Claude Max 20x costs $200 per month. In the United States, that's an expensive but defensible line item for a professional who uses it daily. In many other countries, it's roughly 70% of a monthly minimum wage — a sum that makes the decision to subscribe feel genuinely reckless for anyone who doesn't have a specific, high-demand use case that pays for itself.
That's the friction point that matters most as AI tools move from novelty into normal business infrastructure. The question isn't whether heavy users will keep paying. They will. Someone whose entire output depends on Claude's ability to handle complex, long-context work — the kind that Pro's $20 tier can't sustain — is already locked in by necessity. The question is everyone else.
The Middle Tier Has the Retention Problem
Anthropics's pricing ladder runs from Free to Pro ($20) to Max ($100 or $200 depending on usage multiplier). The gap between Pro and the lower Max tier is $80 per month. For users in markets where wages are denominated in weaker currencies, even Pro at $20 represents a meaningful discretionary expense. Max is out of reach for most.
This creates a segmentation that Anthropic probably didn't design intentionally but has to live with: an affluent core of heavy users in high-income markets who pay the top rates, and a much larger global population priced out of anything beyond the free tier or basic Pro. The people in the middle — those who would pay for more access if the value justified it — are exactly the retention question.
ChatGPT faces the same problem. Google Gemini Advanced does too. None of the major AI labs have introduced meaningful regional pricing, the way Spotify and Netflix have long done to capture subscribers in lower-income markets.
What Actually Drives Retention
The "honeymoon" framing — where users pay for a few months out of curiosity, then cancel when the novelty fades — misses the real driver of churn. People don't stop paying because AI tools became boring. They stop paying when they can't point to a specific, recurring task that would take significantly longer or produce worse results without the subscription.
For a content creator who uses Claude daily to draft, edit, and restructure long-form work, $20 per month is easy to justify. For someone who tested it for two weeks and uses it occasionally to answer questions, $20 is a subscription that keeps getting deprioritized in a budget review.
The durability of AI subscriptions tracks closely with specificity of use. Vague "I use it for various things" subscribers churn. "I use it every morning to process my client feedback and generate first drafts" subscribers stay. That pattern holds regardless of price tier.
What it doesn't account for is access. A user in Brazil or India who has the high-specificity use case and would happily pay might still find that the dollar-denominated price simply doesn't fit their income reality. Those users aren't churning because the value isn't there — they're churning, or never subscribing in the first place, because global pricing hasn't caught up to global demand.