Americans Want AI's Benefits Without Costs

 

“Jay Reed's op-ed "Abandoned Land? Build a Data Center" (Oct. 2) correctly argues for placing data centers on unused land. The data-center conflict is a textbook public-goods problem, in which AI users have an incentive to free-ride.

 

Most Americans want the shared benefits of AI but would prefer that others bear the local costs of data centers.

 

Fifty-two percent of U.S. workers use AI on the job, and 15% use it daily, according to Gallup. McKinsey & Company finds that nearly 90% of organizations use AI for at least one function. Verasight reports that AI adoption has grown roughly twice as fast as smartphone adoption did. If demand indicates value, Americans have spoken loudly. It's no wonder that investment in AI data centers has exploded, with computing infrastructure reaching 1.5% of gross domestic product in the first quarter of 2026.

 

Yet roughly 70% of Americans say they oppose data centers being sited near them.

 

Officials have been quick to respond to constituents' concerns. New York imposed a statewide moratorium on large data centers in July, and Texas's August directive halted as many as 1,800 projects. Pennsylvania lawmakers are weighing a bill that would let municipalities follow suit. More than 400 local moratoriums have been recorded across 44 states. These states are free-riding on other states' advancements in AI.

 

The data-center problem, however, is easier to solve than many other public-goods problems. Compared with other capital investments, data centers have less need to be near population centers. An auto plant must be near workers, a port near water, a hospital near patients. Data centers face no such constraints.

 

The federal government owns about 28% of the nation's land, much of it in the sparsely populated West.

 

Some of this land would be well suited to data centers. The Trump administration has already started down this path by selecting four sites for which it will solicit private data-center development.

 

Disputes over behemoth data centers seem unlikely to last. In 1997 the world's fastest computer occupied about 2,500 square feet, but today we have that computing power in our pockets. AI is advancing even faster. The cost of running a query at GPT-3.5-level performance fell more than 280-fold between 2022 and 2024. Innovation will mean smaller centers, making their location less of a concern.

 

State and local governments are blocking data centers to avoid the perceived local costs of a technology that half of working Americans use. To resolve the free-rider problem, put servers on vacant federal land and keep the rules flexible as the centers shrink.

 

Tomas J. Philipson

 

Chicago

 

Mr. Philipson served as a member and acting chairman of the President's Council of Economic Advisers 2017-20 and serves on Anthropic's Economic Advisory Board.” [1]

 

1. Americans Want AI's Benefits Without Costs. Wall Street Journal, Eastern edition; New York, N.Y.. 07 Oct 2026: A16.