Rapidly expanding AI computing infrastructure is accelerating investment in dedicated power systems as developers confront grid connection delays and unusually demanding electrical loads. The resulting market for microgrids, energy storage and advanced power conversion is attracting major investment.
The rapid expansion of artificial intelligence is creating a new market for electricity infrastructure, with data centre developers increasingly looking beyond conventional grid connections to secure the power needed for large computing facilities.
Microgrids, combining local generation, energy storage and sophisticated control systems, are emerging as a viable option. Unlike conventional backup arrangements, they can be designed to manage electricity continuously, coordinate different power sources and operate alongside the wider network.
Investment in the technology is accelerating. According to research company GlobalData, the United States microgrid market is expected to grow from approximately US$8 billion in 2025 to US$21.2 billion by 2030.
A Reuters report published on 26 October identifies several major transactions associated with the growing market. These include Vertiv's US$2.6 billion acquisition of Utility Innovation Group and EQT's acquisition of Scale Microgrids in 2025.
The investments reveal the changing way data centre operators are approaching electricity supply. Securing sufficient computing equipment is only one part of developing an AI facility. Developers also need reliable electrical capacity, often within timeframes that conventional network expansion cannot accommodate.
A different kind of electrical load
AI data centres present challenges that extend beyond the total amount of electricity they consume. Large clusters of graphics processing units (GPUs) can produce rapid changes in electrical demand as computing workloads begin, end or change intensity. These variations can occur over extremely short periods and create difficulties for equipment designed around more predictable operating conditions.
The American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) identifies millisecond-scale load variations associated with AI training as a potential threat to electrical stability. Managing changes like these requires more than installing sufficient generation capacity to meet the facility's maximum demand.
Power conversion equipment, energy storage and control systems must be capable of responding to fluctuations while maintaining appropriate voltage, frequency and power quality. If poorly managed, large and rapidly changing loads can also affect the electricity network to which a facility is connected, making electrical system architecture increasingly important in data centre design.
From emergency backup to active power management
Traditional data centres have historically relied on uninterruptible power supplies (UPSs) and standby generators to maintain operations during interruptions. However, a microgrid can extend this approach by coordinating generation, storage and consumption during normal operation as well as during outages.
Battery systems can absorb short-duration demand changes, support power quality, and reduce the peak load drawn from the network. Local generation can provide additional capacity, while control systems determine how the different resources operate together. Some configurations also allow facilities to disconnect from the wider grid and continue operating independently for a defined period, provided sufficient generation and stored energy are available.
The US Department of Energy has identified modular microgrids, direct-current architectures and coordinated control as potential ways of accelerating the development of data centres and other large electricity consumers. The department notes that microgrids can offer shorter deployment timelines than conventional transmission and distribution expansion in suitable circumstances.
However, building dedicated power infrastructure does not eliminate every constraint. Equipment availability, fuel supply, permitting, and generation costs can still determine whether a project is commercially viable.
Direct-current systems attract attention
The growth of AI infrastructure is also renewing interest in direct-current (DC) power distribution. Data centre computing equipment ultimately operates on DC electricity, although conventional facilities generally receive alternating-current (AC) power and use multiple conversion stages to supply their equipment.
DC-based architectures can reduce the number of conversion stages and potentially improve efficiency, depending on system design.
ABB is among the companies developing power technologies for increasingly demanding AI facilities, including DC solutions and equipment intended to support grid stability.
Advanced converters and coordinated energy storage may also enable data centres to respond more actively to conditions on the external electricity network. ASHRAE's work on grid-interactive data centres describes how facilities could eventually provide services such as controlled demand reduction, fast load response and other forms of flexibility.
These capabilities depend on appropriate technical systems, operating agreements and market rules. They should not be confused with the basic ability to maintain power during an outage.
A growing infrastructure market
The scale of proposed developments indicates why electrical equipment suppliers and infrastructure investors are paying attention.
Reuters reports that Scale Microgrids is developing multiple projects ranging from approximately 200 MW to more than 1 GW, while other developers are exploring combinations of onsite generation, storage and existing industrial power infrastructure.
For suppliers, this creates opportunities extending beyond generation equipment. Transformers, switchgear, power electronics, batteries, protection systems, monitoring equipment and microgrid controllers all become important components of the electrical architecture.
The speed of investment is also creating pressure on equipment supply chains. Competition for electrical equipment is contributing to longer delivery times and higher costs in some markets.
Implications for South Africa
South Africa's expanding data centre industry faces many of the same questions, although local project economics, electricity market arrangements and network conditions differ from those in the United States. The country already has experience with private generation, embedded energy systems, battery storage and electricity wheeling. These capabilities provide a foundation for more sophisticated power arrangements at large commercial and industrial facilities.
For AI data centres, the challenge is to combine sufficient electrical capacity with the reliability and power quality required by sensitive computing equipment. A dedicated microgrid could help manage onsite generation, storage and rapid load variations, but its value would depend on the site's connection arrangements, operating requirements and access to dependable energy sources.
There is also an important distinction between installing backup equipment and creating an actively managed electrical system. A facility with diesel generators and a UPS is not automatically operating a microgrid capable of balancing generation, storage and demand.
As AI computing expands, the electrical infrastructure serving data centres is becoming a specialised engineering discipline in its own right.
The emerging opportunity is not simply to supply more electricity to computing facilities. It is to design power systems capable of delivering that electricity with the speed, stability and flexibility increasingly required by AI workloads.
