BS Summary: This article contains 1 faulty reasoning type, including Indoctrination, with Indoctrination as the most egregious example at 5.3% saturation with 123 hits. Analysis detected 123 faulty-reasoning hits from 2,309 analyzed words, generating a BS Score of 42.6% and a BS Rank of 34% (9,740 of 14,612 articles). This article is better (less manipulative) than 66.70% of the article peer group.
A Better Way to Build AI
America’s Dominance Depends on Getting Local Communities on Board
An anti–data center sign outside a home in New Carlisle, Indiana, March 2026 Jim Vondruska / Reuters
SARAH KREPS is John L.
Wetherill Professor and Director of the Tech Policy Institute at Cornell University and a Nonresident Senior Fellow at the Brookings Institution.
She served in the U.S.
Air Force from 1999 to 2007.
Print Subscribe to unlock this feature or Sign in .
Save Sign in and save to read later
Article link: https://www.foreignaffairs.com/united-states/better-way-build-ai https://www.foreignaffairs.com/united-states/better-way-build-ai
This is a subscriber-only feature.
Subscribe now or Sign in .
Chicago Cite not available at the moment
MLA Cite not available at the moment
APSA Cite not available at the moment
APA Cite not available at the moment
Request reprint permissions here .
Like many in the community, I value Cayuga Lake not as an abstraction but as something irreplaceable that we have a duty to protect.
At the same time, the region’s economic realities are difficult to ignore: a shrinking tax base, aging infrastructure, and rising public costs that have strained local government for years.
As both a resident who would live with the project’s consequences and a tech policy scholar who studies the governance of digital infrastructure and has a background in environmental impact, I offered to conduct an independent assessment of the project’s likely effects on water use, power demand, employment, and the local tax base.
(The company agreed and compensated me for the time required to conduct the analysis.
It had no role in determining the scope of the study, conducting the research, interpreting the findings, or drafting the report, and it exercised no editorial control over the final product.)
I found that the engineering, on its own terms, was sound.
The proposed closed-loop cooling system would largely eliminate operational water withdrawals from Cayuga Lake, replacing the retired coal plant’s once water-intensive operations.
Similarly, although the data center campus’s electricity demand would be substantial, the site already has considerable transmission infrastructure and is located within a regional grid in which electricity generation is predominantly carbon free and sizable surplus capacity exists.
Regulators and developers must treat local buy-in as an asset rather than a hurdle.
Much of the opposition ran on misinformation and the conflation of project-specific issues with broader anxieties.
Some concerns reflected a general apprehension about AI, corporate power, and environmental change that extended beyond the project itself.
Questions about surveillance, workers being displaced by AI, and environmental harm became intertwined with debates over water use, electricity demand, and land use.
And yet the skeptics were not wrong about everything.
Permanent employment at the new facility would likely settle at around 75 specialized roles, meaningful for a town absorbing industrial closures but far from a major jobs engine.
Estimates of how much tax revenue the data center would generate varied as well, given corporate exemptions and incentives.
More important, the real disagreement was not necessarily about gallons of water or megawatts of power.
It was about control—whether decisions affecting the lake, the grid, and the land would be shaped primarily by local institutions or by national-scale companies and technology developers.
The technical details mattered, but only insofar as they reflected larger questions of trust, representation, and agency.
In that sense, the Cayuga debate reflects a broader dilemma facing the United States.
As AI infrastructure becomes essential to the United States’ technological leadership, the country’s success will hinge not only on capital, energy, and technology but also on the capacity of democratic institutions to reconcile national ambition with local consent.
FAULT LINES EVERYWHERE
The tensions that surfaced in upstate New York are not isolated.
Across the United States, communities that once competed to attract technology investment are now reassessing what kinds of infrastructure they are willing to host.
Data centers —once treated as benign, even invisible—have become flash points in local politics, exposing a growing gap between national technological ambition and local governance.
In Chandler, Arizona, home to Intel’s semiconductor manufacturing campus, the city council rejected a proposed large data center complex backed by Meta, Microsoft, and major venture firms.
Opponents cited concerns about water consumption and heat emissions in a desert city already confronting chronic drought.
What AI advocates framed as economic development, residents and council members recast as a risk to long-term environmental stability.
