An Indianapolis city councilor named Ron Gibson woke before 1 a.m. on a Monday in April to the sound of gunfire. Thirteen rounds struck his home. A note on the doorstep read: "No Data Centers." His eight-year-old son had been playing with Legos at the dining table the day before, steps from where the bullets hit.

Gibson had voiced support for a proposed data center in his district's Martindale-Brightwood neighborhood. That support nearly cost his family their lives.

This is not an isolated case. A 20-year-old man recently attempted to firebomb OpenAI CEO Sam Altman's home in San Francisco. Days later, two more people allegedly shot at Altman's residence, leading to arrests. What started as a planning dispute has turned violent.

Alphabet, Amazon, Meta, and Microsoft are expected to spend more than $650 billion in 2026 alone to expand AI data center capacity, according to Bloomberg. Six hundred and fifty billion dollars. In a single year.

That figure tracks closely with the $690 billion in AI capital expenditure we documented in our State of the Stack analysis earlier this year. The money is real. The ambition is real. The question is whether any of it can actually be built.

According to analysts at Sightline Climate, between 30 and 50 percent of AI data centers planned for deployment in the US this year will be delayed or canceled. Of 140 planned projects representing roughly 16 gigawatts of capacity, only about five gigawatts are under construction.

The first bottleneck is physical. Shortages of critical electrical equipment, transformers, switchgear, batteries, are strangling build timelines. Lead times for high-power transformers have ballooned from around two years before 2020 to up to five years today. Many of these components still come from China, which creates an uncomfortable dependency at a time of escalating trade tensions.

The second bottleneck is human. Communities are saying no. And they are winning.

According to tracking by Project Censored, 188 organized opposition groups now operate across the country. In Q2 alone, 53 active groups across 17 states targeted 30 data center projects. Two thirds of those protested projects were blocked or delayed.

The opposition's most powerful weapon is not environmental concern or noise complaints. It is the utility bill.

Data from the Environmental and Energy Study Institute shows that in Virginia, home to the world's largest data center market, electricity prices have risen by up to 267 percent over five years. Nationally, Brookings found that electricity costs have climbed roughly 42 percent since 2019, well above headline inflation. The research concluded that the cost of grid upgrades needed to power AI infrastructure is being passed on to residential customers.

A Yale Climate Connections analysis made the disparity explicit. Between 2020 and 2024, residential electricity prices rose 25 percent. Commercial users, including data centers, saw far smaller increases. Industrial users actually paid lower average prices than two years prior. Residents subsidize the infrastructure powering the AI boom while the companies driving demand get discounted rates.

The Dallas Federal Reserve has modeled the inflationary impact. Under plausible scenarios, data center electricity demand could add between 0.04 and 0.13 percentage points to annual PCE inflation by 2030. If all proposed projects were connected and fully utilized, the figure rises above a full percentage point.

Utility disconnections are rising. Families are choosing between electricity and food. The companies building these data centers are among the wealthiest in human history.

What makes this movement unusual is its speed and geographic spread. Opposition does not cluster in one region or one political camp. It is everywhere.

In Festus, Missouri, residents filed a lawsuit against their city after a 360-acre data center project was approved, alleging the council held private meetings and gave the public inadequate time to review the proposal. In nearby St. Charles, residents are pushing to ban data centers permanently. In Foristell, a proposed land annexation was amended to preserve agricultural zoning after residents suspected the land would be used for data center development.

In Imperial County, California, a grassroots group called Not In My Back Yard Imperial gathered more than 3,400 petition signatures to fight a data center proposed adjacent to hundreds of homes and within a mile of schools. The county planning commission voted against the project's lot merger in December, though the decision was subsequently appealed.

In San Marcos, Texas, a proposed $1.5 billion hyperscale campus was halted after more than 100 residents spoke against it and only a handful voiced support. The projected power demand would have reached 2.5 times the city's peak electrical load. 2.5 times.

In Tucson, Arizona, residents packed council chambers to oppose "Project Blue," a facility that would have consumed millions of gallons of drinking water from the desert for cooling. The council voted unanimously to kill it. In Wisconsin, voters used a ballot measure to block a proposed AI facility. In Independence, Missouri, voters ousted councilors who had supported a local data center.

Communities are learning from each other, sharing strategies across state lines, building what amounts to a decentralized national movement against concentrated corporate power.

