Nasdaq hits a record high, one truth: The hotter AI gets, the harder it is to build data centers

The real world, however, is beginning to say "no" to AI.

On Monday, Eastern Time, all three major US stock indexes rose, and the Nasdaq hit another all-time high, driven by leading tech stocks such as Nvidia and Microsoft. This marks the second consecutive session of strong gains for the Nasdaq.

The market logic is not complicated. US nonfarm payrolls data for September came in weaker than expected, and markets quickly dialed back expectations for further Federal Reserve rate hikes. Traders now see only a 24% chance of a rate hike at the Fed's October policy meeting, down from 70% just a week ago.

With pressure from interest rates easing and oil prices retreating, tech stocks continue to rise. Once again, Wall Street is betting big on AI.

But while capital markets are constantly driving up AI valuations, an opposite trend is emerging around the globe.

Trump is stressing that the US must win the AI race, and companies like Microsoft, Meta, Amazon, Google, and Nvidia are announcing bigger data centers, larger GPU clusters, and higher capital expenditures.

The problem is, more and more ordinary people are starting to ask a very practical question:

Why do these data centers have to be built next to my home?

This could be the most easily overlooked, yet most critical change in today’s AI investment boom.

The more successful AI becomes, the greater the demand for data centers; the more data centers, the heavier the consumption of electricity, land, water, and grid capacity; and as infrastructure scales up, local resistance from residents only intensifies.

In other words, AI’s biggest bottleneck might not be GPUs in the future, but whether there’s a place willing to let you build.

$42 billion in European investment is now on hold

For the past two years, the market has been concerned about whether there’s enough Nvidia GPUs, enough HBM, or enough electricity.

Now, a new problem has emerged.

Even if you have GPUs, servers, money, and power, you might still not get a permit.

Research shows that due to public opposition, tightened approvals, and regulatory restrictions, about $42 billion of European data center investments have been affected, with many projects delayed or even canceled.

Between January and April this year, more than 70 data center projects in Europe were denied or restricted—already more than the total for all of 2025.

What’s even more noteworthy is that the resistance is no longer just a few residents with signs in the neighborhood.

It’s spreading from local governments all the way to courts, regulators, and even parliaments.

This means that data centers no longer face simple “NIMBY” (Not In My Back Yard), but rather a gradually institutionalized opposition to infrastructure.

Europe has been desperately trying to catch up with America in AI, but now an embarrassing paradox emerges: Europe wants AI but is increasingly unwilling to bear the physical costs behind it.

Olivier Dalmunny, an associate professor at HEC Paris specializing in energy transition, even warned this could be the “last straw” to break Europe’s AI competitiveness.

The reason is very practical. Very few people want a huge data center suddenly popping up next to their homes.

That’s because the benefits of AI are distributed across society, but the costs of a data center are heavily concentrated locally.

AI creates value nationwide or even globally, while noise, power consumption, water use, land occupation, rising electricity prices, and grid pressure land on just a handful of communities.

Those who get the benefits are not the same as those who bear the costs.

So now, conflict is erupting.

Europe is stepping on the brakes for data centers

Ireland once paused data center projects because their power demand was too high.

After that, Scotland also suspended review of some new hyperscale data center plans.

Nordic countries were once the favorite region for data center operators—they have plenty of land, abundant renewables, and a cold climate to reduce cooling costs.

But even the Nordics are tightening up now.

With applications for electricity surging from large data centers, Denmark passed an emergency law putting data centers at a possible disadvantage when applying for grid capacity in the future.

Spain introduced new rules requiring data centers to use at least 80% renewable energy.

Some UK projects have stalled due to vehement local opposition.

On the surface, these policies are meant to regulate data centers.

But from an investment perspective, there’s really only one message: The barriers to building AI infrastructure keep rising.

In the past, the biggest competitive edge in building data centers might have been capital.

In future, it may come down to four things: land, electricity, grid access, and local government support.

And these things are not infinitely replicable.

An AI data center might become a “ghost warehouse”

So why is local resistance only getting stronger?

