A CEO of one of America’s oldest manufacturers told Fox News his company is on a path to double in size, and the reason is the AI boom. Not an AI product. Not an AI feature bolted onto a legacy business. Just the boom itself, arriving as demand for physical things that get built in physical buildings.
I review AI tools for a living. I poke at agent frameworks, I call out inflated benchmarks, I write unkind things about products that promise autonomy and deliver a wrapper around someone else’s model. So my first reaction to a manufacturing CEO crediting AI for doubling his business was skepticism, because that is my default setting. My second reaction was more uncomfortable: this is probably the most honest AI success story I have read this month, and it has nothing to do with software.
The part of the boom nobody demos
Here is what the current news cycle looks like when you line it up. An old-line American manufacturer expects to double. A company called Frontieras North America is publicly reframing coal for what it calls the AI economy. Consumer Reports is examining what AI data centers do to electric bills and water supplies. CNET is writing about data centers coming for land, water, and power. Understanding AI published sixteen charts trying to explain the boom to people who cannot see it.
Those stories are all the same story. The AI industry sells itself as weightless, something that lives in a browser tab and costs twenty dollars a month. The actual thing is heavy. It has a footprint measured in acres, a thirst measured in gallons, and a power draw that shows up on residential utility bills belonging to people who have never opened a chat interface.
When a manufacturer says AI is doubling his order book, he is telling you the shape of the boom more clearly than any model release note. Steel, transformers, cooling equipment, electrical infrastructure — someone has to make it, and the making is happening in old factories that were supposedly obsolete.
Why this matters if you buy AI tools
You might reasonably ask what any of this has to do with whether an agent framework is worth your time. More than you would think.
- Pricing is not stable. Every product you use runs on compute that requires land, power, and water. Those inputs are getting contested. Costs that seem generous today are subsidized by an investment cycle, not by unit economics.
- The physical bottleneck is real. If manufacturers are being asked to double capacity, capacity is what everyone is short on. Availability, latency, and rate limits are downstream of concrete and copper.
- Coal is back in the conversation. Read that Frontieras headline again. A company is pitching coal Whatever you believe about energy, this is not the clean-tech future the industry markets.
- Local backlash is coming. When Consumer Reports and CNET both run pieces on your utility bill and your water table, the political phase has started.
The uncomfortable accounting
I spend a lot of my time telling readers that a tool is not worth what it charges. This story flips the frame. The honest question is not whether a given AI product is overpriced. It is whether anyone is paying the real price at all.
A manufacturer doubling in size is a genuinely good outcome for the people who work there. American industrial capacity coming back online because of demand rather than subsidy is a legitimately interesting development, and I am not going to pretend otherwise for the sake of a contrarian take. But that growth exists because enormous physical resources are being redirected toward compute, and the cost of that redirection is landing on utility ratepayers and local water systems, according to the coverage above.
That is not a reason to stop using AI tools. It is a reason to stop describing them as immaterial.
What I would watch
The manufacturing story and the utility-bill story are the same trend viewed from opposite ends, and they will resolve together. If demand holds, factories expand and bills climb. If the buildout slows, both reverse. The sixteen-chart explainers exist because the boom has become too large to intuit, which is usually a sign that the people inside it have stopped being reliable narrators.
So my advice, as someone paid to be suspicious of AI claims: when you evaluate a tool, add a line to your mental spreadsheet for the infrastructure it assumes. Not because you can do much about it, but because a product whose costs are being absorbed elsewhere is not a product whose price you understand.
An old manufacturer doubling in size is a real signal. It says the boom is made of things. Things run out.
🕒 Published: