Data Centers Are Necessary. That Doesn’t Mean Warren County Needs One

Two recent Royal Examiner pieces approached the data center debate from different directions. Kenneth Roko’s “The Data Center Debate Should Include Data Security” raised concerns about cybersecurity, privacy, and artificial intelligence. “We All Use Data Centers. So Where Should They Go?” made the equally important point that virtually all of us depend on data centers whether we think about them or not.

I agree with parts of both arguments, but I think some important distinctions are getting lost.

I’ve followed computing since the early 1980s. Even as a kid, I programmed, read computer magazines, went to the library, and tried to keep up with a field that seemed to change constantly.

By 1999, I was working professionally at digitalNATION, where we mostly called what we did dedicated server hosting. We had racks of servers running websites, email, and other Internet services, along with dial-up Internet access.

I’ve experimented with many forms of artificial intelligence over the years, built an artificial neural network myself about a decade ago, and today run modern AI models on my own hardware. I also have a virtual private server hosted in Richmond, back it up daily to another off-site service, and have a Bitcoin miner hosted at a location that helps stabilize the energy grid.

Those are all data center uses, but they’re very different.

I’m fascinated by this technology. I was also opposed to the most recent Warren County data center proposal because I don’t believe we’ve sufficiently answered the technological, infrastructure, and financial questions that could affect this community for decades.

Data Centers Aren’t One Giant AI Network

Some truth exists in the idea that AI computers are networked together. Modern AI systems can connect dozens or hundreds of powerful processors using extremely fast connections. A model can even be divided among several computers so their combined memory and computing power can handle something too large for one machine.

I’ve experimented with distributed AI at home on a much smaller scale, having multiple computers process different pieces of a document and return their results.

But that’s very different from assuming data centers scattered across America automatically combine into one giant AI computer or send everyone’s information to a central government repository.

A company or government certainly could intentionally build such a system. The relevant question is whether a particular service actually does so.

Data Security Matters, but It Isn’t New

I agree with Roko that cybersecurity belongs in this conversation. But protecting information in data centers isn’t a problem that began with AI.

I’ve worked on government systems where firewalls separated servers, application logic, and databases, with intrusion detection watching for attacks. I’ve managed databases where we performed daily backups and physically shipped tapes to two different locations so a disaster couldn’t destroy every copy.

Today, backups and server drives can be encrypted, and information traveling between your browser and a website is normally encrypted.

None of that makes a system impossible to hack. Security has always required layers.

The better privacy questions are: Who has access to my information? What are they allowed to do with it? How long do they keep it? What do they collect?

And perhaps more importantly: What can they learn about me from information I didn’t realize I was providing?

Can seemingly unrelated information establish patterns about my activities or whereabouts? Can it reveal my interests or habits? Can information collected for one purpose be combined to create a profile and then sold, shared, or used for another purpose?

Those are legitimate privacy concerns. But they concern the applications collecting information and organizations controlling it, not simply the building containing the computers.

Government hasn’t ignored cybersecurity either. NIST has cybersecurity and AI risk-management frameworks and recently published a draft specifically addressing AI data-center security.

Roko raises an important distinction: government guidance isn’t necessarily mandatory. If Warren County believes particular cybersecurity protections are necessary, we should ask which can actually be required of the project.

Blockchain Isn’t a Magic Shield

I would also be careful about treating blockchain as the answer.

Blockchain has become a buzzword covering cryptography, private keys, distributed networks, smart contracts, and many other technologies. There are legitimate uses and plenty of scams.

A blockchain can make records difficult to alter, but even that depends upon its consensus system. The network must agree about which version represents the truth. Depending on the blockchain, that may depend upon who controls computing power, cryptocurrency stake, or another mechanism.

Blockchain also doesn’t prevent someone from stealing a private key, exploiting software, compromising a computer, or tricking someone into authorizing something.

Blockchain doesn’t eliminate trust. It changes where trust is placed.

Knowing is half the battle.

AI Is Changing Remarkably Fast

My bigger concern for Warren County is financial risk.

Today’s largest AI systems require enormous amounts of computing hardware, electricity, and cooling. But those requirements are changing remarkably quickly.

Some AI already runs directly on phones. Increasingly capable models can run on gaming computers. Developers are reducing memory requirements and dividing work among GPU memory, ordinary system RAM and CPUs.

We’ve seen sudden changes before. In 2012, AlexNet used neural networks and GPUs to dramatically outperform competing approaches in the ImageNet image-recognition competition. Neural networks rapidly came to dominate the field.

In 2016, AlphaGo defeated Lee Sedol, one of the world’s strongest Go players, again showing how quickly assumptions about AI capabilities can change.

Maybe today’s enormous GPU clusters remain necessary for decades. Maybe AI usage grows so rapidly that efficiency improvements are overwhelmed and we need even more data centers.

Or maybe today’s systems eventually look like a brute-force stage on the path to something much more efficient.

I don’t know.

That’s precisely the point.

