You should protest against the datacenters

tldr: people protesting data centers are right, but not for the reasons they think they should be. it’s not about taxes or water or noise. it’s about who gets to control the infrastructure everyone needs to exist – and whether that control stays closed in three companies or opens to whoever runs the hardware. liberty first, economy second.

Build up, expect questions

There are currently people in US who are protesting against the data centers. And while I am never a luddite or do I ever protest against such an infrastructure, I am currently inclined to agree with them.

And many more…

The protesters are right. But not for the reasons they think they’re right.
I think they are misled to go against the infrastructure without thinking how exactly its being used.

The bad deal

Let’s set aside the moral argument for a minute and just look at the ledger. What does a community actually get when a data center moves in?

Before we compare factories, we should agree on the basic definition. Every factory does some environmental damage and creates some local effects. That does not automatically make it bad. If the factory produces more net good than environmental damage, then it may be worth having. Everyone should do that math locally. The question is not whether production has a cost. The question is whether the good produced is large enough, broad enough, and genuinely available enough to justify the cost.

During construction: jobs, local spending, some tax revenue. A couple years of activity. Then the building is finished, and something shifts. The data center sits there – a sealed box of servers – generating enormous revenue for its owner. It employs maybe a few dozen people long-term, mostly security and maintenance. It draws massive amounts of power from the grid everyone pays into. It draws water from shared aquifers. The land it sits on was probably rezoned or tax-subsidized to lure the company in.

Not all tech is equal. A factory is not automatically a public good just because it produces something technological, and a data center is not automatically the same kind of thing as another data center. You have to ask who the technology can actually serve.

A factory making chips for consumer computers can eventually make computers, RAM, and graphics cards cheaper for ordinary people. The product can spread through the economy and become useful to anyone who can buy it. But a factory making only enterprise-grade servers serves a much narrower group.
You are unlikely to see those servers sold for people to host at home. The hardware may be physically close to your community while its benefits remain available only to the large companies that can afford and control it.

Now consider solar panels. A solar-panel factory is a neutral good regardless of where it is installed.
The panels serve their purpose whether they end up on your roof or in a utility-scale power plant. The benefit is not tied to the factory’s owner or trapped in the place where the panels were manufactured. The product is useful wherever it goes.
Monopoly can arise as concentration of goods rather than the value of each good in itself.

The same distinction applies to data centers. A data center serving open models can provide a tool that many different operators, businesses, universities, and communities can use. A data center serving closed models can take the same land, electricity, water, and shared human output and turn them into a service controlled by one company. The label is the same – factory, server facility, data center – but the technology is not equal in terms of whom it can serve.

Open models can work in the same way. They are produced in one place but can be hosted and used by many operators, businesses, universities, and communities. Unlike a physical product, an open model has no physical supply limit. One copy can serve the world.

But these data centers are increasingly used to train and serve closed models owned by a handful of companies. The output does not return as a shared good. Public inputs are converted into a closed asset, and the public is asked to rent access to it.

Enterprise-grade hardware does not benefit ordinary people merely by existing near them. The physical presence of the facility does not give the community meaningful control or access.

This is the bad deal for everyone. A community gives up land, power, water, and grid capacity. In return it gets, at best, a few dozen maintenance jobs and the privilege of paying a monthly subscription to access the thing built using resources extracted from that community. You host the factory, power the factory, cool the factory, and then get charged rent to use the product the factory made from your inputs.

Okay, yeah, consider the case.
The data centers almost everywhere bad – raise electricity prices water and all those arguments.
They raise it everywhere, so we can average the negative here and keep it as a constant.
You suffer for something anyway. But in case when those DCs are used for training something open – you get the benefit, everyone else gets the benefits, all economies get the benefit from abundance and diversity.

Suffering for something is different when its for private vs common good.

