For products with power users & high marginal costs
For products with power users and high marginal costs
One mistake/untenable competitive equilibrium that hot AI startups like Cursor, Anthropic have recently been running is an unlimited usage subscription model deeply unaligned with their underlying business. It’s a fixed revenue, highly variable costs (which are mostly deeply negative). Their most profitable users (i.e. those that barely use the service) churn first. Their power users make tokens go burrrrrr for no additional cost to themselves.
Despite hyper around models getting cheaper over time, users switch instantly to the latest most capable model. So token prices haven’t fallen at all in years:
Meanwhile, number of tokens per request has gone vertical with inference-time compute:
Revenue per user: constant.
Costs: exponential.
The natural business model-product-fit for many high-usage, compute-heavy AI products is usage-based.
https://ethanding.substack.com/p/ai-subscriptions-get-short-squeezed
https://ethanding.substack.com/p/windsurf-gets-margin-called
https://hypersoren.xyz/posts/smart-squeeze/
https://www.michaeldempsey.me/blog/2025/10/03/sequencing-vs-equal-odds-applied-research/