Because hyperscale AI data centers are no longer the same thing as legacy Internet data centers.

People used to imagine data centers as basic buildings filled with servers.

But in the age of AI, data centers are swiftly becoming industrial-grade infrastructure.

Hundreds of thousands of GPUs running 24/7 demand huge power supplies, cooling systems, backup generators, batteries, substations, and massive tracts of land.

Once built, these facilities occupy huge spaces, consume enormous electricity, but might not create that many permanent jobs.

This brings up a sharp question:

What do local residents actually get?

A data center that consumes enough electricity to power a city may drive up land demand, stress grid investments, and use water resources, but doesn’t necessarily create thousands of long-term jobs like a manufacturing plant would.

This is why some European academics call them huge “ghost warehouses.”

They look massive from the outside, filled with billions of dollars’ worth of GPUs, but don’t really need that many people working there long-term.

The more valuable the AI industry becomes, the more this contradiction stands out.

US Investing Network believes this will likely become one of the key factors in AI infrastructure repricing over the next few years.

Previously, everyone talked about a “GPU shortage.”

In the next phase, the real scarcity may not be GPUs, but places where GPUs can be legally, rapidly, and affordably deployed.

The truly scarce AI resources are shifting from chips to Land + Power + Interconnect—in other words, land, electricity, and grid access.

The same problem is emerging in South Korea

If you think this is just a European issue, think again. Similar tensions are appearing in South Korea.

With Samsung Electronics and SK Hynix, South Korea is one of the most critical links in the global AI semiconductor supply chain.

This June, the Korean government even ranked data centers alongside semiconductors and Physical AI as one of three priority investment areas. The central government clearly wants rapid expansion of AI infrastructure.

But on the local level, things look very different. In some regions, residents have been protesting data center construction for extended periods.

In July, South Korean officials announced plans requiring some proposed data centers to obtain the consent of a majority of residents within 200 meters, while introducing new project review and dispute mediation mechanisms.

In one Seoul neighborhood, residents even demanded the government revoke an already-issued building permit.

These residents protested outside the local government office for months, and as of mid-August, the demonstrations had lasted 172 days.

In Gwacheon City, southern Seoul, local legislators proposed new regulations to limit potential safety risks from round-the-clock data center operations, including backup battery fires.

This highlights a crucial issue. Although AI infrastructure is a national strategic priority, data centers ultimately require a physical city, a particular plot of land, and a specific community.

The nation wants AI, but local communities don’t necessarily want data centers.

This will become an increasingly significant contradiction during future global AI expansion.

Can a single community really block $10 billion??

A manager at Alvarez & Marsal once made a point investors should seriously consider:

A single community may have the power to halt a $10 billion data center plan. This really highlights a huge future risk for the AI industry.

Microsoft, Meta, Amazon, and Google can approve tens of billions in capital expenditures. Nvidia can keep rolling out more powerful GPUs.

TSMC can keep adding advanced fabrication capacity. SK Hynix and Micron can keep ramping up HBM production.

But when you get to the industry’s downstream, all those chips have to be plugged into servers, and servers need physical data centers.

The data center needs to connect to the grid. The grid needs regulatory approval. The project needs land. And local residents have to agree.

If even one link is blocked, hundreds of billions in capital expenditure up the chain can’t actually translate into computing power.

This is why US Investing Network has always argued AI’s most valuable focus is not just the chips.

Electricity, transformers, power distribution, storage, cooling, land, grid access, and data center operators that already have power resources may all carry rising strategic value in future.

Because AI doesn’t live in the “cloud.” What we call “the cloud” is ultimately made up of real buildings, physical cables, substations, and hundreds of thousands of power-hungry GPUs.

The most ironic scene is now emerging

Trump and America’s tech giants all emphasize that the US cannot lose the AI race. Europe is also emphasizing AI sovereignty.

South Korea has elevated AI infrastructure to the level of national strategy.

But when these countries actually start large-scale AI infrastructure construction, they discover the biggest resistance comes from their own residents.

So the entire AI era now faces a very strange picture.