Is AI Just a Fad? The Honest Answer Isn't Yes or No

No, AI isn't a fad — but that's the wrong question. The right one is two questions wearing a trench coat: is the technology durable? and is the money around it a bubble? The honest answers are "almost certainly yes" and "quite possibly." Both can be true at the same time. We know this because it already happened once — the internet was completely real, and 1999's stock prices were completely insane.

I get asked this constantly, usually some version of: "Ricky, is this AI stuff gonna blow over like 3D TVs?" It's a fair question from people who've watched a lot of Next Big Things come and go.

So let me put my cards on the table first: I sell AI setup and training. That means you should take my enthusiasm with the appropriate grain of salt — which is exactly why I'm going to give you the bear case in full, with real numbers, before I tell you what I think. If I only gave you the sunny version, I'd be a salesman, not your tech guy.

The question underneath the question

When somebody asks "is AI a fad," they usually mean one of three different things:

Mixing these up is how people get hurt. In 2000, plenty of folks correctly figured out that internet stocks were overpriced — and then concluded the internet itself was a fad. They were right about the bubble and catastrophically wrong about the technology.

Is anybody actually using AI, or is it all noise?

Here's where I have to be careful, because the honest numbers are lower than the headlines. The gold standard is the US Census Bureau's Business Trends and Outlook Survey — an actual probability sample of American businesses, not a vendor poll. As of May 2026, it found 19.8% of US businesses reported using AI in their business functions. About one in five.

And two honest caveats that most articles skip:

Now the part that matters for Main Street: adoption is heavily tilted toward big companies. Census found AI use at 37% among firms with 250+ employees, but under 20% among firms with fewer than 20 people — and between December 2025 and May 2026, use rose among firms with 20+ employees while showing no statistically significant change among the small ones. If you run a five-person shop in Hahira and feel like you're behind, look around: most of your peers haven't moved either.

But here's the counterweight, and it's a strong one. JPMorganChase Institute looked at actual bank transaction data from over 4.6 million small businesses — real money leaving real accounts, not survey answers. They found 17.7% were paying for AI services by the end of 2025, up from 1.7% in 2019. And the adoption speed is startling: the 2025 cohort of small businesses hit a 10% adoption rate in six months. The 2019 cohort took 77 months to get there.

That's roughly thirteen times faster. Fads don't usually accelerate like that — they spike and collapse.

Will people run out of ideas for it?

This is the part of the worry I can most confidently set aside. Nobody ever "ran out of ideas" for electricity, or the internet, or the telephone. That's not how a general-purpose technology behaves — and "general-purpose technology" is an actual economic category, not a marketing phrase. The classic 1995 definition from Bresnahan and Trajtenberg gives three tests: is it used as an input across many different industries, does it keep improving, and does it make innovation easier in the industries that adopt it?

Whatever else you think about AI, it clears all three bars. Usage bears that out at a scale that's hard to picture: Google went from processing 9.7 trillion AI tokens a month in May 2024 to over 3.2 quadrillion by May 2026. Microsoft processed over 100 trillion in a single quarter, five times the prior year.

If anything, history says the opposite of "running out of ideas" — it says we're early, and the useful ideas come slower than expected. Electricity was available in the 1880s, but manufacturing productivity didn't measurably accelerate until the 1920s. Economist Paul David's famous explanation: factories had to be physically rebuilt around small individual motors instead of one giant central steam engine driving overhead belts. Just swapping the power source changed almost nothing. The gains came four decades later, once people redesigned the work itself.

The genuinely unknown part — and I'm not going to pretend otherwise — is whether the AI models themselves keep improving at the pace of the last few years. Smart, credentialed people disagree about that, loudly. Anybody who tells you they're certain which way that goes is selling something.

Now the part nobody wants to hear: the money might be a bubble

I'd be doing you a disservice if I skipped this. There's a serious, well-documented case that the investment around AI has run ahead of the business reality.

