The AI Gold Rush Will End Like Airlines: Great for Users, Terrible for Operators

I've been thinking a lot about airlines lately. Not because I'm traveling more, but because every time I use ChatGPT or Claude or any of the AI tools I've woven into my daily workflow, I see the same pattern emerging that turned aviation from a glamorous frontier into a brutally commoditized business.

And if I'm right, we're about to watch the AI industry follow the exact same trajectory.

The Uncomfortable Truth About Infrastructure

Here's what I keep coming back to: the best infrastructure businesses are terrible businesses to be in.

Airlines are the perfect example. They require massive capital expenditures. They operate on razor-thin margins. They're completely commoditized—most travelers pick based on price and schedule, not brand loyalty. Running an airline, frankly, sucks.

But the unlock airlines provide? That's massive. Global business travel. Tourism economies. Supply chains. Family connections. The entire modern world depends on cheap, reliable air travel existing.

The same pattern played out with shipping containers. Before containerization, global shipping was boutique, expensive, and labor-intensive. Then we standardized it all. Now it's a low-margin, commoditized business that most people never think about.

But that commoditization unlocked something extraordinary: modern global trade. Your ability to order something manufactured in Shenzhen and have it on your doorstep in Ohio a week later depends entirely on shipping being cheap and boring.

I think AI is heading down this exact path.

Why AI Will Become Boring Infrastructure

Look at what's already happening:

The models are converging fast. Six months ago, there were meaningful differences between GPT-4, Claude, and Gemini. Now? They're all basically the same for most use cases. The gap is closing weekly.

Open source is catching up. We're one or two breakthroughs away from having open-source models that match the commercial offerings for 90% of use cases. When that happens, the pricing pressure becomes brutal.

The capital requirements are insane. These companies are burning billions on compute, data centers, and talent. The circular spending between big tech companies—where Microsoft invests in OpenAI, which then spends that money on Azure compute—can't last forever.

It's starting to feel like the dot-com bubble. I lived through that one. I've got enough gray hairs to recognize the pattern. Massive investment. Sky-high valuations. Everyone convinced they're going to be the winner. Then reality hits.

And here's the kicker: AI spending is currently propping up the US economy. When this bubble deflates—and I think it will, probably not as catastrophically as dot-com but it will deflate—it's going to be painful.

The Good News (For Everyone Except AI Companies)

But here's what makes this bearable: the infrastructure will still be there.

After the dot-com crash, we didn't lose the internet. We lost a bunch of overvalued companies, sure. But the fiber optic cables stayed in the ground. The protocols kept working. The infrastructure remained, and ten years later we built entirely new businesses on top of it.

The same thing will happen with AI. The models will become commoditized utilities. Running an AI company will be a low-margin infrastructure play. But the unlock for the rest of us will be extraordinary.

I'm already seeing it in my own work. I use AI extensively—Notion AI for research and organization, ChatGPT for drafting and ideation, various other tools for production workflow. It's genuinely the most interesting technology I've played with in years.

My goal is that people see it in the velocity of my output, not in the writing itself. It's just me, but more. And that multiplier effect is real.

What This Means for MSPs and IT Services

Here's where it gets relevant for the channel:

In the short term, there's going to be turbulence. When AI valuations correct, it will ripple through the economy. Small businesses will feel it. IT services spending will take a hit. The timeline is unclear—things move faster now than in 2001, but "faster" doesn't mean "four months."

In the medium term, AI becomes a capability, not a product. You won't sell AI services. You'll use AI to deliver everything else more efficiently. Just like you don't sell "internet" today—you assume it exists and build services on top of it.

In the long term, the intelligence unlock is transformative. The ability to automate cognitive tasks that used to require human judgment? That's massive. But it also means the commodity technical work that many MSPs still rely on becomes even less defensible.

This is why I keep hammering on the same theme: MSPs need to move up the value chain now. When AI makes password resets and basic troubleshooting trivially easy, you need to already be positioned as a strategic advisor, not a help desk.

The technology is enabling that transition. But there's a window, and it's closing.

The Airline Lesson

So yes, I'm bullish on AI. I use it constantly. I think it will unlock enormous value across the economy.

But I don't want to run an AI company any more than I want to run an airline.

The best infrastructure is the kind you never think about. It just works, it's cheap, and it enables you to do more interesting things.

AI is going to become that. It's going to be great.

Just don't be surprised when the companies building it discover they've created a commodity business with airline economics.

What do you think? Are we heading for an AI winter, or is this time different? I'm curious where you see this playing out in your business.