Artificial intelligence has rapidly moved from innovation labs into the core of corporate strategy. Boards are approving AI investments to improve productivity, accelerate decision-making and create new business models. Yet beneath the excitement lies a less discussed challenge that is increasingly attracting the attention of policymakers, energy regulators and risk professionals worldwide: AI’s growing appetite for electricity.
For years, digital transformation was largely viewed as an efficiency story. AI, however, is introducing a new variable into the risk equation. The infrastructure required to train and operate advanced AI models is significantly more energy intensive than traditional computing. As organisations scale AI adoption, energy availability is emerging as a strategic consideration rather than merely an operational cost.
The numbers are becoming difficult to ignore. According to the International Energy Agency (IEA), global data-centre electricity consumption is expected to more than double to around 945 terawatt-hours (TWh) by 2030, slightly exceeding the current annual electricity consumption of Japan. AI-driven computing is expected to account for a substantial portion of this increase.
This is no longer simply an environmental debate. It is increasingly becoming a business continuity and infrastructure resilience issue.
Across the United States, power regulators are already grappling with the implications of AI-driven demand growth. The Federal Energy Regulatory Commission recently directed grid operators to review rules governing connections for large electricity consumers, particularly data centres, amid concerns about grid reliability and infrastructure capacity.
The challenge is not confined to North America. Ireland has become a widely cited example of how concentrated data-centre growth can place pressure on national electricity systems. Data centres accounted for approximately 17% of Ireland’s electricity consumption in 2022, highlighting how digital infrastructure can become a material component of national energy demand.
Technology companies themselves are responding. Microsoft has entered long-term nuclear-energy arrangements, while Amazon, Google and other hyperscalers continue to invest heavily in renewable energy procurement and more efficient cooling technologies. These initiatives reflect a growing recognition that access to reliable power may become as strategically important as access to computing capacity itself.
For boards and risk committees, the implications extend well beyond sustainability reporting.
First, there is infrastructure risk. AI strategies increasingly depend on data-centre capacity that may be constrained by local grid limitations. Delays in power availability could slow expansion plans, increase project costs or create concentration risks in specific regions.
Second, there is transition risk. Many organisations have committed to ambitious net-zero targets. As AI workloads expand, energy consumption and associated emissions could rise unless matched by renewable-energy sourcing and efficiency improvements. Investors and regulators are likely to scrutinise this gap more closely in the years ahead.
Third, there is resilience risk. Research increasingly suggests that AI infrastructure is becoming intertwined with power-system planning itself. Localised shortages, grid disruptions or energy price volatility could have direct implications for AI-dependent operations.
The lesson for boards is straightforward. AI governance cannot focus solely on ethics, cybersecurity and model risk. Energy resilience must become part of the conversation.
The next phase of AI adoption will not be determined only by algorithms, talent or investment capital. It will also depend on whether businesses and nations can secure the electricity required to power an increasingly AI-driven economy.
In the race to deploy AI, the real constraint may not be computing power. It may be power itself.