A similar dynamic has unfolded in northern Virginia, home to the world’s largest concentration of data centers, with more than 250 facilities built in the past decade and hundreds more proposed or under review.
There, resistance to data center expansion has become a defining political issue.
Opposition has shaped recent county elections and mobilized a broad coalition of residents concerned about noise, power demand, and water use.
What was long taken for granted as an economic boon, with little public debate over its tradeoffs, has become the subject of open contention, fueling broader skepticism toward the scale and speed of the region’s technology-driven growth.
Bulldozers outside a data center being built in New Carlisle, Indiana, March 2026 Jim Vondruska / Reuters
These anxieties surface differently depending on who is speaking.
At the national level, policymakers speak of “AI leadership” and “innovation ecosystems.”
At the local level, those abstractions often translate into fears of job displacement, diminished control over essential resources, and rising energy prices.
Opposition to data centers is rarely opposition to computing itself.
It is opposition to the perceived asymmetry between communities asked to host infrastructure and companies whose decision-making seems elite-driven and opaque.
In one local online forum, critics of a project openly acknowledged using AI tools to draft their posts—a small but telling illustration of dependence mixed with distrust.
These tensions will intensify as AI’s energy demands grow.
By some estimates, data centers could consume roughly eight percent of U.S. electricity by 2030.
Yet no federal directive—whether an executive order, a task force, or an incentive program—can compel a local zoning board to approve a project.
This decentralized approach is a deliberate feature of American federalism , not an accident, but it can create friction when national priorities, such as expanding AI infrastructure, conflict with local preferences over land use.
That reality makes these disputes more than parochial conflicts.
In 2025, the White House issued an executive order emphasizing AI leadership as essential to U.S. economic strength and national security and directing the creation of a coordinated framework to sustain the United States’ technological edge.
The resulting AI action plan organizes federal strategy around three pillars: accelerating innovation, building AI infrastructure, and leading in international diplomacy and engagement aimed at forging technological alliances that can counter Chinese ascendancy and extend U.S.
AI dominance.
But these ambitions collide with institutional limits.
Federal policy can accelerate research, investment, and standards setting, but it cannot override local land-use authority.
BUILDING WITH BUY-IN
A more durable approach would rest on several elements.
The first is standardized transparency.
Before granting permits, local authorities should require companies proposing large-scale data centers to publish verifiable information on water use, power demand, emissions, and expected tax contributions.
Independent analysts should review those figures, rather than letting them rest on corporate marketing materials alone.
That would be a reasonable standard, more substantive than the minimal oversight common in the past but far less onerous than the layered environmental and permitting reviews that can make building anything prohibitively difficult in some parts of the United States today.
Transparency does not eliminate disagreement, but independently reviewed company disclosures on a data center’s core effects can reduce the scope of pure speculation and ground public debate in verifiable data rather than worst-case assumptions or promotional claims.
Another element is a clear mechanism for local benefit.
Even a small portion of a multibillion-dollar capital investment can generate tangible, long-term value for communities.
Examples include funding technical training programs at nearby colleges, expanding local broadband, supporting energy-efficiency upgrades, and investing in infrastructure that reuses waste heat from data centers.
Such measures do more than offset costs; they integrate digital infrastructure into the economic and social life of the communities that host it.
That integration is critical to legitimacy.
AI developers cannot ask residents to bear the burdens of technological change without sharing in its gains.
Finally, the firms behind large infrastructure projects should establish standing liaison bodies composed of residents, local officials, and company representatives.
These bodies would monitor compliance, address grievances, and adjust commitments as conditions change.
Unlike one-time hearings or negotiated concessions, ongoing forums institutionalize dialogue and reduce the likelihood that minor disputes escalate into broader opposition.
None of these measures require new federal authority.
They require coordination, shared standards, and credible commitment from firms and local governments alike.
The payoff is predictability.
Projects that begin with transparency, tangible benefit, and institutionalized engagement are less likely to encounter protracted resistance.
More broadly, such frameworks let the country pursue its technological ambitions without undermining the local legitimacy on which their success depends.
THE LEGITIMACY ADVANTAGE
Tech firms often treat community engagement as a compliance exercise or a public relations risk to be managed.
In practice, however, firms that approach it as an afterthought frequently encounter costly delays, political backlash, or outright rejection.