The Bank for International Settlements — the central bank for central banks, not exactly a doom blog — put it bluntly in its 2026 annual report, grouping the AI buildout with history's great manias: "The canal mania of the 1830s, the British railway mania in the 1840s, the electrification exuberance of the late 1920s… and the dotcom boom of the late 90s all shared one common trait: a genuine technological breakthrough that attracted" more capital than it could productively absorb. They note the five largest hyperscalers are set to spend over a trillion dollars on AI capex across 2025–2026, and that these commitments "are outpacing earnings and the free cash flow of these firms, leading some to issue debt."

Some specifics that gave me pause:

Now the other side, fairly stated, because this isn't one-sided:

My read? A bubble in the financing is entirely plausible. A collapse of the technology is not. Which brings us to the most useful history lesson available.

What the dot-com crash actually teaches

Everyone reaches for this comparison and then draws the wrong lesson from it. Here's what actually happened.

The Nasdaq peaked at 5,048.62 on March 10, 2000, and fell to about 1,114 by October 2002 — roughly a 78% wipeout. Brutal, real, and it took 15 years to make back on paper.

Meanwhile, over that exact same stretch: US e-commerce sales grew every single year. Census Bureau figures put 2002 e-commerce at $45.6 billion — up 26.9% over 2001 — while the stock market was still smoldering. The technology never stopped compounding. Only the stock prices crashed.

And the single best cautionary tale is Cisco. Cisco's thesis was correct. They said the world would run on internet networking gear, and the world does. They're a healthy, profitable company today. But if you bought Cisco at its March 2000 peak of $80.06, you didn't get back to even until December 10, 2025 — a twenty-five-year wait — and its market cap is still well below the roughly $555 billion it touched at the top.

One more myth to retire: the dot-com bust is often described as a graveyard of doomed companies. Peer-reviewed research in the Journal of Financial Economics found the five-year survival rate of dot-com firms was about 48% — comparable to or better than early automobiles, tires, televisions, and penicillin. The authors argued there may have been too few dot-coms, not too many.

But real fads absolutely do exist

I'm not going to pretend everything sticks. Plenty of confidently-hyped technology died, and some of it got much further than people remember:

So what separates these from the internet? Every one of them was a product looking for a problem. Nobody's daily work got measurably easier because of a 3D TV. The tell isn't how much hype something gets — it's whether ordinary people quietly keep using it when nobody's watching, because it genuinely saves them time or money.

By that test, AI is behaving much more like the internet than like 3D TV. Which doesn't mean every AI company survives. Most won't.

About that "95% of AI projects fail" study

You've probably seen this one, and it deserves a straight answer, because it's the single most misused statistic in this whole debate.

The number comes from an MIT-affiliated report, and here's what it actually measured: custom and vendor-built generative AI pilots inside large enterprises. Not small businesses. Not everyday AI use. Big-company IT projects with big-company budgets.

And the same report found something that almost never makes the headline — the opposite result for cheap, general-purpose tools. Employees at roughly 90% of companies were regularly using ordinary tools like ChatGPT for work, while only about 40% of companies had actually bought official subscriptions. There's even an anecdote about a corporate lawyer who preferred a $20-a-month chatbot to her firm's roughly $50,000 specialized legal AI tool.

A few honest caveats on the report itself: it wasn't peer-reviewed, its methodology was never fully explained, and the group behind it works on AI infrastructure — so it has some interest in the problem it diagnosed. Treat it as a useful signal, not scripture.

But if you take one thing from it, take this: the expensive custom AI projects are what fail. The cheap general-purpose tools are quietly working. For a small business, that's just about the most encouraging finding possible — you were never going to buy the $50,000 thing anyway.

So what should a South Georgia business actually do?

Here's my honest advice, given everything above:

The short version

AI is not a fad. The money around it may well be a bubble. Both of those things can be true, and the dot-com era is the proof. The internet was real and the 2000 stock prices weren't; e-commerce grew right through the crash while Cisco shareholders waited 25 years to break even.

Nobody's going to "run out of ideas" for a general-purpose technology — but the useful ideas will arrive slower and more boringly than the hype promises. For a small business in Valdosta or Hahira, the winning move isn't to bet big or to sit it out. It's to use the cheap tools for the dull jobs, keep your data in your own hands, and stay flexible enough that it doesn't matter much who wins.

Frequently Asked Questions