Who Builds the Data Centre? The Consumer Paradox Driving Infrastructure Conflict
Every time someone opens Google Maps, streams a film, or asks ChatGPT to draft an email, they are placing a demand on physical infrastructure that many of them would rather not see built near their homes. That contradiction is now generating real political friction — and the data centre industry is beginning to reckon with it.
At a US congressional hearing earlier this year, Representative Alexandria Ocasio-Cortez held up two jars of murky water drawn from private wells in Morgan County, Georgia, near one of Meta’s data centre campuses. Meta has disputed any causal link between its construction activity and the contamination. But the image landed, and it illustrates a tension that is becoming structurally harder to dismiss as hyperscalers accelerate their buildout programmes across North America, Europe and Asia-Pacific.
Mehdi Paryavi, founder and CEO of the International Data Center Authority (IDCA), is direct about where responsibility originates. “We are the reason that data centers get built and expanded,” he says. Consumer appetite for cloud compute, AI services, streaming, e-commerce and digital banking has created demand that the industry is simply fulfilling. Facebook accounts, Amazon shopping carts, Netflix queues — none of these exist without physical infrastructure somewhere absorbing the load.
The Paradox Has a Name, and It Points Back at Users
Paryavi frames the public opposition to data centre projects as a form of digital hypocrisy, though he stops short of using that framing to excuse the industry. When residents protest a new facility by posting their grievances on social media platforms that themselves depend on data centres, the logical contradiction is apparent. When activists push back against infrastructure expansion while using AI tools for daily tasks, the tension compounds. This is not a rhetorical point designed to silence critics — it is a structural observation about how modern digital consumption is organised, and how poorly that organisation is communicated to the public.
The more substantive argument is about location and governance, not existence. Paryavi is clear that the debate has moved past the question of whether additional compute capacity is needed. Hyperscalers plan buildouts years in advance; chipmakers are running significant backlogs; enterprise and consumer AI demand is not softening. The operative question is where that infrastructure goes and how affected communities are consulted before ground breaks, not after opposition mobilises.
On the persistent misconception around water usage, Paryavi offers a figure that rarely surfaces in mainstream coverage: data centres worldwide consume approximately 0.1% of water available for human use. Modern AI and cloud facilities increasingly use closed-loop cooling systems that require no operational water. He acknowledges the complexity, however — a data centre powered by nuclear energy may emit no greenhouse gases directly, but nuclear generation itself draws heavily on water resources. IDCA’s position is that these interdependencies require broad, informed public engagement rather than simplified headline figures that strip away context.
Governance, Siting and the Case for Digital Hubs
The IDCA’s preferred policy response is master planning at the national or regional level — the deliberate designation of concentrated digital infrastructure zones that aggregate compute capacity in locations chosen for their suitability rather than land-cost arbitrage. Paryavi points to Loudoun County in Virginia as a precedent: a jurisdiction that built an entire economic identity around data centre density, creating community rather than disrupting it.
The case for this model is partly logistical. Concentrated hubs reduce the unit cost of provisioning power, cooling and connectivity. They create economies of scale that dispersed, ad hoc development cannot match. They also simplify a national security calculus that the industry is increasingly taking seriously — securing a small number of defined digital zones is materially more tractable than protecting thousands of distributed sites during periods of geopolitical tension.
Where public distrust has hardened into active opposition, Paryavi identifies a consistent failure mode: providers that did not invest in early community dialogue. His prescription is straightforward.
On the question of naming — whether “data factory” would better capture what large-scale AI facilities actually do — Paryavi is dismissive of the rebranding impulse. The term “data centre” encompasses a spectrum that runs from enterprise server rooms to hyperscale campuses to USB drives. What differentiates facilities is not size but the type of data they house or process; a small enterprise centre holding sensitive health records may be more mission-critical than a large AI training cluster. Collapsing that spectrum into a manufacturing metaphor serves simplicity at the expense of accuracy.
The deeper issue is that the industry’s communication deficit has allowed others to define it. Jars of murky water in a congressional hearing room are a more compelling narrative than closed-loop cooling specifications. Until operators invest in closing that gap — systematically, not reactively — the paradox Paryavi describes will continue generating friction that no amount of technical justification will resolve on its own.