By contrast, a small number of companies have begun to recognize local legitimacy as a competitive advantage—one that can reduce friction, stabilize operations, and improve long-term project viability.
Several recent cases illustrate this distinction.
Last year, Google announced a major expansion of its existing data center campus in Midlothian, Texas, as part of a broader $40 billion statewide investment in cloud and AI infrastructure.
The company paired the project with a new solar and battery storage development facility, education grants to local schools and colleges, and a $30 million fund that Google and community partners will use to support workforce training programs, energy-efficiency initiatives, and projects aimed at improving energy affordability for Texas communities.
In Iowa, Microsoft partnered with nearby community colleges to train data center technicians, creating a modest but durable local talent pipeline that links regional workforce development to the firm’s operational needs.
In Umatilla, Oregon, the effects have been more pronounced.
After years hosting Amazon data centers, the local school district has seen tangible fiscal benefits from taxes and community service fees that the company pays, along with direct sponsorships and investments in robotics, STEM labs, and talent pipeline scholarships.
Over time, residents came to associate the presence of digital infrastructure less with extraction and more with reinvestment.
These cases remain exceptions rather than the norm.
Most data center projects still offer little local employment and rely on incentives that can deepen skepticism rather than build trust.
But as resistance to large-scale digital infrastructure intensifies, models based on reciprocity and sustained partnership are likely to prove decisive.
Projects that integrate community benefit into their core design stand the best chance of securing durable consent; those that do not will face mounting political and regulatory headwinds.
This is also where third-party intermediaries can play a constructive role.
Independent actors—universities, regional planning bodies, and nonprofit technical organizations—can help translate between the technical language of infrastructure development and the civic concerns of host communities.
Because no two projects or communities are identical, legitimacy cannot be fully standardized.
But evidence-driven mediation can identify areas of mutual gain, reducing conflict and allowing strategically important infrastructure to move forward.
The physical foundations for large-scale digital infrastructure already exist in many parts of the United States.
Former industrial sites, legacy transmission lines, and surplus power generation could anchor the next phase of the digital economy.
What is missing is not engineering capability but the institutional capacity to secure local consent for nationally significant projects.
The United States has confronted comparable challenges before.
During the New Deal, the Tennessee Valley Authority combined federal investment with local participation, extending electricity and economic opportunity to rural communities while embedding projects in regional institutions.
The lesson was not simply that large projects could be built but that shared benefit and accountability could make them politically durable.
A similar balance between national purpose and local stewardship will be necessary as digital infrastructure becomes a pillar of economic and strategic power.
What is required now is not another executive order or rhetorical commitment to technological leadership but a governance model that treats local legitimacy as a prerequisite rather than an obstacle.
Without it, the United States risks ceding a consequential advantage to more centralized systems.
China’s ability to plan and execute AI infrastructure at national scale may prove decisive precisely because it reduces uncertainty and delay.
In a competition increasingly defined by computing power, energy, and physical capacity, the ability to build may matter as much as the ability to invent.
In that sense, the debates playing out in New York, Arizona, Virginia, and across the country are not just about water use, power demand, or employment projections.
They are a test of whether the United States can still coordinate complex, nationally significant projects within a decentralized political system marked by declining trust.
These contests will shape not only where data centers are built but also how—and whether—the country can translate technological ambition into durable power.
Science & Technology
Artificial Intelligence
Research & Development
Baghdad Goes Its Own Way
The Ukraine Lesson Taiwan Keeps Missing
It’s Not the Drones—It’s Everything Around Them
David Petraeus and Clara Kaluderovic
The Strong Do What They Can—and Suffer What They Must
What Thucydides Really Thought About Power
Trump Is Remaking Latin America
The Short-Term Results and the Long-Term Risks of the “Donroe Doctrine”
How to Win the Defense Innovation Contest
America and Its Allies Must Pool Their Efforts
Colin H.
Kahl and Tobias Vestner
The Military and the Republic
What America’s Armed Forces Can—and Cannot—Do for Democracy
Charles Q.
Brown, Jr. , Peter D.
Feaver , and Andrew Kragie
How Europe Can Get Putin’s Attention
The Continent Must Overcome Its Russia Predicament
Speakers
Analysis
Hover over highlighted words in the article to view the associated bias or fallacy analysis.