UCS flags grid and consumer-cost risks from planned US AI data-center growth

In an analysis of the US power-system implications of AI data-center expansion, the Union of Concerned Scientists argues that large new data-center loads will require utilities to build additional transmission infrastructure and generating capacity.

The group frames the issue not only as a question of supplying power to new campuses, but also of allocating the cost of grid reinforcements. Its publication raises the risk that infrastructure built to serve large data centers could shift costs to other electricity customers.

As an indication of the scale under discussion, the article cites an estimate that Amazon’s planned data centers could require 44 GW of power. The estimate applies to planned facilities, not operational capacity, and is presented by the source as a reported requirement rather than an official Amazon disclosure.

The publication also cites a speculative estimate of up to $7 trillion in spending on data centers and grid requirements over the next decade. That figure is explicitly qualified as dependent on assumptions, including AI-chip refresh cycles and the success of AI business models.

For operators and developers, the analysis suggests that grid interconnection, generation availability and the treatment of network-upgrade costs could become increasingly important determinants of development timing and project economics. For utilities and regulators, this could increase the prospect of tariff disputes and delays to supporting infrastructure projects.