Technology companies can gamble billions on which future occurs. Warren County can’t.

Electricity Is Where the Risk Becomes Real

Virginia’s JLARC study found that data centers are driving an enormous projected increase in electricity demand.

Its analysis concluded that satisfying projected demand would require substantial new generation and transmission infrastructure. Even meeting half of unconstrained demand would be difficult.

The modeled solutions involve major additions of solar, wind, natural gas, battery storage, and transmission. JLARC found that both modeled scenarios would also rely partly on what it called “as yet unproven nuclear technologies.”

Today’s demand is real. The Department of Energy estimated that data centers consumed about 4.4% of U.S. electricity in 2023 and could consume between 6.7% and 12% by 2028.

The question isn’t whether demand exists. It’s who assumes the risk when long-lived infrastructure is built around forecasts of future demand.

JLARC estimated that residential customers could eventually pay an additional $14 to $37 per month by 2040 for generation and transmission costs associated with data-center demand.

That’s where this stops being an abstract technology discussion.

So Where Should Data Centers Go?

The other Royal Examiner commentary asks a fair question: We all use data centers, so where should they go?

My first instinct was to say they should expand where the infrastructure already exists. But Northern Virginia demonstrates why it’s more complicated: it has extraordinary fiber and data-center infrastructure while its electrical infrastructure is increasingly strained.

A better principle is to put major data-center development where independent analysis shows that the cost and community impact of providing electricity, transmission, fiber, water, and other infrastructure are acceptable—and where the parties creating the demand bear an appropriate share of those costs and risks.

We all use airports, power plants, warehouses and water-treatment plants. That doesn’t mean every community needs one.

Promises and Guarantees Are Different

This was my biggest concern with the most recent Warren County proposal.

What revenue is actually guaranteed? What infrastructure must be built, and who pays? What happens if demand doesn’t materialize, the facility changes owners, or the equipment becomes obsolete sooner than expected?

JLARC found that much of the economic activity associated with data centers occurs during construction. Industry representatives described a typical 250,000-square-foot facility as employing about 50 full-time workers, roughly half contractors, compared with around 1,500 during peak construction.

That doesn’t mean data centers provide no benefit. It means temporary construction activity, projected economic impact, and guaranteed long-term benefits aren’t the same thing.

Any proposed payment-in-lieu-of-taxes, or PILOT, agreement deserves the same scrutiny. Does the payment increase with inflation and rising county costs, or does an amount negotiated today remain fixed decades from now? When does the agreement expire? Is it periodically renegotiated? What happens if ownership changes? How does the county’s long-term revenue under the agreement compare with ordinary taxation?

A payment that sounds impressive today could look very different twenty or thirty years from now.

The same principle applies to local impacts. If a developer says neighbors won’t hear the facility, put a measurable noise limit in the conditions. If buffering is promised, specify it. If economic benefits are promised, distinguish estimates from legally enforceable commitments.

The more confident everyone is in those promises, the less objection there should be to putting them in writing.

Why I’m Cautious

I’ve watched computing reinvent itself since I was a kid in the early 1980s when my family purchased an Atari 400. I’ve watched technologies that seemed revolutionary become commonplace, dominant approaches get replaced, and things expected to take years arrive much sooner.

Today I run AI models on my own hardware and distribute AI workloads across computers. I’m fascinated by AI and use it every day.

That’s not despite my caution about making decades-long infrastructure commitments around today’s AI requirements. It’s one reason for it.

Technology companies and investors are betting on the future value of AI infrastructure. They’re entitled to make that bet with their money.

My concern is making sure Warren County and its residents aren’t making the bet on their behalf.

We can’t predict what AI will look like ten or twenty years from now.

We can make sure that if those predictions turn out to be wrong, Warren County isn’t left holding risks that should belong to the companies making the investment.

Lewis E. Moten III
Front Royal, VA

About Letters to the Editor

Letters to the Editor reflect the views and opinions of the individual authors and do not necessarily represent the views of the Royal Examiner, its staff, affiliates, or advertisers.

The Royal Examiner welcomes a wide range of viewpoints and encourages thoughtful, respectful discussion on matters of local and public interest. Publishing a letter does not mean the Royal Examiner endorses the opinions, claims, or conclusions expressed by the writer.

Statements and claims made in letters are the responsibility of the authors. The Royal Examiner may not independently verify every factual assertion contained in a submitted letter, and readers are encouraged to consider the information and opinions presented with their own judgment.

Not all letters submitted will be published. The Royal Examiner reserves the right to edit letters for clarity, length, grammar, and compliance with editorial standards. Letters containing personal attacks, offensive language, unsupported accusations, or material that may violate the law may be rejected.

By submitting a letter, the author grants the Royal Examiner permission to publish, edit, reproduce, and distribute the content in print, online, and other formats.

We value reader participation and welcome differing perspectives. Letters to the Editor are an important part of community conversation, and we encourage readers to share their views on issues that matter to Front Royal, Warren County, and the surrounding region.