America's artificial-intelligence infrastructure boom has run into a problem that no semiconductor shortage or cloud-computing bottleneck can easily solve.

People living where the giant facilities are being built are increasingly saying they do not want them.

New research from Data Center Watch shows that 45 U.S. data-center projects worth approximately $68 billion were blocked or delayed by local opposition between April and June 2026. The projects represented more than half of the new large-scale developments the research group began tracking during the second quarter.

The numbers reveal a growing tension at the heart of the AI boom.

Technology companies are racing to build enormous computing facilities because increasingly powerful AI systems require extraordinary amounts of processing power.

But communities have to live beside those facilities.

And many residents are questioning whether the economic benefits of AI infrastructure are worth the demands it can place on electricity grids, water supplies, land and local infrastructure.

This is no longer a handful of isolated neighborhood disputes.

Data Center Watch says there are now 843 active opposition groups across 49 states, with Hawaii the only state without a group identified in its current tracking. Around 30 state legislatures have also introduced or adopted measures dealing with data-center siting, electricity and water issues.

The AI buildout is becoming a local political issue.

Much of the public discussion around artificial intelligence focuses on software.

The newest model.

The latest chatbot.

The next AI agent.

But behind every major AI system is a much more physical reality.

Data centers need land.

They need electricity.

They need cooling.

They need fiber connections.

They need substations and transmission infrastructure.

And they often need large quantities of water, depending on the cooling technology being used.

The companies building these facilities argue that they are essential to maintaining America's technological competitiveness.

Amazon, Meta, Microsoft and Alphabet are among the companies investing enormous sums in data-center infrastructure. The companies say these facilities are necessary to develop increasingly sophisticated AI and compete globally, particularly with China.

For investors, the logic is compelling.

More AI adoption means more computing demand.

More computing demand means more data centers.

More data centers mean more infrastructure spending.

But local residents see the equation differently.

Opposition groups have raised concerns about electricity consumption, water use, noise, land use, environmental impacts and whether local communities receive enough economic benefits in return.

In some areas, residents worry that data centers could increase pressure on power systems and eventually contribute to higher utility costs.

That concern has become particularly relevant as AI companies sign enormous electricity contracts and utilities plan new generation capacity.

The facilities may create construction jobs and eventually generate substantial tax revenue.

But those benefits do not necessarily arrive equally or immediately.

A community may face years of construction disruption before receiving meaningful fiscal benefits.

And if a utility has to make expensive grid upgrades, residents may question who ultimately pays for them.

That is helping fuel the backlash.

The latest $68 billion figure actually represents an improvement from the beginning of the year.

Data Center Watch recorded 75 projects worth roughly $130 billion that were blocked or delayed during the first quarter.

That suggests the pace of disruption eased during the second quarter.

But the underlying opposition did not disappear.

The 45 projects affected between April and June still represented more than half of the large-scale developments tracked during that quarter.

In other words, a substantial portion of new projects is encountering resistance before construction can proceed smoothly.

That creates a new risk for the AI industry.

The limiting factor may no longer be whether companies can afford to build.

It may be whether they can obtain permission to build.

The spread of local opposition is increasingly translating into formal policy.

Some communities have imposed construction moratoriums.

Others are tightening planning requirements.

State legislatures are considering rules related to electricity consumption, water usage and siting.

That creates another layer of uncertainty for developers.

A company planning a multibillion-dollar data center may need to make long-term commitments years before the facility becomes operational.

If local rules can change during that period, project economics become less predictable.

That can potentially increase costs.

And higher costs eventually matter to the entire AI investment cycle.

Electricity may be the most difficult issue of all.

An advanced AI data center can consume enormous amounts of power.

As more facilities are proposed in the same region, utilities have to determine whether existing infrastructure can support the additional demand.

In some locations, the answer may be no.

That means new generation and transmission capacity has to be constructed.

This can take years.

AI companies, meanwhile, want computing capacity much faster.

That creates a timing mismatch.

Technology companies want rapid expansion.

Energy infrastructure moves slowly.

Local opposition can slow both even further.

Cooling advanced computing equipment requires heat to be removed continuously.

Different data centers use different cooling methods, and water use can vary substantially.

Still, water has become one of the most politically sensitive aspects of new facilities, particularly in regions already facing drought or competing demands from agriculture and households.

Residents increasingly want developers to provide detailed information about expected consumption.

That is changing the permitting environment.

A project that once might have been viewed primarily as an economic-development opportunity is now being evaluated as a long-term resource decision.

Data Center Watch said its Q2 report was the first time its research formally tracked opposition outside the United States.

It identified resistance across Europe, Australia, South Africa and other regions.

That suggests the underlying conflict is not uniquely American.

The same basic questions are appearing wherever AI infrastructure expands:

Who pays for the electricity?

Who supplies the water?

Who owns the land?

Who gets the jobs?

Who receives the tax revenue?

And what happens to the community if the promised economic benefits do not materialize?

Those questions become more difficult as the facilities grow.

The financial numbers involved are enormous.

A single hyperscale AI campus can require billions of dollars.

Delays therefore do not simply inconvenience developers.

They can disrupt construction schedules, financing arrangements and agreements between technology companies and infrastructure providers.

A six-month delay in a giant AI project can affect the deployment timetable for the computing systems that were supposed to operate there.

That can create ripple effects through chip suppliers, cloud companies, power providers and construction firms.

This is why data-center permitting is increasingly becoming an investor issue.

The debate over data centers is also changing the political conversation around artificial intelligence.

For years, AI policy focused on issues such as jobs, copyright, privacy and model safety.

Now infrastructure is becoming just as important.

Communities are asking whether the physical footprint of AI deserves the same scrutiny as the algorithms.

That means AI companies may increasingly have to build a public-relations strategy alongside their construction strategy.

They need local support.

They need regulatory certainty.

And they need to demonstrate tangible benefits.

One unusual consequence of the data-center bottleneck is renewed interest in existing crypto-mining sites.

Some Bitcoin mining operations already have access to power infrastructure and permitted industrial facilities.

That can make them attractive locations for AI developers looking for computing capacity without waiting years for new grid connections.

Yahoo Finance reporting noted that Bitcoin miners are becoming unlikely “landlords” in the AI infrastructure market because some already-permitted power sites can potentially be repurposed for AI computing.

That creates an unexpected connection between two industries that were once viewed as competitors for electricity.

Bitcoin mining could become part of the infrastructure solution for artificial intelligence.

The biggest lesson from the $68 billion in delayed projects is that AI is no longer simply a software story.

The next stage depends on land, power, water and permits.

Companies can build better models.

They can design faster chips.

They can raise billions in financing.

But none of those things matter if the physical infrastructure cannot be built.

For investors, that means the AI growth story now has a new variable.

Local acceptance.

The AI industry may have the money to build the future.

The question is whether the communities hosting that future are willing to let it happen.

And with hundreds of opposition groups now organized across nearly the entire country, that question is becoming harder to ignore.