The Mainframe Never Dies — It Just Gets Hijacked by Enthusiasts

By Arick West | Published: June 24, 2026

I’ve been around long enough to hear the exact same lies they’re telling us about AI today.

If you were around in the 1960s or 70s (or just old enough to remember your parents talking about it), you know the pattern. The industry’s first instinct for any new computing technology is to make it a Mainframe with Terminals.

The Pattern: “Just a Terminal in Your House”

When the microcomputer revolution started (yes, I was there for the whole 60 years), the experts — the guys from IBM, DEC, Burroughs, the whole establishment — had a very clear vision:

“Every house will have a terminal that connects to the big central computer in the city. That’s how it works. You don’t compute at home. You just access the computer.”

It made sense to them. Mainframes were expensive, rare, and powerful. That’s where the value is. Keep it locked in the cloud, give people dumb keyboards to punch cards and read punch-out tape, and keep the money flowing in subscription rents.

Sound familiar?

How That Actually Ended

Here’s what the experts in 1975 absolutely, positively did not see coming:

The PC revolution didn’t start because executives decided PCs were a good idea. It started because a bunch of electronics nuts with crazy ideas built their own machines in garages.

• They didn’t wait for IBM’s permission.
• They didn’t wait for a “secure enterprise model.”
• They bought parts, soldered boards, and built machines the mainframe folks said were worthless — 4K of memory, primitive displays, no real security, but theirs.

And here’s the critical detail: the commercial PC industry didn’t win because it was better than the mainframe. It won because it was home-built first.

The enthusiasts built the prototype — thousands of them, independently, across the country. They proved the model worked in practice, not in a boardroom vision. Then someone (Apple, Commodore, Tandy) realized the hobbyist market was the market and commercialized what the hobbyists had already proven.

The “mainframe in the city” model didn’t die because it was obsolete. It died because enthusiasts refused to stay in their terminals.

We Are Here Again

The AI industry right now is having the exact same conversation IBM had in 1975:

“AI will be a centralized resource. You’ll have an apartment-sized terminal, pay per token, and never touch the compute directly. That’s the future.”

They’re not doing this because it’s better. They’re doing it because centralization is how they make money.

But I’m telling you, having seen it once before: it doesn’t work that way.

When people learn they can run a model on their own hardware that’s good enough (and let’s be honest, they’re getting very good at good enough), the whole subscription model starts to look like what it is — a protection racket selling access to technology that’s being made free by the community.

The Enthusiast Wave Is Coming

Just like in the 70s, the real innovation in AI isn’t coming from the enterprises trying to replicate the mainframe model.

It’s coming from:

• The hobbyists running LLMs on 2080s and 3090s in their basements
• The engineers building AI appliances that look nothing like the “AI terminal” the VCs drew in a pitch deck
• The people fine-tuning models on their own data without asking permission

And when the commercial winners emerge (and they will), they won’t look like Apple or Commodore. They’ll look like what the enthusiasts already built.

The “mainframe AI” model doesn’t die by being defeated. It dies by becoming irrelevant when millions of people realize they’ve been doing it better at home for free the whole time.

“I’ve seen this movie. Trust me — bet on the terminal builders, not the mainframe owners.”