The dirty parts of the computing world, revisited

Bulletin of the Atomic Scientists | Nathan Ensmenger

Data center expansion across the United States has transformed computationally intensive digital technologies into a major environmental and political battleground. Massive power usage and freshwater consumption from artificial intelligence infrastructure now trigger fierce local opposition, state ratepayer protection laws, and federal policy scrutiny. This growing friction stems from treating hyperscale facilities as weightless digital nodes rather than resource-heavy industrial factories.

Data centers operate as extractive enclaves. While host communities absorb severe groundwater depletion, backup generator air pollution, and strained utility grids, corporate profits and computational services flow directly to distant technology hubs. Local planning commissions in regions like Loudoun County currently struggle to regulate national energy infrastructure and industrial policy using simple special-use permits. Without binding national regulatory frameworks, independent environmental monitoring, and equitable cost-sharing rules, localized municipal rejections will fail to reduce overall ecological harm, merely rerouting industrial burdens to economically vulnerable communities and less regulated international jurisdictions.

Comment

The rapid concentration of hyperscale AI infrastructure in Northern Virginia exposes severe vulnerabilities in grid interconnection queues and transformer manufacturing supply chains. In Loudoun County, massive power allocations for AI workloads strain existing transmission capacity faster than regional utilities can construct high-voltage lines. This bottleneck demonstrates how digital infrastructure scaling is constrained by physical industrial manufacturing rather than software innovation.

This physical capacity wall creates systemic delays for dual-use defense computing and federal national security workloads competing for the same regional power allocation. As commercial data centers absorb available megawatt capacity across PJM Interconnection, federal facilities face extended procurement lead times and higher base electricity tariffs. Strategic computing expansion across Loudoun County ultimately hinges on industrial transformer production capacity rather than semiconductor availability.

Strategic Question for Discussion

Which factor presents the greater long-term constraint on national hyperscale computing capacity — regional grid interconnection delays in hubs like Loudoun County or global supply bottlenecks in high-voltage industrial transformer manufacturing?

The available evidence points toward transformer manufacturing lead times as the primary structural bottleneck, as production cycles for high-voltage equipment currently exceed four years. While regional grid operators across PJM Interconnection can adjust local queue priorities, physical equipment shortages directly cap total grid additions regardless of regulatory policy. This trajectory indicates that industrial manufacturing throughput, rather than local zoning approvals, will dictate the pace of digital infrastructure expansion.

Share your assessment in the comments below.

To read in details click here