Data centers are the new public enemy number one. 

“They’re stealing our water!”

“They’re going to crash the power grid!”

And after communities across the country started pushing back, the AI data center boom is supposedly grinding to a halt.

There’s just one problem: none of these claims are true.

They’re three of the biggest myths surrounding the most important infrastructure buildout in decades.

So today, let’s separate fact from fiction and dive deeper into the biggest myths surrounding America’s data center boom.

Myth #1: Data Centers Are Draining Our Water Supplies

This is the most common myth I hear.

AI chips generate a tremendous amount of heat which requires massive cooling systems.

Some of these cooling systems use water.

But this is where the numbers can get misleading.

There are two different ways to think about a data center's water footprint: direct water use and indirect water use.

Direct water is the water the data center itself uses.

The biggest source is cooling. Some facilities use evaporative cooling systems that consume water to remove heat. Others use dry cooling systems that operate mostly as closed loops and consume very little water after they're initially filled.

Indirect water is water used somewhere else to support the data center.

The biggest example is manufacturing the chips inside it. Semiconductor factories use highly purified water to manufacture GPUs, CPUs, memory chips, and other components.

Research firm SemiAnalysis dug into the water myth by accounting for both direct and indirect water use.

They compared SpaceX's massive Colossus 2 data center in Memphis, Tennessee with a single In-and-Out restaurant location. 

Colossus 2 currently has about 400 megawatts of computing capacity – enough to power 330,000 average American homes for a year.

SemiAnalysis estimates Colossus 2 has a total water footprint of 346 million gallons per year.

This includes water used directly at the facility and indirect water associated with manufacturing its chips.

346 million gallons sounds terrifying.

Until you apply the same accounting to a hamburger.

An In-N-Out restaurant has direct water use too.

There's the obvious stuff: water used inside the restaurant for cleaning, cooking, sinks and other operations.

But that's not where most of a burger's water footprint comes from.

It's the indirect water.

Cattle need feed. Growing that feed requires water. Producing tomatoes and lettuce requires water. All of that water gets counted toward the burger's total footprint.

Using the same accounting methodology, SemiAnalysis estimates the burgers sold by an average In-N-Out restaurant have a water footprint of 147 million gallons per year.

Now compare the two:

Colossus 2 has the same total water footprint as roughly 2.5 In-N-Out restaurants.

Put in perspective, the idea that data centers are going to drain America’s water supply starts to look a lot less convincing.

Myth #1: Busted.

Myth #2: Data Centers Are Going to Collapse the Power Grid

This one has a little more truth behind it.

But the mainstream narrative is incomplete at best. 

AI data centers consume an incredible amount of electricity, and America's power grid wasn't built for dozens of gigawatt-scale facilities suddenly showing up and asking to plug in.

But there's another option:

Don't plug into the grid.

More data centers are moving toward what's called behind-the-meter power.

Instead of waiting years for the local utility to provide enough electricity, developers can build power generation directly at or near the data center.

Bloom Energy (BE), GE Vernova (GEV), and Solaris Energy Infrastructure (SEI) are leaders in this space. 

SemiAnalysis expects behind-the-meter power to supply more than half of new U.S. data-center capacity beginning in 2028. 

It has already identified 75 gigawatts of firm orders for behind-the-meter power equipment – enough to power 62 million American homes for a year.

Yet this part of the data center power story often gets ignored.

AI is absolutely creating a historic increase in electricity demand.

But that doesn't mean we're simply piling all of that demand onto today's grid and hoping for the best.

The AI boom is forcing America to build more power.

Most of it is behind the meter. 

Myth #2: Busted. 

Myth #3: Data Center Projects Are Getting Canceled Everywhere

This is another case where big numbers make great headlines.

More than 300 local governments have adopted data center moratoriums or bans. 

This certainly sounds like the data center boom is being stopped.

But counting moratoriums tells you almost nothing about how many data centers are actually being delayed.

Research found 20 gigawatts of planned data center capacity located within areas covered by local restrictions.

But only 1.5 gigawatts of that capacity is actually being delayed by the restrictions.

That’s 7.5%.

Why?

Because announcing a moratorium doesn't magically stop every data center in the area.

Many restrictions only apply to new permit applications. If a project already has its approvals, construction continues.

Other moratoriums will expire before projects need permits. Some don't cover the exact parcel where the data center is being built. And developers can sometimes redesign or relocate projects. 

That's why "300 communities have restricted data centers" and "300 data center projects have been canceled" are completely different statements.

Meanwhile, estimates suggest 38 gigawatts of new U.S. data center computing capacity will be delivered in 2027 – more than twice 2026's total.

Of that, 22 gigawatts are already under construction. 

And much of the remainder is already financed and undergoing site preparation.

Put simply: the data center boom isn't being canceled. It's getting bigger.

Myth #3: Busted.

You’ve Been Lied To About Data Centers

The myths above make great headlines.

But the numbers tell a very different story.

• The water usage has been exaggerated. 

• The power problem is being addressed with an enormous buildout of behind the meter systems. 

• And more than 300 local restrictions have done little to slow construction.

In the face of these myths, America’s AI infrastructure buildout is accelerating. 

I recommend you stay invested.