For years, we have been sold the illusion that the "Cloud" is a weightless, ethereal place. We speak of our data as if it exists in a digital void, floating in a seamless expanse of connectivity. But the Cloud is not a cloud; it is a physical entity. It is a sprawling network of concrete monoliths, humming server racks, massive cooling arrays, and a staggering amount of copper wiring. It is made of steel, silicon, water, and an appetite for electricity that is currently pushing the global power grid to its breaking point.

Recently, the conversation around data centers has shifted from technical curiosity to a matter of urgent public policy. We see the headlines: Big Tech companies pivoting toward nuclear energy to sustain their AI ambitions, local governments fighting the construction of new data centers due to water scarcity, and power grids in Virginia and Ireland struggling to keep the lights on for residential neighborhoods because the "compute" demand is too high. The world is waking up to the fact that the digital economy has a massive, physical footprint.

However, most of the current drama focuses on the "AI Gold Rush." We talk about the energy cost of training a Large Language Model or the water required to cool a GPU cluster. But there is another, more insidious drain on our planetary resources—one that doesn't solve a complex problem or generate a creative image, but exists solely to track us. This is the Surveillance Tax.

The Parasitic Layer: Useful Compute vs. Parasitic Compute

To understand the energy crisis of the modern web, we have to distinguish between two different types of computation. First, there is Useful Compute. This is the processing power required to render a page, send a message, calculate a trajectory, or run an AI model that helps a doctor diagnose a disease. This is the work the user actually wants performed.

Then, there is Parasitic Compute. This is the invisible machinery of the ad-supported internet. It is the layer of tracking pixels, bid requests, real-time auctions, analytics, fraud checks, profiling, and measurement that triggers every time you load a website or open an app. The user didn't ask for this. The user doesn't see it. But the data center still has to process it.

We have built a global infrastructure where a significant portion of our total compute capacity is dedicated not to serving the user, but to harvesting the user. Across the internet as a whole, a defensible estimate is that advertising and tracking may account for something like 5–15% of infrastructure energy, with one 2016 academic estimate placing online advertising near 10% of total internet infrastructure energy. In individual data centers, the share depends entirely on what they host: near zero for some enterprise workloads, but potentially substantial inside ad-funded search, social, news, video, and app ecosystems. We are burning power-plant-scale energy to power a system that essentially asks, "Who is this person, and what can we sell to them right now?"

The Machinery of Waste: Real-Time Bidding (RTB)

The primary engine of this waste is a process called Real-Time Bidding (RTB). To the casual user, an ad is just a banner on a screen. But behind that banner is a high-frequency auction that happens in milliseconds. When you click a link, your data—your location, your device ID, your browsing history, your interests—is broadcast to dozens, sometimes hundreds, of potential bidders.

This is not a single request. It is a cascade of network calls. Each bid request triggers a chain reaction: servers must be queried, databases must be searched, identity matching must occur, and fraud detection algorithms must run to ensure the "impression" is real. This process happens hundreds of times per person, per day.

In the United States, it is estimated that a single person’s online activity and location are exposed through RTB roughly 747 times per day. That number does not mean 747 visible ads. It means hundreds of hidden data broadcasts, many of which happen before the user has even finished loading the page. In Europe, the average is around 376. Multiply those hundreds of auctions by billions of users, and you realize that the "invisible" part of the internet is actually a massive, churning engine of energy consumption. The waste is not the ad itself; it is the machinery underneath the ad.

The Math of the Surveillance Tax

When we look at the energy cost per person, the numbers seem negligible. But the danger of the cloud is that it hides scale. When you multiply a "tiny" amount of energy by a billion people, you aren't looking at a rounding error—you are looking at a systemic crisis.

Based on current infrastructure overhead, the modern internet likely spends between 0.125 and 0.225 kilowatt-hours (kWh) per person, per day on the machinery of advertising and surveillance. For one individual, that is a drop in the bucket. But let's scale that to a population of one million daily users.

