There is a moment when it is clear that something has subtly changed in the American economy while standing outside a partially constructed data center in the Virginia suburbs, complete with exposed cabling, concrete floors, and the smell of cut copper. There are no laptops in the hands of those who are working on the most pressing tasks here. They have pipe wrenches in their hands.
For many years, the narrative we told ourselves about success in America followed a very particular path: pursue a degree in computer science, get a job at a tech company, and earn a salary that made your parents proud over Thanksgiving. There is some truth to that story. However, in ways that are a little difficult to ignore, it is starting to fall short.
| Category | Details |
|---|---|
| Topic | Skilled Trades vs. Software Engineering — Salary & Demand Shift in the U.S. |
| Key Figures Referenced | Jensen Huang (Nvidia CEO), Dario Amodei (Anthropic CEO), Alex Karp (Palantir CEO) |
| Industries Affected | Construction, Electrical, Plumbing, HVAC, Software Engineering, AI/Data Centers |
| Geographic Scope | 30+ U.S. States |
| Salary Range (Trades) | $80,000 – $150,000+ annually (experienced tradespeople) |
| Key Driver | AI data center infrastructure boom requiring massive physical buildout |
| Projected U.S. Investment | $500 billion+ in data center construction commitments |
| Reference Source 1 | Nvidia Official Newsroom — AI Infrastructure |
| Reference Source 2 | U.S. Bureau of Labor Statistics — Electricians & Plumbers Wage Data |
In 30 U.S. states, the average salary of electricians and plumbers is now higher than that of many mid-level software engineers, and the difference is growing in ways that hardly anyone could have predicted five years ago. According to some reports, salaries in skilled trades have almost doubled in some markets. This may not be fully explained by a single factor, but the boom in AI infrastructure comes very close.
The CEO of Nvidia, Jensen Huang, stated unequivocally at the World Economic Forum in Davos earlier this year that the need to build the data centers that power and train artificial intelligence will drive six-figure salaries for plumbers, electricians, and construction workers.
He wasn’t being philanthropic or giving. He was describing what Nvidia, a company that is expected to sell nearly $200 billion worth of data center chips in 2025, is witnessing in real time. “Salaries have gone up nearly double,” he said. “Everyone ought to be able to earn a decent living. A PhD in computer science is not required to accomplish this.
At Davos, that statement was overlooked in favor of louder discussions about chip exports to China and AI regulation. However, it was not given the attention it deserved.
The scale may not be simple, but the explanation is. Data centers are necessary to build AI. Land, steel, cooling systems, high-voltage electrical infrastructure, and miles of pipe and cable are all necessary for data centers. That doesn’t come from a prompt. It is not based on a large language model. It takes skilled human hands, and there aren’t enough of them at the moment.
In the United States, there is a real shortage of qualified electricians and HVAC technicians, and the buildout that corporations like Microsoft, Meta, Amazon, and Google are demanding has made the shortage resemble a bidding war. Over the next few years, tech companies have committed to leasing data centers for a total of $500 billion. All those structures require wiring. Cooling is necessary for every server rack. Plumbing is necessary for every cooling system.
It’s difficult to ignore the irony. The same AI systems that pose a threat to entry-level coding positions are also creating a huge demand for non-coding workers. Dario Amodei, CEO of Anthropic, has issued a warning about a “white-collar bloodbath” that could eliminate half of entry-level knowledge worker positions. The pressure is already being felt by junior software engineers. But what about the electrician who comes to wire the server room that powers those AI systems? He is negotiating an increase in his rate.
This change feels more than just a statistical anomaly because it validates long-held beliefs among those in the trades. This work contains genuine intelligence. When designing a new system from the ground up, a plumber must use spatial reasoning, problem-solving skills, and hard-won experience to run supply lines, account for pressure differentials, and think three problems ahead.
The person who brushes it off as “just plumbing” is typically the one who has never had to sort through someone else’s disorganized system of pipes that run in all directions without any logic. Those are the systems that malfunction the most. They are also the most costly and difficult to repair.
Some self-employed plumbers and electricians make well over $500,000 annually. That is uncommon and difficult; the owner bears a large portion of the responsibility, and there are real physical demands and scaling challenges. However, it is taking place. Additionally, working tradespeople now have a much higher floor. In a market dominated by data centers, a licensed electrician is barely making ends meet. They are accumulating wealth.
In the meantime, the employment market for software engineers is going through something new: real uncertainty about its own future. The number of entry-level positions is declining. Renegotiations are taking place for mid-level positions. It appears that the skills that were meant to be recession-proof are less certain.
There’s a feeling that the American economy is correcting something it got slightly wrong for a generation, which may be unsettling to many. We encouraged a large number of young people to pursue four-year degrees in computer science, told them that the trades were a less desirable path, and then watched in slow motion as that less desirable path started to produce comparable, and in many cases, better, financial results without the student debt.
It’s still unclear if this is a transient spike brought on by a very particular infrastructure buildout or a permanent rebalancing. However, observing a group of electricians running conduit in the late afternoon light outside that incomplete data center in Virginia, it’s hard to avoid thinking that the more crucial question isn’t whether this lasts. It’s whether we were ever measuring the correct things in the first place.

