For over a decade, the digital economy has been driven by a key assumption: that computing power is scalable, cloud infrastructure is flexible, and innovation can grow almost without limits. That belief is now being challenged. As artificial intelligence boosts the demand for data processing, electricity — often considered a background input — is becoming a vital strategic issue.
At the same time, global energy systems are entering a new phase of volatility. The escalating conflict in the Middle East, and notably the closure of the Strait of Hormuz, once again exposes the fragility of our energy supply chains. In this context, the exponential growth of energy-hungry AI infrastructure is no longer just a technological or environmental issue; it is increasingly a matter of geopolitical risk, economic resilience, and ultimate survival.
Big Tech companies try to secure their energy
Large technology companies are already responding. Microsoft has signed an agreement to restart a reactor at the Three Mile Island Nuclear Generating Station in the United States. Meta has announced plans to secure up to 6.6 gigawatts of nuclear power capacity. Meanwhile, Google and Amazon are investing heavily in Small Modular Reactors (SMRs), an emerging nuclear technology unlikely to be deployed at scale before the end of the decade.
These moves signal a structural shift. Access to reliable electricity is no longer simply an operational concern — it is becoming a source of competitive advantage and strategic autonomy.
AI models are growing larger and more energy-intensive, while hyperscale data centres require vast and continuous power. Yet expanding generation capacity, reinforcing transmission networks or deploying new nuclear technologies takes years, sometimes decades. The mismatch is stark: the demand curve for computing is rising far faster than the supply curve for energy — precisely at a moment when geopolitical instability is making long-term energy planning more uncertain.
A new political economy of AI
This dynamic is giving rise to a new political economy of AI. The debate is no longer confined to technological leadership or market share. It is increasingly macroeconomic, infrastructural and geopolitical.
Policymakers might soon face difficult questions. AI development could come to be treated as a priority use of electricity, comparable to historically protected industrial sectors. Differentiated pricing for data centres may also be considered as a way to manage pressure on national grids. Over time, governments might even explore sector-specific limits to ensure that digital infrastructure does not crowd out essential household or industrial consumption.
Indeed, AI was widely framed as a productivity-enhancing general-purpose technology. However, it is also becoming a major area of consumption—one that directly competes with other economic and social priorities.
As a result, the foundations of competitiveness are evolving. While human capital, research ecosystems and venture capital remain essential, they are no longer sufficient on their own. A country’s ability to secure a stable, abundant and low-carbon electricity supply to meet this increased demand is becoming just as critical, particularly in a more volatile geopolitical environment.
From this perspective, industrial policy and energy policy are once again becoming inseparable. Governments seeking to attract AI investment must now think not only about regulatory frameworks and innovation incentives, but also about generation capacity, grid resilience and exposure to global energy shocks. Technology strategy is converging with infrastructure planning in ways that recall earlier industrial eras.
The end of unlimited growth
The narrative of unlimited digital expansion is giving way to the reality of physical constraints. Electricity is not infinitely scalable. Transmission networks cannot grow at the pace of venture capital funding rounds.
In this context, calls for “digital sobriety” — reducing unnecessary computational use — may gain traction. Yet restraint is unlikely to emerge primarily from ethical appeals. More plausibly, it might result from economic adjustments, political negotiations and the reallocation of scarce resources.
Ultimately, the question facing both governments and companies is not just technological — it is deeply political. As AI becomes more embedded in the economy and in everyday life, societies may have to make clearer choices about how energy is prioritised.
In a world of accelerating digital demand, fragile energy systems and mounting geopolitical tensions, the dilemma could increasingly be framed in very human terms: heating or AI?
This is not just about electricity costs or technological competitiveness. It raises a broader question about how essential resources are allocated in an era of systemic uncertainty. Policymakers may soon face difficult choices, as disruptions to energy markets, climate pressures on infrastructure and food systems, and growing competition for resources all converge. At that point, the issue becomes less abstract — and more about what societies are prepared to prioritise.
The expansion of AI infrastructure may begin to intersect directly with other fundamental needs — from food production and industrial activity to housing, mobility and basic public services. Governments and citizens alike may find themselves confronting trade-offs that were previously unthinkable.
