How to Build a Pro-AI Majority

Artificial intelligence is becoming a political problem before it has finished becoming an economic one.

For a technology whose champions call it the most consequential of the century, AI is strikingly unpopular. Roughly seven in ten Americans now oppose an AI data center in their area, with nearly half strongly opposed. That level of resistance has surged by more than 30 points in less than a year and now exceeds opposition to a nearby nuclear plant. Of respondents from 32 countries in a recent Ipsos survey, Americans were the most likely to expect AI to make their economy worse. The country building the future is, by this measure, the most anxious about it.

That is a problem far larger than bad public relations for a few large technology companies. A hostile public produces hostile politics: moratoria, punitive taxes, licensing schemes, permitting refusals, and local vetoes that can throttle the deployment of AI across the economy. It already has done so. New York’s legislature passed a first-in-the-nation one-year moratorium on permits for large data centers in June, Maine’s legislature passed one of its own before a gubernatorial veto, and dozens of local moratoria have been adopted nationwide, with community opposition blocking or delaying tens of billions of dollars in planned projects.

We think this is a mistake. AI is one of the most valuable technologies ever to emerge. There’s a real risk that America will squander its advantage in this field by failing to spread what it builds. The economic gains from AI depend on whether that diffusion occurs. Public opinion will help determine whether it does.

For those of us who believe that AI is a vital part of America’s and the world’s future, the political task is to build a durable constituency for AI deployment. Against the AI backlash, we need a new, pro-AI majority, and a new politics to fuel it.

That constituency cannot be assembled along the usual partisan lines, because AI does not divide the country the way most issues do.

Some on the Right see AI as an engine of innovation, growth, and opportunity. Another part of the Right has become one of the most determined sources of AI opposition, driven by concerns about children’s safety, social disorder, and the power of big tech.

The Left is divided, too. Some are hostile on grounds of jobs, inequality, climate change, and corporate power. Others remain broadly optimistic about technology, progress, and abundance.

A pro-AI majority, therefore, will have to be built across partisan lines from groups that agree about AI while disagreeing about almost everything else.

Recent work from the polling firm Echelon Insights makes that coalition map unusually clear. Combining views across 13 industries, it finds that Americans cluster into seven distinct “tribes,” each with a coherent worldview about technology, energy, and finance.

Two of those tribes form a natural pro-AI base. The Aggressive Deployers—young, male-skewing voters enthusiastic about nearly every emerging technology—register +84 net favorability toward AI and account for 13 percent of the electorate. The Center-Right Abundance tribe, a largely Republican constituency broadly optimistic about innovation, records +56 net favorability and makes up another 17 percent.

Together, they account for 30 percent of voters. That’s a meaningful base, but not a governing majority.

Add to them the Center-Left Abundance tribe. College-educated, suburban, and broadly pro-technology, it represents 12 percent of the electorate. Its members are cool on AI specifically, but only slightly, at -6 net favorability.

For a pro-AI coalition, these voters are reachable with the right effort. Win them over, and AI supporters reach roughly 42 percent of the electorate.

Beyond the Center-Left Abundance voters sits the largest single tribe: Passive Youth. They are 18 percent of voters, younger and female-skewing, skeptical of AI but not deeply opposed to it. These voters make up the swing constituency. Bring along a meaningful share of them, and a pro-AI coalition can move from sizable plurality to definitive majority.

The Echelon data show that building a pro-AI majority does not require converting the technology’s most committed opponents. Rather, it requires constituencies already inclined toward technological optimism but still unsure about AI itself. A compelling message for these on-the-fence audiences should begin with what the technology actually does.

The strongest argument for AI is also the most accurate: it disperses capability. AI takes functions that were once expensive, specialized, and cordoned off and places them within reach of ordinary people and small firms at a fraction of the old cost.

The research evidence supports the idea that AI’s gains are biggest for those starting furthest behind. In the largest field study to date, customer-support agents using a generative-AI assistant resolved 14 percent more issues per hour. But the effect was 34 percent for the least experienced and least skilled workers and close to zero for the top performers.

That pattern recurs across professions. A randomized study of professional writing found that AI cut the time required by 40 percent while raising quality and compressing the gap between the strongest and weakest performers, with participants who scored lowest on an initial unassisted task seeing the largest gains. GitHub developers completed a coding task 56 percent faster with an AI assistant. In a field experiment with consultants, AI raised output quality on suitable tasks by roughly a third, with the largest gains going to the lowest-scoring consultants.

This narrowing of the skills gap will have reverberating economic effects. A working software application that once required a team of engineers and tens of thousands of dollars to commission can now be assembled in days by one person with no formal training. The same compression is reaching design, legal analysis, accounting, tutoring, and the back office of nearly every small business.

Venture capitalist Michael Bloch has sketched the optimistic version of this future. He imagines a professional displaced by AI using these tools to build (in just a few weeks) the niche product her former employer was too large to bother with. Soon, she’s earning more than she did in her old job.

AI, in other words, is the most powerful solvent yet invented for the barriers to starting things. It lets the small compete with the large and the founder with an idea compete with the incumbent who has everything except the idea.

AI’s promise for entrepreneurship speaks to every group the coalition needs. To the Center-Right Abundance voter, it is a story about markets and the dignity of initiative. To the Center-Left Abundance voter, it is a story about access and breaking the grip of incumbents. To the Passive Youth voter, it is a story about agency.

What policy program emerges from our account of what AI does? We like to think of it as “diffusionism.”

The aim is simple: place AI’s capabilities in as many hands as possible. The goal is not just economic growth, but the broad distribution of capability itself.

Most of the current AI debates begin by asking who will capture the returns from AI and how those returns should be redistributed. But a society cannot redistribute prosperity it has failed to generate, and it cannot democratize opportunity if a powerful technology remains concentrated in a few large firms and institutions.

The first task is therefore diffusion—getting the capability into as many hands as possible. That is what makes broad participation in the gains possible in the first place.

Diffusionism offers a better frame than either boosterism or backlash. It does not deny the real anxieties around AI: the power of Big Tech, the pressure on workers, the strain on energy systems, the risks to children, and the fear that ordinary people will be acted upon rather than empowered. Instead, it answers those anxieties by insisting that the technology’s benefits must be broadly accessible rather than narrowly captured.

The politics mirrors the technology. AI disperses capability, lowers barriers, breaks through gatekeepers, and rewards initiative over credential. A diffusionist politics tries to accelerate those same effects. It pushes power outward, widens participation, and rejects a politics built around managing scarcity.

What does diffusionism look like in practice? First, it means confronting the physical constraints on AI’s expansion: the energy and transmission needed to power computation, the land-use and permitting rules that determine whether anything gets built, and the water and grid demands that increasingly decide a project’s fate.

The hardest immediate test is the data center. It is the first major diffusionist policy fight. The case has to be made to communities in terms of tangible local benefit rather than asserted over their objections.

The coalition for such an effort crosses partisan lines. That is unusual. But it can be built.

What is not available is unlimited time. The longer the story of concentration goes unanswered—the story of AI as something vast, corporate, and imposed on ordinary people—the harder the persuadable groups will be to reach. Hostile politics will harden into moratoria and refusals that throttle diffusion just as the country most needs it.

The window for building a pro-AI majority remains open. But not indefinitely.