Enclosure of the Commons

There is even a name for this pattern, and it’s five hundred years old: enclosure.
(guess who is discovering this stuff already has a name? Me, armchair philosopher)

A data center draws on genuinely public resources. The land may be rezoned or subsidized. The electricity comes from a grid everyone pays into. The water comes from shared aquifers and reservoirs. The training data comes from the internet, books, code, conversation, and the collective output of human civilization.

Then the output – the trained model – gets fenced off. The weights are closed, the API is metered, and access is rented back to the public whose resources and culture helped build it.
You do not get the model.
You get a chat window, a usage limit, and a monthly bill – along with a company deciding what the system is allowed to say.

The pattern is the same as the 18th-century enclosure acts. Swap hedgerows for API keys and grazing rights for inference credits. The mechanism has not changed. Only the surface has.

Who owns the box matters

Who owns the data center, and what is it working for?

An open facility can host models and sell compute or inference to businesses, universities, public institutions, and communities. It can compete on price, quality, privacy, or local service. That remains infrastructure others can use.

A facility owned by a model company and built to serve that company’s closed system is different. It turns public resources into a private edge and makes the public dependent on one provider. The question is not whether the facility draws power. The question is whether it leaves people with more access and control, or less.

I would protect a data center that uses its infrastructure to provide an open, broadly useful good.
I would oppose a data center whose purpose is to strengthen a closed monopoly and make the public dependent on it.

Scale is not the same as public good

The scalers will say that concentrating compute under one company makes it more efficient and brings lower prices. Maybe it can. But efficiency is not a promise, and there is no contract requiring a company with no competitors to pass the savings on.

Once one company controls infrastructure people need and there is nowhere else to go, it can dictate access, terms, prices, and the conditions under which everyone else exists. That is the old monopoly pattern with new machinery: extract value from the surrounding system, concentrate ownership, remove the possibility of exit, and call dependence efficiency.

The important distinction is vertical integration. A large provider that sells open compute or inference to everyone can still be infrastructure others use. A company that owns the data center, the model, and the only route to inference is using infrastructure to make its own closed system unavoidable.

Essential infrastructure and the kill zone

The same logic applies to the internet. We call it a network, but for anyone whose business, communication, or public life depends on it, it is infrastructure. Compute is becoming this generation’s essential facility. If your business depends on AI inference, then whoever controls access to compute can decide whether your business survives.

Compute is becoming this generation’s essential facility. It has direct, measurable influence on whether your business can exist. If you’re building anything that depends on AI inference – and increasingly, that’s everything from customer support to drug discovery to legal research – then who controls access to compute controls whether your business survives.

And this is where it gets worse than the railroads.

The railroads could control price and access. They couldn’t watch what you were shipping, figure out which goods were most profitable, and then launch their own competing shipping company that undercut you using the data they collected from your shipments.

That is what people miss when they worry about AI taking jobs.
The threat is not only that a machine replaces you at your desk – that is inherent to the tech itself.
It is that the company controlling the closed model and the infrastructure replaces your company – your reason for existing as a business at all.

If the models are open, you can use them to reform your operations, serve your customers better, and automate on your own terms. Adapt or grow on your terms.

If the infrastructure is closed, the Google playbook becomes the logical endpoint: replace the worker, then the business, then the sector, and sell access back to the customers who used to be yours.

The essential facilities doctrine exists because we have been here before. Railroads received public land grants, and courts felt justified compelling access because the infrastructure was built on public concession. Data centers receive public tax breaks, public land, public grid capacity, and public data. Yet the profits and control are collected by the few.

Liberty first, economy second

The real argument is about liberty.
Taxes are not the core issue. A company can pay every tax correctly and still use essential infrastructure to close off access, concentrate power, and make everyone else dependent on it.

A tax exemption could even be defensible if the facility were producing a genuinely universal public good.
The question is who benefits from the environmental and infrastructural costs, and whether the result leaves the public with more freedom or less.

The enclosure acts took centuries to reverse. We’re at the beginning of this one.
So, protest against the monopolies and closeness, not the datacenters themselves.