Michelle Yin: America is Fighting About the Wrong End of the Wire
The real broadband fight isn't over data centers but over whether $21 billion in leftover BEAD savings goes toward helping people actually use the networks being built.
If you want to see what Americans think of the AI economy, skip the earnings calls and head to a county zoning meeting. In the first three months of this year alone, local opposition blocked or delayed 75 data-center projects worth $130 billion, as many as in all of last year, and more than a hundred communities have enacted moratoriums.
Members of Congress have proposed pausing large-scale AI data-center construction nationwide even as the White House pushes to accelerate it. Hyperscalers will pour roughly $700 billion into AI infrastructure this year, and the country is in open revolt over where it lands, what it does to electric bills, and who actually benefits.
I understand the anger, but as an economist who studies what digital infrastructure actually does to people’s economic lives, I think this fight is aimed at the wrong end of the wire. The supply side of the buildout will get sorted by markets, courts, and utility commissions.
The question that will decide whether ordinary Americans gain or lose from the digital economy is on the demand side — who can afford to connect, on what device, with what skills, in a labor market increasingly run by algorithms. And that decision is sitting quietly on a desk at the Commerce Department right now.
The evidence on what is at stake is finally in. With my colleague Diego Guerrero, I studied the Connect America Fund, a $9 billion federal program that wired rural America between 2015 and 2021. Because eligibility was set by an engineering formula built on terrain, soil, and bedrock rather than by politics, the rollout functioned as a natural experiment, letting us isolate the causal effect of broadband from everything else that changes when it arrives.
The results upend the standard story. For the general population, new broadband barely moved employment at all. For workers facing steep barriers to job search, measured here among workers with disabilities, labor force participation jumped by as much as 4.7 percentage points, several times the measured employment effect of the Americans with Disabilities Act. The mechanism was not remote work. If anything, telework grew less for these workers than for the general workforce.
The internet mattered because it let people who had been priced out of searching for accessible employers and jobs within reach finally find work near home, in construction and food service as much as in offices. All told, the program created jobs at roughly $21,000 to $47,000 each, a fraction of the $92,000 per job-year the government’s economists attributed to the 2009 stimulus.
Infrastructure returns on whether it gets built
Infrastructure’s returns are not set by how much gets built. They are set by who is on the other end of it, and whether they can actually use it.
The same study carries a warning. When the wire arrived, better-off households upgraded to fast plans while households that included a person with a disability did not. They took the cheapest tier and owned fewer devices to begin with, so the same infrastructure produced diverging outcomes.
Build the network and nothing else, and you widen the very gap you claimed to close at exactly the moment when AI-screened hiring, online-only job postings, and digital-first government services are raising the price of being on the wrong side of it.
That brings us to the decision nobody is protesting. Last year, the Commerce Department restructured the $42.45 billion Broadband Equity, Access and Deployment program to award grants on lowest cost alone, stripping out the requirements for low-cost plans, digital adoption, and workforce development — the entire demand side of the program.
That restructuring produced roughly $21 billion in projected savings, and the National Telecommunications and Information Administration is right now deciding what to do with it. The options on the table include more construction, middle-mile projects, and, crucially, digital adoption and workforce programs. Construction on the main awards begins this summer, so the concrete is literally being poured while the question of who will actually be able to use it remains unanswered.
There is a version of this decision for every persuasion. For fiscal conservatives, our targeting simulations show that weighting broadband dollars toward communities with high concentrations of search-constrained workers, including people with disabilities, workers reentering after incarceration, and displaced workers, generates about 6 percent more jobs per dollar than allocating by coverage gaps alone, at zero additional cost.
For anyone worried that AI will concentrate wealth and power, this is the rare lever that pushes the other way, because the documented gains from connectivity flowed overwhelmingly to the most marginalized workers, in the kinds of jobs no data center will ever automate away. And for the communities packing those zoning meetings, it finally answers the question they are really asking, which is what any of this buildout does for them. A tax abatement for a server farm that employs forty people never will.
My suggestion to Congress and the administration is straightforward. Keep fighting about siting and ratepayers if you must, since those fights have merit. But do not let the loudest infrastructure argument crowd out the most consequential one. Direct a serious share of the $21 billion to adoption, devices, and digital workforce capacity, targeted where the barriers to using the network are highest. Require states to report who actually subscribes, at what tier, and on what devices, not just where the map turned green.
The supply side of the AI economy is being built at historic speed either way. Whether it becomes an engine of broad opportunity or a $700 billion monument to divergence is being decided now, on the demand side, in an allocation memo most Americans will never read. It deserves the crowd.
Michelle Yin is a director and professor at Northwestern University’s School of Education and Social Policy, where she leads the RISEI Lab, which studies pathways to employment for people with disabilities and other underserved populations. Her study of the Connect America Fund’s labor-market effects, coauthored with Diego A. Guerrero, is available as a working paper. The views expressed are the author’s own. This Expert Opinion is exclusive to Broadband Breakfast.
Broadband Breakfast accepts commentary from informed observers of the broadband scene. Please send pieces to commentary@breakfast.media. The views expressed in Expert Opinion pieces do not necessarily reflect the views of Broadband Breakfast and Breakfast Media LLC.