To put that into perspective, the average U.S. household uses about 10,500 kWh per year. The hidden ad-tech layer for just one million people consumes enough electricity annually to power 4,300 to 7,800 U.S. homes.

And then there is the water. Data centers require immense amounts of water for cooling to prevent server meltdowns, as well as indirect water used in the generation of the electricity they consume. When we scale this to the global internet population—roughly 6 billion users—the numbers become staggering. The global ad-tracking layer potentially consumes between 274 and 493 Terawatt-hours (TWh) per year and billions of gallons of water every single day.

The ad-supported internet does not just sell your attention; it burns electricity and water to chase it.

The False Solution: Bigger Power Plants

The current industry response to the data center energy crisis is to build more power. We see a rush toward Small Modular Reactors (SMRs) and a revival of nuclear energy. While clean energy is a necessity, there is a dangerous logic at play here: the belief that we can simply "engineer our way out" of waste by increasing the supply of power.

But adding more power to a parasitic system doesn't solve the problem; it only subsidizes the inefficiency. If we build new power plants to support an internet that still spends a meaningful share of its energy on real-time bidding auctions for toothpaste ads, we aren't innovating—we are just fueling a leak. The solution isn't just more energy; it is less wasted compute.

The Local-First Mandate: Efficiency as a Virtue

This is where the philosophy of Local-First software enters the conversation. For a long time, "Local-First" and "Privacy-First" have been framed as ethical or legal choices. We argued that your data should stay on your device because it is your right to privacy. But there is a second, equally compelling argument: Local-First is a sustainability strategy.

The current model of the web is hyper-centralized. Every tiny action—a like, a scroll, a search—can require a round-trip to a massive, water-cooled data center. By moving the logic, the storage, and the intelligence to the "Edge"—the user's own device—we reduce the need for that round-trip. We stop the parasitic drain on the central grid.

Imagine an internet where your personal data doesn't need to be broadcast to 747 bidders a day because the "matching" happens locally on your device, or better yet, doesn't happen at all because the app doesn't track you. By stripping away the surveillance layer and optimizing the UI to be lightweight and efficient, we aren't just protecting the user's privacy—we are reducing the global demand for data center expansion.

At Clairos, this is the core of our architectural mission. When we build tools like our Journal app, we aren't just choosing "local storage" to be polite about privacy. We are choosing it because it is the only sustainable way to build software. We believe in Cognitive Ergonomics—tools that fit the human mind without requiring a planetary sacrifice to operate.

Conclusion: A Smarter Edge

We are at a crossroads. We can continue building a centralized, ad-funded internet that treats attention as the product and infrastructure as someone else's problem, or we can move toward a model of software that is lighter, more local, and more respectful by design.

The data center drama of today is a warning. Global data center electricity demand is projected to more than double by 2030, reaching around 945 TWh per year, driven heavily by AI and other digital services. That does not mean every data center is wasteful, or that all cloud computing is bad. It means the internet is no longer abstract. It is physical infrastructure with physical consequences: power lines, water systems, land use, cooling equipment, and grid pressure.

That is why the surveillance tax matters. If we are going to spend more electricity on computation, that computation should serve the user. It should help people think, create, communicate, heal, learn, and solve real problems. It should not be burned on invisible auctions, behavioral profiling, and hundreds of daily data broadcasts that exist only to squeeze more value out of a human being's attention.

We do not need a bigger cloud for everything. We need a smarter edge. We need software that keeps personal data close, reduces unnecessary round-trips, strips away parasitic tracking, and respects both the user's boundaries and the planet's limits. Local-first architecture is not just a privacy choice. It is an efficiency choice. It is a sustainability choice. It is a refusal to waste planetary resources on systems people never asked for.

The next time you load a page and see a flickering ad, remember that it is not just a nuisance on your screen. It is a signal from a much larger machine: one that burns power, water, and trust to chase a data point that should have stayed in your pocket.