Originally published in the Dakota Scout, reproduced here permission of Mr. LaFollette
Let me start with a confession: I’m a huge fan of artificial intelligence.
I use ChatGPT daily. I teach informal classes on it. And after a career spent at the intersection of accounting and technology — working with Thomson Reuters, the AICPA, and serving as editor of CPA Technology Advisor — I genuinely believe AI is one of the most powerful tools ordinary people have ever been given.
That’s why I want to say this now, while everyone is still smiling — AI is going to get a lot more expensive.
You’ve lived through Netflix, Amazon Prime, Uber, Microsoft Office and cell phone plans, so you already know how this story ends.
A familiar pattern
This is the “freemium” adoption model — the most powerful customer-acquisition engine modern tech has ever invented. You give people something incredibly useful for little or nothing. You let it become woven into their habits, their workflows, their thinking. Then, once dependence is established, you begin charging for comfort, convenience, performance, privacy and scale.
The pattern works in four steps:
- Be extremely affordable
- Capture the market
- Become a habit
- Raise prices
Here is the economic rule behind it all: Prices don’t follow costs. Prices follow dependency.
Every major technology platform of the past two decades has followed this script.
When Netflix launched streaming in 2007, it cost $7.99 a month — less than two movie rentals. Today its ad-free Standard plan is $17.99, Premium is $24.99, and even the ad-supported tier rose again in early 2025.
Uber and Lyft nearly gave rides away in their early years, subsidized by venture capital and the promise of future dominance. In 2024, the median U.S. ride-share fare rose more than 7 percent. Fares have climbed over the past several years well beyond what inflation alone would explain.
Cell phone service tells the same story. Early plans came with strict limits — a few hundred minutes, per-text fees, roaming charges and brutal overage penalties. Then came “unlimited.” Today that word means throttling, deprioritization, hotspot caps and tiered service levels. The technology improved. The pricing model matured.
Perhaps the clearest parallel is Microsoft Office. For decades you bought it once and used it until your computer died. Then Microsoft introduced Office 365. Now it’s $6.99–$9.99 a month for individuals, $99–$129 annually for families, and far more for businesses. The software didn’t suddenly cost more to build — but once it became infrastructure, Microsoft could shift from selling a product to renting access.
This isn’t villainy. It’s the business model.
Where AI stands today
Serious consumer AI currently costs about $20 a month — ChatGPT Plus, Microsoft Copilot Pro, Anthropic Claude, Google Gemini and similar offerings. Behind the scenes, the real money already flows through tiers and usage fees. Better models cost more. Heavy usage costs more. Business features cost considerably more. Free tiers exist to make you comfortable and capture market share. They are marketing, not charity.
Nearing conclusion on step two of the pattern — market capture — we’re already on our way to widely habitualizing AI.
Some might think, “But computing costs are falling, so AI prices should fall too.”
That misunderstands how pricing works in technology.
Yes, the cost per computation declines over time. But AI’s demands are enormous: training large models requires constant operation of thousands of specialized graphics processing electrical circuits known as GPUs.
Inference costs money with every query, and these expenses are real and persistent.
More importantly, once a product transitions from novelty to necessity, cost becomes irrelevant. When AI becomes baked into schoolwork, email, research, office life, customer service, scheduling, content creation, accounting and personal organization, it stops being a cool app and becomes infrastructure — and infrastructure always ends up priced like a utility.
Economists call it lock-in. Accountants call it switching cost. Consumers experience it as, “I can’t live without this anymore.” Whatever you call it, the result’s the same: pricing power shifts to the provider.
Some will argue AI is different: real competition exists, open-source models offer alternatives, regulators may intervene.
Sure, maybe AI is different. There’s real competition — OpenAI, Google, Anthropic, Meta are all fighting for users. Open-source models exist. Regulators might step in. Maybe. But based on historical patterns, we can make some predictions.
Over the next two years, clear consumer tiers could emerge — Basic, Plus, Pro, Business. Free versions would become noticeably limited. Serious users moving up the ladder often will not realize when choice becomes necessity.
By 2029, AI will become bundled into everything: office software, email, security, accounting, marketing, customer-relationship systems. Your total monthly tech bill will rise even if no single line item looks outrageous. Death by a thousand subscriptions — now with AI surcharges.
2030 and beyond
Ad-supported AI for the masses will be standard.
Those able to pay premiums will have clean, private, powerful AI.
At that point, no one will be debating whether AI is worth it. The only debate is which tier people can afford.
This creates an equity problem worth noting. When AI becomes essential for competitive work, quality education and efficient living, pricing tiers become access tiers. Those who can pay get better tools, faster responses, more privacy and fewer interruptions. Those who can’t get the limited, ad-supported version. We have seen this movie before — with internet access, quality devices and education technology. AI will amplify it.
None of this makes AI bad. It makes it important. And it makes today’s pricing phase exactly what every technologist eventually recognizes. It’s an introductory offer.
I’m still teaching people how to use AI. I still encourage everyone to learn it, experiment with it and build it into their workflows. The productivity gains are real. But I also want people to understand the moment we’re in.
The honeymoon pricing won’t last, the bills will arrive, and AI will still be worth it — just no longer cheap.
So yes, learn it now. Get comfortable with it. Build skills while the playing field is still relatively level. Understand where it genuinely helps and where it’s just impressive theater.
But don’t mistake today’s prices for tomorrow’s reality.
The cable bundle taught us that.
The cell phone “unlimited” plan taught us that.
The streaming wars taught us that.
Microsoft Office taught us that.
AI won’t be different. It will just be faster.