This backlash is scrambling traditional political lines.

According to a Quinnipiac University poll, 65 percent of Americans oppose AI data centers in their communities. A Pew Research Center poll found that more Americans say data centers have a negative effect on the environment, home energy costs, and quality of life than a positive one. Only 26 percent of Americans hold a favorable view of AI overall.

The electoral impact is already measurable. Democrats Mikie Sherrill and Abigail Spanberger won gubernatorial races in New Jersey and Virginia in 2025 partly on promises to bring utility bills under control, according to CNBC. In Georgia, two incumbent public utility commissioners lost their seats after residential electricity prices climbed 41 percent in four years. In Michigan's open Senate race, candidates from both parties are campaigning explicitly on data center policy.

The legislative response is accelerating. The AI Data Center Moratorium Act, introduced in late March 2026, would temporarily pause new AI data centers at the federal level. Maine's House passed a moratorium 82 to 62, halting large data center construction until 2027. Oklahoma has proposed a moratorium until late 2029. New York Senator Liz Krueger is pushing for a statewide pause. Delaware is advancing legislation to charge data centers higher rates. Florida is creating new rate structures.

Senator Josh Hawley, a Republican from Missouri, and Senator Richard Blumenthal, a Democrat from Connecticut, have teamed up on a bill requiring data centers to find their own independent power sources, according to inewsource. Bipartisan cooperation on tech policy is rare. It signals how potent this issue has become.

As Lawfare put it: utility costs have become the new "cost of eggs" concern, deeply intertwined with cost-of-living anxieties that will dominate the 2026 midterms.

This story may look like a land-use fight. It is not. It is a supply-side constraint on the entire AI infrastructure buildout, and every forecast for agentic commerce, AI-powered fraud detection, and machine-to-machine payments depends on that buildout proceeding on schedule.

We have spent months documenting the demand side of the AI compute crunch. Anthropic outages. GPU prices up 50 percent. Enterprise customers queuing for inference capacity. The backlash covered in this article is the supply side. Together, they form a physical constraint on the agentic commerce buildout that no protocol, no standard, and no software update can solve.

The $1.5 trillion agentic commerce forecast that underpins valuations across the AI sector assumes the compute exists. It assumes models can be trained, fine-tuned, and deployed at scale. It assumes inference capacity grows in line with demand. Every one of those assumptions requires data centers to get built, powered, and connected to the grid.

When two thirds of protested projects are blocked or delayed, those assumptions start to crack. The models powering agentic checkout cannot run without GPUs. The GPUs cannot run without power. The power cannot flow without transformers that take five years to deliver and data centers that communities refuse to host. The implications cascade.

For processors, networks, and platforms betting on AI-native commerce, this is not someone else's problem. If the compute does not get built, the models do not scale. If the models do not scale, the agentic commerce layer that Visa, Mastercard, Stripe, and every major processor is investing in stays a demo.

Big Tech is aware of the problem. Microsoft has built a "Community-First AI Infrastructure" framework for its data center projects. Amazon commissioned research claiming its data centers do not raise local electricity bills and, in some cases, contribute surplus payments that help modernize grids. Google and others are exploring self-supplied power through nuclear and renewable energy to sidestep grid constraints entirely.

These olive branches face deep skepticism. Communities have watched developers use non-disclosure agreements to keep details about water use, electricity demand, and siting plans hidden from the public until deals were nearly finalized. Harvard's Electricity Law Initiative has documented dozens of special contracts between utilities and hyperscalers that give billion-dollar companies cut-rate deals behind closed doors, then spread the infrastructure costs to everyone else.

The core tension will not resolve itself. The United States wants to lead the global AI race. That requires compute infrastructure on a massive scale. But the communities being asked to host that infrastructure are paying the price in higher bills, strained water supplies, industrial noise, and diminished quality of life. They receive little in return. Data centers create few permanent jobs. The economic development promises ring hollow when the construction crews leave and the facility operates with a skeleton staff.

Something has to give. Either the industry finds a way to genuinely share the benefits and bear the costs of its expansion, or the grassroots movement that has already blocked two thirds of protested projects will keep growing. The political incentives now favor opposition. Every election cycle, the case for saying no gets stronger.

Is your community facing a data center proposal? What would it take for Big Tech to earn your neighborhood's consent?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.