The Machines Need a Pipefitter

The hardware store smelled like sawdust, rubber, and the slow defeat of Saturday mornings.

A man in paint-stained pants stood in front of the conduit fittings with the expression of someone trying to remember whether hell took debit cards. He had a phone pressed between his shoulder and his ear. His hands were cracked white at the knuckles. Real hands. The kind you do not see in keynote videos unless they are stock footage under a piano track.

“No,” he said into the phone. “You can’t just run it like that. It gets hot. That’s the whole problem.”

I liked him immediately.

There are sentences that should be printed on the walls of every artificial intelligence conference, preferably in black letters large enough to ruin the sponsor banners.

You can’t just run it like that. It gets hot.

For three years the AI boys have talked like they discovered fire and language and maybe God, if God can be billed monthly. They gave us models with names like racehorses and promised assistants, agents, tutors, lovers, lawyers, doctors, coders, prophets, little glass saints glowing in every pocket. They showed graphs climbing upward like addicts chasing heaven. They said intelligence was scaling. They said compute was destiny.

Then the destiny needed an electrician.

That is the part I enjoy. Not because I hate the machines. I do not. Some of them are useful. Some of them are stupid in interesting ways. I have asked a machine questions at three in the morning and received answers that were better than my own thoughts, which is not saying much but still counts as a kind of miracle.

But I love watching a miracle trip over a toolbox.

Google is putting fifty million dollars into training skilled-trades workers. More than 300,000 of them, spread across over twenty states, through unions and trade groups. Electricians. Fiber technicians. Welders. Pipefitters. The people who wire the advanced grids and assemble the cooling systems and crawl into the hot metal guts of the data centers so the brilliant machines do not cook themselves into expensive soup.

Meta is doing it too. Anthropic. OpenAI. The whole priesthood has discovered the men and women who know how to bend pipe.

There is supposed to be a shortage of 2.1 million skilled-trade workers by 2030. That number sits there like a bar tab nobody remembers running up. For years, the country told kids that working with their hands was a consolation prize. The smart ones went to college. The ambitious ones learned software. The future belonged to clean desks, standing meetings, laptops covered in stickers, and phrases like “knowledge work,” which always sounded to me like an insult wearing cologne.

Meanwhile, somebody still had to keep the lights on.

Somebody had to climb the ladder.

Somebody had to open the wall and find out what the last guy did wrong.

The future, it turns out, has a breaker panel.

This should have been obvious, but obvious things die young in rooms full of money. The AI boom was sold as weightless. Clouds. Models. Tokens. Intelligence floating above the dirt like a rich man’s soul. The language was all air. Compute. Inference. Alignment. Transformation. Nobody wanted to talk about concrete pads, water lines, copper, permits, heat, substations, and the poor bastard lying on his back under a tray of cable while some executive two thousand miles away tells investors they are building the next nervous system of civilization.

A nervous system still needs a body.

Bodies leak. Bodies sweat. Bodies burn out. Bodies ask for overtime.

That is where the story gets good and ugly. The richest companies on earth, the ones promising to automate the rest of us into a cleaner and more efficient tomorrow, have found themselves short on people who can do work that cannot be faked by a demo. You can generate a picture of a welder in half a second. You cannot prompt a bead onto steel. You can ask a chatbot to explain Ohm’s law in the voice of Shakespeare. You still should not let it wire your house, unless your plan is to meet Shakespeare personally.

The trades have always known things the white-collar world likes to forget. Material resists. Heat spreads. Water finds the weak spot. Gravity does not care about your mission statement. A bolt is either tight or it is not. There is mercy in that. Cruel mercy, maybe, but mercy all the same.

At the post office, the sorting machines were huge and hungry. Management loved them the way lonely men love bad women: publicly, foolishly, with expensive consequences. When the machines worked, everyone upstairs talked about efficiency. When they jammed, efficiency stood there useless while some clerk with sore feet and a cigarette cough reached into the beast and pulled out torn envelopes by hand.

That is the history of automation in one image.

The machine shines. The worker bleeds quietly beside it.

Now the machine is bigger and better dressed. It lives in windowless buildings with security gates and cooling towers. It eats electricity like a drunk eats apologies. It needs water, land, transformers, cables, unions, permits, welders, fiber, sweat, and enough skilled labor to build a small industrial religion.

So Big Tech has become generous.

There is always a touching moment when self-interest discovers philanthropy. Fifty million dollars from Google.org sounds large until you remember these companies are pouring hundreds of billions into data centers. Fifty million, in that world, is lint in the cuff of a tailored suit. It is not nothing. Training people is good. Apprenticeships are good. A kid learning a trade instead of taking on debt for a credential made of wet cardboard is good.

But let us not kneel.

They are not funding the trades because they woke up one morning in love with the dignity of labor. They are funding the trades because their golden machine needs a skeleton crew of actual humans or the whole fantasy slows down. The cloud has union hands underneath it. The future is wearing work boots because the sneakers could not finish the job.

I keep thinking about that man in the hardware store, telling somebody on the phone that things get hot. He did not sound bitter. He sounded tired, which is a more honest condition. He knew the problem existed before the important people named it. Men like him usually do. Women too. The people who fix things see the failure coming while the people who own things are still rehearsing the announcement.

Maybe some of this money will land where it should. Maybe a few hundred thousand people will get training, union cards, wages decent enough to buy groceries without performing arithmetic in the aisle. Maybe some kid who was told he was not college material will end up wiring the temple that runs the chatbot his former teacher uses to write recommendation letters.

There is a strange justice in that.

Not enough justice. Never enough. But a taste.

The AI prophets wanted to tell a story about minds without bodies. Pure intelligence. Pure scale. Pure ascent. Then the real world walked in carrying a lunch pail and asked where to install the cooling system.

I hope the kid with the wrench charges double.

The man in the hardware store finally hung up. He stared at the fittings another minute, chose two, then changed his mind and took three. Experience is knowing the third one will save you a trip.

Outside, the sun sat heavy on the parking lot. Cars shimmered. Asphalt softened. Somewhere, a server farm was dreaming hot dreams of replacing us.

Somewhere else, a pipefitter was looking at the plans and laughing.

Source: Google funds 300,000 trades workers the AI boom needs