Hey! Let’s slow the pace of artificial intelligence (AI) development! Otherwise, in a year or two we’ll have Skynet. Then the Terminator. And we know how that turns out. 

Let’s dig in…

Every Bubble Finds Its Pin

For several years, Tech Bros and their banker-enablers have shoveled big bags of money into the roaring furnace of AI Capex; a trillion dollars and more. 

The sector has contractually committed itself to suck dry a Great Lake or two worth of water; plug deep into the American power grid; buy up every GPU rack that GPU rack-makers can produce; and justify it all with the promise of a sidekick robot in every house, cooking that proverbial chicken in every pot due to the vast wealth it’s all going to create. 

“Sit back, and leave the driving to us,” they say. And they mean it, considering the promises of automated driving! 

But now, those same Tech Bros advise caution. Dario Amodei, the chief executive of Anthropic, just called for a global slowdown in AI development. He joins the ranks of Sam Altman, Elon Musk, Steve Wozniak and a parade of other researchers and policy influencers who want to tap the brakes. 

An organization called The Future of Life Institute has urged “all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4.” It warns of an “out-of-control race” to build systems “that no one — not even their creators — can understand, predict, or reliably control.”

Major media have also climbed into the pulpit to preach salvation through safety. The New York Times has headlined the issue. The Wall Street Journal recently ran an article entitled “Pause AI for Humanity’s Sake.” Holy smokes… do it “for humanity!”

How convenient. Because when a boom moves this fast, burns this much cash, and still can’t outline its saleable product, let alone name the actual deliverable, it’s time to pull the fire alarm, if not head for the exits.

Or put it this way: at one point in life, I practiced bankruptcy law. And I know what it looks like when a company has blown out the cash, maxed the credit line, and now can’t pay its bills. The board gets some “financial religion,” so to speak. And the duly appointed Vice President for Public Remorse appears in federal court before a bankruptcy judge: “Your Honor,” he says, “We never expected that this would happen.”

No, not at all. Nobody ever saw this big, smoke-belching train rolling down the tracks, going “Whooo… Whoooooo!”, like the Union Pacific Big Boy last summer. Everybody totally missed that vibe in the rails, right?

Well, maybe not. Let’s just ask, for argument’s sake, what’s really going on? Will AI kill us all, if not at least destroy civilization? Or is this newfound AI corporate humanitarianism and caution more like the recognition of mathematical reality, aka revenge of the spreadsheets? 

In other words, the AI boom is no longer just a tech story. It's become a massive debt story, a water-and-power story, a permitting story, a copper story, and, above all, a “Show Me the Money” story.

“Safety!”, Say the Guys Who Set the House on Fire 

Apparently, the go-go, never-stop tech industry wants you to believe that AI has moved too fast. 

And maybe, just maybe… they’re not wrong: “It’s the most potentially dangerous human development since the atom bomb,” according to an old friend who works deep in the heart of no less than Carnegie Mellon University, on exactly this segment of computer science. And I believe him.

But there’s not just one angle here. Yes, beware of Skynet and the Terminator; but what else is going on behind the curtains? 

Begin with the fact that big players have committed vast amounts of money. One outfit called the Futurum Group estimates that the five largest U.S. cloud and AI infrastructure providers are on track to spend not quite $700 billion just in 2026; atop what they spent in 2025, 2024, etc., and whatever is planned for 2027. The buy is for AI compute systems, data centers, networking, and supporting power stacks. 

Perhaps it really is just the inevitable March of Progress. Indeed, we always have some manner of innovation; but then again, from another angle, much of AI resembles the Yellowstone Park Fires of 1988, which burned up a century of dry tinder. 

And for all the AI money spent so far, what’s the output? A chatbot that writes memos? A search box that argues with you? A customer-service entity that makes you beg for a real human being?

Okay, just kidding (sort of). The issue isn't whether AI is useful; yes, it is, just like railroads, radio, fiber optics, and the internet. In fact, AI has already solved some otherwise unsolvable math problems, and just last week it helped the Houthis in Yemen target Saudi oil installations, per news accounts. 

But the real issue is whether today’s AI buildout will ever generate enough high-margin, recurring revenue to pay back invested capital. Never mind repaying the next trillion dollars that are programmed to roll down the conveyor belt into the Capex crusher, concentrator, and blast furnace.

And now we come to why the timing of this current “AI Pause” sermon is so convenient; it gives many Tech Bro CEOs an elegant excuse. Cuz nope, they didn’t overbuild, nor misprice their business idea, nor confuse venture-cap theology with earnings power. No-no-no… They were responsible stewards of the investors’ money! And then along came “society,” to demand a timeout. And hey, we have those busybody government regulators, too. 

The modern AI Luddites, those High Priests of Precaution, pulled the air brakes and saved us all from the Terminator. And, purely by coincidence, they saved management from explaining why their business models are moving fast to the point of insolvency, and why they can't pay their looming power and water bills.

