I made a video on this because I keep seeing the question everywhere: is $200 a month on Claude Code too expensive? I am on the plan, so I pulled my actual usage. Here is the gist and my take, with the full cost math so you do not have to watch to get it.
The honest comparison
In the past 30 days I pushed about 5 billion tokens through the $200 Max plan. At pay-as-you-go API rates that same usage runs over $4,500. You are paying $200 for something that would cost more than twenty times as much a la carte. So for a heavy user, worth it is not really the question.
The reason it is this cheap is that these companies are in a land grab. OpenAI is projected to lose $44 billion between 2023 and 2028, and Anthropic is burning through about $3 billion this year. It is the Uber playbook, back when a ride was five dollars because VC money paid the difference.
The catch: you have to actually use it
The Max tier only pays off if your usage is real. Anthropic also sells $20 and $100 plans, and for most people those fit better. On the $20 plan you hit limits before lunch. On the $100 plan I once hit the wall mid-day and had to wait three hours for a reset, which is what pushed me to Max. Paying for Max and touching a sliver of it is how you overpay, which is the opposite of the point.
How I would actually play it
Tokens are a vanity number on their own. The test is what you ship. Here is the plan I ran and would run again:
- 1. Get the $200 plan, not the cheaper ones. You cannot learn what this thing does if you are constantly rationing.
- 2. For 30 days, use it for everything. Code review, docs, research, testing, content, not just code.
- 3. After 30 days you will know which three to five workflows actually produce value. Keep those.
- 4. Optimize those for quality, not volume. Aim for output that would impress someone who knows the domain.
- 5. Point that output at revenue. A product, a service, content, consulting, whatever your thing is.
I built learn.cloudyeti.io, briefly.givingback.ai, and earnyeti.com with Claude Code, including the automated system that produced the video itself. I judge the plan by the return on those, not the token counter.
Why the price won't last
Nick Turley, who runs ChatGPT at OpenAI, has said the current pricing was accidental and there is no world where it does not change a lot. Axios ran a headline in March that AI may never be as cheap as it is today. Jensen Huang said at GTC that engineers should get token budgets worth half their salary, and some firms are already adding six-figure token budgets on top of pay. Use this window to build durable workflows, and do not assume the number stays this generous.
Work with me
I specialize in helping teams choose the right AI plans and strip out spend that is not earning its keep. I spent 6 years at Amazon before going independent. If you want a straight read on which tier fits, book a call at cloudyeti.io/meet.
Book a call