The IEA’s own reporting through 2026 is consistent and unambiguous: the three biggest structural drivers of rising global electricity demand are air conditioning, data centres, and electric vehicles, growing global consumption at 3.6-3.8% a year — around fifty percent faster than the average rate over the previous decade. Look closely at who actually owns and uses each of those three things, and an uncomfortable question follows naturally: is the world’s electricity infrastructure being expanded, at everyone’s cost, primarily to serve a wealthy minority’s consumption?
Air Conditioning: A Genuine Divide, With a Real Nuance Worth Getting Right
Start with the starkest number available: the United States alone uses as much electricity cooling buildings as the entire continent of Africa uses for everything combined. That’s not a comparison of similar things — it’s one country’s air conditioning bill against a continent’s total electricity consumption.
Global AC ownership sits at roughly 27% of households today, projected to reach 41% by 2050. But that average conceals a genuinely stark distribution. Projections show at least 80% of the world’s richest 10% of households will have air conditioning by 2050 — compared with somewhere between 2% and 23% of the poorest 10%, depending on the region. In East Asia and the Pacific, AC ownership runs at roughly 22% among the lowest-income households against nearly 80% among the wealthiest — in the same countries, under the same climate. In India, only around 16% of households currently own a unit, despite among the most severe heat exposure on the planet. In sub-Saharan Africa, ownership outside the wealthiest quintile is described as “extremely limited,” not primarily due to cost, but because a meaningful share of the population still lacks reliable access to electricity at all.
It’s worth being fair here, though: unlike data centres or EVs, AC access genuinely is, in significant part, a life-and-death issue rather than a comfort upgrade — the WHO attributes almost half a million deaths a year globally to heat, and expanding cooling access to the world’s poorest, hottest regions would save lives, not just add convenience. The real inequality isn’t that AC exists. It’s that the wealthy already have it saturating their homes and workplaces at multiples of what any developing nation’s poorest households can access, while the grid investment needed to expand that access globally gets priced and financed as though the benefit were evenly shared.
Data Centres: A Luxury by Almost Any Definition
This site has already covered this territory in detail, and the numbers bear repeating in this context specifically: AI data centres are projected to consume enough water by 2030 to meet the basic domestic needs of every one of the 1.3 billion people in Sub-Saharan Africa for a year, and nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria — over 650 million people. Data centre electricity demand accounts for less than 10% of the total growth in this IEA forecast, smaller in raw terms than AC or EVs — but it’s the most geographically concentrated and the most narrowly beneficial. The internet access gap runs 94% in high-income countries against just 23% in low-income ones; over 2.2 billion people remain offline entirely. The infrastructure straining global grids to power AI systems serves, overwhelmingly, people who already have every other advantage this list describes.
Electric Vehicles: Bought Almost Exclusively by the Comfortable
The academic research on this is remarkably consistent across countries: EV adoption is concentrated among higher-income, better-educated households, in single-family homes, with home charging access — and this isn’t a temporary early-adopter phase that later evens out. Studies from the US, Australia, and China all find the same pattern: EV buyers are wealthier, own more vehicles per household already, and disproportionately benefit from tax incentives and charging infrastructure that itself gets built preferentially in wealthier neighbourhoods. One Australian study bluntly summarised its own finding: unlike solar panels, which lower-income households adopted specifically to cut energy costs under financial pressure, EV uptake ran in the opposite direction — a purchase made by households with spare financial capacity, not one made out of need.
Who Actually Pays for the Grid That Serves This
Here’s where the inequality compounds rather than simply sitting alongside itself. Expanding electricity grid capacity to meet this demand growth — new generation, new transmission infrastructure, new storage — is a cost that gets socialised across entire national electricity systems, built into the price every household pays per unit, not billed specifically to the households and data centres actually driving the growth. A household that owns no air conditioner, drives no EV, and has never used a cloud AI service still pays its share of the grid investment required to serve the neighbours, corporations, and data centres that do. In the poorest households — those least likely to own any of the three things actually driving demand growth — a rising electricity bill is a larger share of already-stretched income, funding infrastructure disproportionately built to serve consumption they don’t and often can’t participate in.
The Honest Conclusion
This isn’t a case for banning air conditioning, data centres, or electric vehicles — each of them serves real purposes, and in AC’s specific case, expanding access to the world’s poorest hot-climate regions would be a genuine, measurable public health win. But the framing of “electricity demand is growing, therefore we all need to invest more in the grid” obscures something the underlying data makes fairly plain: a significant share of that growth is driven by consumption patterns heavily concentrated among people who already have the most, while the cost of building the infrastructure to support it gets spread across everyone, including people who will likely never own any of the three things actually driving the bill upward. That’s not a technology problem. It’s the same distribution question this site has raised about AI’s resource costs more broadly — who benefits, and who pays — just showing up again, this time in the electricity bill itself.
