AI Doesn't Live in the Cloud. It Lives in Buildings.
The most software-driven industry in history has a plumbing problem.
Elon Musk's Latest Job Posting Says More About the Future of AI Than Any Product Launch
A few days ago, Elon Musk posted an unusually short job listing.
SpaceX is hiring engineers and skilled tradespeople to build the world's most powerful AI supercomputer clusters. Send three bullet points and a résumé to an email address. That's the whole application.
He subtly said two things at once: he’s creating the best data centers ‘off Earth,’ but also that he’s not just hiring engineers, but he specifically mentioned tradespeople: Plumbers. Electricians. Construction professionals. The people you'd hire to build a hospital or a factory, being recruited to build the machines that will run the next generation of artificial intelligence.
This isn't one billionaire being eccentric. Jensen Huang, whose company makes the chips half of Silicon Valley is fighting to buy, has been saying versions of the same thing for over a year. When he's asked what the AI boom will produce, he doesn't talk about coders. He talks about plumbers and electricians and steelworkers. At Davos, sitting next to the CEO of BlackRock, he called it "wonderful" that so many of the new jobs will be tradecraft.
Two of the most powerful people in the technology industry are quietly telling us the same thing. The AI economy runs on hands.
We've Been Looking in the Wrong Direction
For twenty years, career advice went one way. Sit at a keyboard. Learn to code. Get away from anything that involves lifting. The world was going digital, and the safest place to stand was inside a screen.
Now the companies building the actual digital world are, by their own admission, desperate for the people we told to go somewhere else.
Here's the piece almost nobody explains, and it's the part that changes how you see all of it.
AI has weight.
A large language model, running at the scale of ChatGPT or Claude, doesn't live in a cloud. It lives in a building. That building holds racks of GPUs — physical chips — arranged in rows that consume more electricity than most towns. The waste heat from those chips would cook them in an afternoon, so the building needs cooling systems more elaborate than most hospitals. Some data centers run on chilled water pumped through miles of copper. Others use immersion tanks where the servers sit submerged in liquid.
None of that gets built by writing prompts.
Every AI Breakthrough Becomes a Construction Project
Every AI breakthrough you read about — every new model, every capability jump, every headline about how much data centers are consuming — is a downstream order for concrete, steel, copper wire, high-voltage switchgear, industrial HVAC, chilled water systems, backup generators, fire suppression, structural welding, and grid interconnection. There are more than 3,000 operational data centers in the United States right now, and another 1,500 in development. Every single one requires the trades to exist before it can run a single line of code.
Behind the buildings is a power grid that has to grow to feed them, and the grid needs electricians. In front of them are chip fabs the size of small cities, and those fabs need plumbers and HVAC technicians in numbers nobody's fully budgeted for yet. Around them are new substations, cooling towers, and transmission lines that will keep being built for the next thirty years.
The pattern isn't that AI creates some trade jobs. The pattern is that every serious step forward in AI raises the ceiling on how much physical infrastructure the world has to build. There is no software-only version of this future.
The Career Advice We Got Wrong
For a long generation, we told young people the trades were the fallback. Something you did if college didn't work out. Something you did with your hands because you couldn't do it with your head. This was always insulting and always wrong, and now it's about to become an expensive mistake.
The companies deciding the future of AI are telling us, plainly, that the constraint on their business isn't compute. It's construction. Not software. Buildings. Not engineers. Everyone who builds the world engineers rely on.
You could not design a more dramatic reversal of twenty years of career advice if you tried.
Notice what this means for anyone in a trade right now, or thinking about entering one. You're not choosing a "safe fallback." You're choosing an industry that a handful of the most well-capitalized companies in the world are about to spend a decade fighting over. The demand curve for skilled trades over the next twenty years does not look like a comfortable, steady rise. It looks like the demand curve for railroad workers in 1870.
It also means something for owners of trade companies. The plumber, the electrician, the HVAC contractor — you are not running local service businesses anymore, at least not entirely. You are running the on-ramp to national infrastructure. Some of your best techs, five years from now, will be recruited to build data centers. Some will leave to run their own crews on those builds. Your entire labor market is about to be tugged into a different orbit by companies most of your customers can't even name.
Whether that's good news or bad news for you depends on how you're built. A shop that pays what techs are worth, trains what apprentices need, and treats the work as skilled — that shop just became strategic infrastructure. A shop that treats the trades as commodity labor is about to lose people to companies who know exactly what they're worth.
Why This Matters to Contractors
I think the deeper story isn't about AI, actually. It's about a country slowly, awkwardly, remembering something it forgot.
The physical world is not a solved problem. It never was. We spent twenty years pretending everything important was moving into software, and the whole time, the servers running that software were sitting in buildings that had to be plumbed, wired, cooled, and maintained by people we told to sit at a keyboard instead.
Now the people who wrote the software are hiring the people we told to go the other way.
The next time someone tells a kid the trades are a fallback, I hope somebody in the room is honest with them. The tradesperson who wires the next generation of data centers, or plumbs the next generation of chip fabs, or keeps them from cooking themselves in a heat wave — that person isn't below the software engineer in some invisible hierarchy of importance.
They're the reason the software engineer has a job at all.
AI may write the code. Someone still has to build the world it runs on.
That's not a fallback. That's the foundation.