Okay, hold that thought…

China Marches to a Different Drumbeat 

Now we get to an inconvenient comparison: the U.S. boom-boom approach to AI versus what’s happening in China-China-China. 

That is, in America the AI pitch has been “one model to rule them all.” Build a giant frontier system and cut it loose within the Internet. It’ll do everything: write code, summarize libraries, create term papers for college kids, draw funny pictures, diagnose rashes, plan vacations, pass the bar exam, flatter executives and maybe discover new frontiers of physics, all between quarterly updates. 

In other words, it’s the swing-for-the-fence model: expensive, glamorous, vague and wonderful for stock promotion and political rah-rah. USA! USA! USA! 

China, by contrast, has aimed AI development towards its factories. No poetry or metaphysics; just apply this new tech to make factories run better. News accounts reveal that Chinese smart factories already deploy tens of thousands of vertical AI models across production scheduling, predictive maintenance, digital twins, machine vision, quality inspection and industrial software. China’s goal is to tighten tolerances, catch defects earlier, reduce scrap, learn from production mistakes, and build better products faster so that their vendors can sell them at lower cost.

In short, China is not off on some Quixotic mission to seek what Tech Bros call “total addressable market.” Nope, just better margins. And as America spends fortunes following some Yellow Brick Road to find the Grand AI Oracle in Emerald City, China slipstreams behind American developers to scrape, data-mine, query, and distill whatever it can from U.S. models to accelerate its own work. 

Indeed, many U.S. players have accused Chinese firms of industrial-scale IP theft, labeled “distillation.” Brazenly, Chinese systems extract billions of AI tokens from American frontier models via millions of daily requests. And when called out, Beijing’s answer is the usual shrug: hey, this kind of thing is common; China adds its own special content; therefore, this is not theft, merely “innovation with Chinese characteristics.” 

It’s an old story: copy the expensive part, add a new hood ornament and call it domestic Chinese achievement. It’s how a General Motors Chevy became a Chinese “Chery” automobile back in the 1990s.

Drag the Grid Into the Circus Tent 

Now, here’s a part I actually like. Because AI may be a bubble, but it’s not an imaginary bubble. It’s concrete and copper, transmission lines, substations, electrical equipment, and much more. In a sense, data centers are big industrial plants with better public relations.

The fact is that America’s power grid limits AI growth. Indeed, a firm called Grid Strategies has warned that data-center electric demand currently runs far ahead of what utilities can connect. In essence, the AI crowd has finally uncovered a platform that won’t download: the power grid, which is old, overstretched, and often managed as if electrons can be summoned by flipping a switch. (Hint: no, it doesn’t work like that.) 

Thus, AI has forced the issue. The U.S. needs new everything: gas and nuke generation, power lines, transformers, switchgear, substations, backup power, cooling systems, electrical steel, copper, silicon carbide, control systems, skilled trades. Indeed, the whole neglected electrical-industrial stack is now back on the national policy menu.

Watch the Picks-and-Shovels 

If –when! – the AI bubble pops, the Tech Bro glamour trades will get a buzzcut. But physical bottlenecks won’t disappear just because stock promoters ran out of Silicon Valley bullsh!t. Just follow the money, which leads to a long list of shortages: skilled labor, energy supply, power equipment, grid hardware, and cooling capacity.

Specific sectors to watch: 

• Electric utilities with real load growth and regulatory support; 
• Transmission and distribution contractors; 
• Transformer and switchgear manufacturers; 
• Copper, electrical steel and specialty metals; 
• Gas turbines, backup generation and grid-scale power systems; 
• Cooling and HVAC contractors; 
• Power semiconductors, especially silicon carbide; 
• Nuclear fuel, uranium and reactor services; and 
• Data-center infrastructure firms that get paid to build, connect and cool the beast rather than pretend it will someday write poetry for profit.

We have many investable names here in our stable of Paradigm Press pubs, so stay subscribed. Meanwhile, never chase a company just because you saw a news article and some CEO said “AI” a few times during an earnings call. Follow the facts: look for backlogs, pricing power, domestic manufacturing advantage, regulated returns, scarcity of skilled labor, and contracts tied to hard infrastructure instead of vaporware pitches. 

Wrap Up

The AI pause is being sold as ethics, safety, and even wisdom: “Save humanity! Defeat the Terminator!”

Okay, that’s sweet. But first, follow the money. When an entire industry has spent like a drunken empire and still can’t point to a profit engine large enough to justify the buildout, the pause becomes a bit too timely and convenient.

Looking ahead, AI evangelists will keep selling their snake oil and digital rapture. But as you plan your life, be sure to ask who gets paid out of all this, and who is set up to lose a metric tonne of money. Because when the bubble bursts, chatbots everywhere will vanish into clouds of depreciation and bankruptcy filings. 

On the other hand, the transformers, turbines, cables, cooling plants and skilled hands necessary to electrify the next economy are still real and will stay that way. In a mania built on digital promises, the best opportunity may be hiding in plain sight, as old-fashioned hardware.

That’s all for now. Thank you for subscribing and reading.