Data Centres Are Driving Up Power Bills Globally — What a New Study Means for Pakistan's AI Push

A new global study reviewed in Engineering Review concludes that the rapid expansion of data centres — particularly those built to host artificial intelligence training and inference workloads — is becoming a measurable contributor to electricity demand growth and, increasingly, to upward pressure on consumer power bills. For Pakistan, sitting at the early stage of a national AI build-out and with utility tariffs already at the top of South Asian comparisons, the question is whether the same effect will materialise locally over the next decade.

Globally, data-centre electricity consumption has roughly doubled in five years. In the United States, several state regulators have already approved or are reviewing tariff structures that pass elevated grid investment costs through to retail customers, with utilities arguing that residential and commercial users should not subsidise the connection of hyperscale data centres whose load profiles strain the existing infrastructure.

Why data centres matter to the grid

Data centres have three properties that make them different from typical industrial loads:

• Magnitude — a single hyperscale facility can draw 100 to 500 MW continuously, comparable to a large industrial city's full load.
• Constancy — unlike a textile mill that ramps up and down, a data centre runs at near-constant load 24/7, with the grid needing to provide that baseload regardless of solar availability.
• Reliability sensitivity — data centres need higher reliability than typical industrial loads, often justifying utility investments in dedicated feeders, redundant transformers, and reinforced transmission paths that other customers ultimately help finance.

The mechanism by which bills rise

When data centres connect at scale, three cost categories climb: capacity charges (because total system peak demand increases), transmission and distribution investment (because the network must absorb the new load), and increasingly the marginal cost of generation (because the cheapest existing capacity is fully utilised, pushing dispatch up the merit order to more expensive generators). All three flow through to consumer tariffs over time via the regular review process.

The Pakistani context

Pakistan's data-centre footprint today is small relative to its overall industrial load — a handful of operators serving banking, telecom, and government workloads. But two simultaneous policy moves are pointing toward rapid growth: the federal government's plan to deepen the digital economy infrastructure, and the recently announced Chinese AI hub framework that is bringing IBI Beijing United Technology and other Chinese tech firms to set up Pakistan-based AI operations.

What that could mean for Pakistani tariffs

If Pakistan attracts several hundred megawatts of data-centre load over the next five years — a credible scenario given the policy direction — the same global dynamics will start playing out locally. NEPRA will face the same regulatory question other regulators are facing: should hyperscale customers be required to pay direct connection charges that reflect their actual cost-of-supply, or should those costs be socialised across the consumer base and recovered through the regular tariff process? The answer materially affects what residential and small-commercial customers see on their bills.

The renewables angle

The global pattern shows that hyperscale operators are increasingly choosing to underwrite renewable generation projects directly — power purchase agreements with solar and wind developers — to lock in cost-stable supply and meet sustainability commitments. If Pakistani data-centre operators follow that pattern, the side effect would be accelerated solar and wind buildout to specifically serve their load, which dampens the cost-pass-through effect on other customers. That is the optimistic version of the same story.