Because it's kind of true, but also mostly wrong. People conflate two things: "Subsidized" subscriptions as a monetization model, and (unsustainable) subsidization through outside investors.

When you max out your Claude Max plan every billing period, are you consuming more tokens than you could realistically pay for with the same money spent on API inference? Definitely! This works because many Claude Max subscribers don't use it to the max (and you probably aren't using it to the max every single time either), so overall subscriptions might still be profitable. Technically, the subscribers subsidize each other, but this is entirely sustainable (see any other subscription model that is based on its consumers not maximizing their usage, e.g. gyms).

What people mean (or want to say) when they say inference is subsidized is that it is subsidized in an unsustainable way (e.g. with investments that will not last forever - if this is true at scale it should worry us!). What happens now is that they take the fact that an individual subscription used to the max by itself is a financial drain for the company offering it, attach the "subsidized" term to it, and infer from that a bunch of stuff that is usually associated with the term (such as unsustainability) but doesn't really apply here.

Because the subscriptions grant you much more usage than what you get in tokens for the same money via the API. That being said, they probably don't have ruinous losses per subscription, if any. Don't forget that many users are casual and do not use Claude Code, so a casual user and a power user level out

When you say that something costs $10k worth of tokens at API pricing, it sounds impressive.

Some people then assume that this price means the subscriptions are deeply unprofitable at least for users that make heavy use of their usage limit.

If you look at estimates based on serving Kimi K3, margins at API pricing must be well north of 95%.

And that's assuming frontier labs don't serve inference more efficiently than Kimi K3.

I believe most people dramatically overestimate the inference costs.

It doesn't help that in many discussions about profitability, people mix many different metrics, including other costs like R&D.

Because they are, I pay anthropic 200 a month, if I were paying by the token like an enterprise it would be close to 11 grand a month. Yes I use 100 percent of my allocation on every window.