Ep. 50: The AI Price Hike Is Coming
This entire episode is dedicated to discussing a prediction I’ve brought up eleventy billion times: AI prices are gonna go up, and it’s not a matter of if, it’s a matter of when. I break down Microsoft Copilot Cowork’s usage-based pricing rollout, what it signals for the rest of the industry, the concept of tokenmaxxing, and how Anthropic’s existing usage limits and Fable 5’s pay-per-usage model are early signs of what’s ahead.
We celebrating!
We did it, fam! Episode 50! Let’s take a quick moment to celebrate, because it’s dope, I’m grateful, and it’s important to celebrate your wins.
Thank you for being here, thank you for reading, thank you for responding, thank you for listening to the podcast if you tune in. Just thank you for being on this journey with me.
It’s pretty wild to see how much has changed with AI since episode 1 dropped on August 7th last year. The AI advances are slowing in some regards while AI capability is rapidly increasing, and I really do believe that this tech is here to stay. So here’s to another 50 episodes..
Speculation, Extrapolation, and Common Sense
My thoughts surrounding an AI price hike are more speculation, extrapolation, and common sense than it is some sort of insider knowledge, but I’ve brought this up like eleventy billion times before on the podcast, and I’m gonna bring it up again today and dedicate the entire episode to it, because it’s not a matter of if, it’s a matter of when.
This episode, and my most recent speculations about an impending AI price hike, were fully inspired by Microsoft’s recent rollout of Copilot Cowork to the general public. They announced that users would be billed for Cowork on a usage-based basis, meaning charges will be determined by the tasks that are run.
For those who don’t know, Copilot is Microsoft’s LLM, and Cowork is their agentic version of the model, which means it can actually execute tasks for you, not just answer questions. It runs in the cloud, as opposed to locally on your computer, and operates within the Microsoft ecosystem.
This usage-based pricing is a huge shift from the predictable flat-fee-per-month pricing model utilized by every AI company. To clarify, regarding Microsoft’s Copilot Cowork specifically: users will pay a flat subscription fee to have access to Cowork, and then on top of that they will be billed based on their usage of Cowork.
My Prediction
My prediction is that the rest of the companies follow suit.
There’s already evidence pointing that way. Anthropic already has usage tiers and usage limits (which I support), and for a very brief stint there was talk about them removing Claude Code from the Pro tier all together (the lowest of their paid subscriptions).
Additionally, if you’re thinking that the name “Cowork” sounds familiar, it’s because it’s exactly what Anthropic calls their agentic model created for the everyday user. Coincidence? I think not. Microsoft worked with Anthropic to create Copilot Cowork and integrated the technology used for Claude Cowork.
Real quick, speaking of Anthropic, Fable 5 is still shut down with no reports on if or when it will be live again. But, relating to all this price hike talk, when Anthropic released Fable 5 they announced it would be pay-per-usage billing following that two-ish week trial period where, like a good dealer, they were going to let everyone get a taste.
So, back to Microsoft an co., once one company sees another company doing something, especially when those companies are already working together, it feels inevitable to me that the other companies will follow suit. When exactly will the other companies switch to this usage-based model? Again, I have no idea. This is all just a compilation of speculation, extrapolation, and common sense. But it’s absolutely coming.
The Bros Were Tokenmaxxing
Allow me to introduce you to another term that further speaks to this impending pricing-model change: tokenmaxxing.
Tokenmaxxing is the word used to describe the most tech bro way of using AI: burning as many tokens as possible per interaction with AI…just because.
Taking it back to the way early episodes, here’s a quick reminder of what tokens actually are:
- Words are broken down into sequences of characters referred to as tokens
- Tokens also serve as the billing unit of AI models, with usage being measured in tokens
Thankfully this very brochacho approach to using AI, is officially coming to an end.
The biggest offenders of tokenmaxxing seemed to be enterprise users, meaning people whose companies paid for their AI access and usage. Many of these companies gamified usage, pushing their teams to use the best models and move the fastest, costs be damned.
CEOs started singing a different tune this year as all-you-can-eat pricing was replaced by token-based pricing, which, paired with the advent of more advanced token-hungry models, had companies like Uber burning through their entire 2026 AI coding budget by April.
NGL, feels like a bit of poetic justice that Uber got Ubered.
The Approach I’ve Always Recommended
So what we’re seeing now is these companies doing what they should have done in the first place: encouraging employees to be mindful about usage. No shit.
Honestly, it’s just a good approach to life: why use more than you need?
Additionally, the most basic AI model that an LLM has is very likely more than sufficient for your needs. As a Claude user, Sonnet 4.6 is both my daily driver and what I use for vibe coding. I’ve never felt the need to use the token-hungry Opus model, and I don’t want to speak for you, but my guess is that it would be a bit like having a legit military-grade Hummer when all you do is drive to the mall. Yes, I’m saying this as someone who owns a Jeep Wrangler. FWIW it’s the most basic model! 😉
Would You Pay More?
The AI price hike is coming, but I gotta say that I’m not opposed to paying more. I find the tool incredibly useful, and for all that it does, especially when comparing it to other software that I use for my business, paying $100/month wouldn’t be unreasonable.
Do I enjoy paying less? Absolutely. Do I absofuckinglutely hate this bait and switch approach? With all my heart. Do I have plans to learn about the open weight models and try to future-proof myself? You bet.
How bout you? Would you be willing to pay more to access your favorite LLM If so, how much more? I’d really love to hear your two pennies about all of this and if you’d be willing to pay more for AI. Hit reply and fill me in. I’m genuinely interested your thoughts.
How I Used AI This Week
Each episode I share a quick example of how I or someone I know used AI that week.
This week I’m talking about my guy John, who goes by JCode, and he actually used AI, more specifically ChatGPT, to save his Mac.
JCode got himself in a bit of a pickle trying to use an external hard drive as an internal hard drive, and long story short, everyone at the Genius Bar told him there was no hope. Everything was corrupted and his only option was to wipe it all, and he would lose everything.
My guy teamed up with ChatGPT, spent a few hours in the computer (like, a good amount), fixed everything, didn’t lose anything, and his Mac is fully functional.
He shared this story on IG and gave me permission to use it here, and while your mileage may vary, this is a damn good use case.
Da Wrap-up
The AI price hike is coming. When exactly? Who knows. AI money continues to be funny, VC funding is drying up, and valuations for these AI companies are artificially inflated and sprinting towards a bubble.
So while I can’t say when the price hike will happen, it’s definitely going to happen, so why not do your best to “get ahead” of it right now in any way you can.
As always, endlessly grateful for you and your curiosity.
Catch you next Thursday.
Maestro out.