

The rapid development of new AI-focused data centres is turbocharging investment in new power generation, energy storage and grid projects. Yet the scale of the opportunities on offer is matched by growing uncertainty over regulation, electricity pricing and which projects will actually materialise. 

AI still accounts for a very small proportion of global power demand, but it is already changing power sector economics in areas of most intense development. For investors, this is creating opportunities far beyond the data centres themselves. A Wood Mackenzie Power & Renewables briefing on the subject held in Seoul in August examined the main issues. It focused on the US, which provides the clearest indication of the scale of the challenge, but also the APAC region, where AI demand is having a growing impact. The event also offered lessons for the power sector across the entire world. 

Wood Mackenzie expects a supercycle of infrastructure investment across the power sector, including renewables, gas-fired generation, energy storage, transmission and new technologies. Yet the range of possible outcomes has also widened, increasing the importance of understanding market structures, policy and project-specific risks.  

US large load capacity, including data centres, with signed construction or electricity supply agreements has reached 195 GW, while there has been advanced discussion over another 107 GW. Once projects at an earlier stage of development are thrown into the mix, the pipeline becomes much larger, underlining the danger of assuming that announced capacity will necessarily translate into operating assets.  

This matters for banks, investment funds and private equity because electricity availability is increasingly becoming a constraint on data centre investment. The intersection between data centre operators who place an exceptionally high value on reliable electricity and power systems that are struggling to provide additional capacity at pace is creating scope for new investment and commercial models.  

A power generation supercycle 

The result is likely to be an unusually broad expansion in generation. In the US, gas is forecast to provide 52% of the incremental energy production through 2035 and renewables another 45%. Annual gas-fired capacity additions could reach 15-20 GW through 2030 as manufacturing constraints ease.  

The opportunities on offer vary sharply by technology. US utility-scale solar additions are likely to plateau at around 26 GW per year as the industry navigates tax credit changes, supply chain challenges relating to import restrictions, and interconnection queues. Storage offers a smooth trajectory, with annual additions forecast to reach more than 22 GW by 2035, supported partly by demand from utilities and hyperscalers.  

Advanced nuclear could eventually provide another source of firm generation for data centres that require round-the-clock supply. Yet investors need to balance that potential against construction risk, possible cost overruns and shortages of skilled labour. The wider lesson is that rapidly growing electricity demand does not automatically make every generation technology – or every project – equally attractive.  

Grid investment is likely to be just as important. Long interconnection queues are already encouraging data centre developers to investigate behind-the-meter generation, interruptible connections and alternatives that make better use of existing infrastructure. Virtual power plants, for example, can aggregate batteries and flexible loads to create grid headroom, while grid enhancing technologies – such as dynamic line rating – can increase the capacity available on existing transmission infrastructure. 

What is the impact in the APAC region?  

The same trends are becoming increasingly visible in the Asia-Pacific region, which is emerging as the next growth frontier. Even allowing for some planned projects not being completed, Wood Mackenzie expects APAC data centre electricity consumption to more than triple from 260 TWh in 2025 to 980 TWh in 2030. That would be equivalent to about 6% of APAC power demand in 2030, roughly a Japan-sized block of new load.  

China accounts for the lion’s share of this increase, with annual data centre electricity consumption forecast to rise from 196 TWh in 2025 to 779 TWh in 2030. Outside China, demand is projected to increase by about 26% a year, from 63 TWh to 205 TWh over the same period.  

The region provides some extreme examples. For instance, on Indonesia’s Riau Islands, off Sumatra, the data centre utility capacity pipeline is on course to exceed the region’s existing installed generation capacity by 2027. New generation is therefore needed quickly if much of that pipeline is to proceed on schedule.  

Headline project pipeline figures can also give investors a misleading impression of the opportunities on offer. In Australia, the exceptionally high level of proposed capacity suggests that the current pipeline there is highly inflated. 

For investors, constraints can be both a risk and an opportunity. They can undermine the economics or timetable of an individual data centre development, while also increasing the value of generation, storage, grid connections and other enabling infrastructure. 

Australia illustrates the wider opportunity 

It is important to acknowledge that data centres only partly explain rising investment levels. Australia combines a large electricity market, a rapidly rising renewable energy penetration rate, ageing conventional generation projects and growing battery storage capacity, alongside rising data centre investment. 

The evolving generation mix creates opportunities for capital to replace retiring assets, provide flexibility and exploit increasingly volatile electricity prices. Battery storage illustrates how investment quality can depend on more than scale. Wood Mackenzie’s analysis of the ten highest revenue storage projects on the National Electricity Market, which accounts for 90% of Australian power supply, shows that smaller, well optimised assets can outperform larger projects, with trading strategy, location and price exposure all influencing returns.  

Who pays for the required power infrastructure? 

The scale of the new demand also raises a more fundamental financing question. US policymakers, for example, are increasingly concerned that residential consumers could bear some of the cost of supplying electricity to data centres. Regulators and utilities are responding with large load tariffs, new procurement mechanisms and rules governing interruptibility and grid connections for large loads. Large load tariffs are designed to ensure that data centres bear the costs associated with meeting their electricity demand, while new procurement structures can allow data centre operators to support specific generation investments.  

The same challenge is emerging in APAC. Investors must project not only future wholesale prices but whether tariffs are predictable, whether sufficient generation and transmission capacity will be available, and whether governments could change the balance of costs between data centres and other consumers.  

Malaysia demonstrates the complexity. Its Corporate Renewable Energy Supply Scheme allows data centres to procure renewable electricity but adds substantial grid charges. The cost of renewable energy generation, system access and conventional electricity tariffs can interact to produce very different final power costs in different markets. 

What opportunities does this create for suppliers? 

This investment cycle also extends into equipment manufacturing, with the rapid growth of battery storage highlighting both the opportunities and the risks. China currently dominates manufacturing capacity for many battery components, including anodes, separators and electrolytes, increasing pressure on other economies to develop more diversified supply chains.  

Governments are responding with industrial policies and localisation strategies, potentially creating domestic manufacturing opportunities. Yet successful localisation requires much more than subsidies. Technology readiness, research and development, manufacturing scale, trade relationships, skilled labour and access to low cost finance all determine whether potential new manufacturing centres can become competitive. That creates an important role for institutional capital, infrastructure funds and green finance, but also reinforces the need for carefully targeted investment. 

The emerging power sector investment cycle is therefore not simply a story of soaring electricity demand. The AI boom is accelerating a much broader reallocation of capital across generation, grids, storage and manufacturing. The winners will depend on where power can actually be delivered, how regulations allocate costs and which technologies can generate acceptable risk-adjusted returns. For investors, understanding those drivers may prove as important as forecasting the growth of AI itself. 

Fill in the form on this page to download a selection of slides from the Seoul presentation, containing more detailed insights, graphics and charts.