$200 subscription vs $7,470 of API usage

Some context: My previous plan was hitting limits too fast, so I decided to subscribe to Anthropic's max plan ($200/month) for a single month, specifically to test newest models and build as many apps as I could in that window. Along the way I put together a small script that scans my project folders, pulls this month's sessions, deduplicates them, and totals the tokens. I'm not estimating anything, the API returns exact token counts on every response and Claude Code writes them to the session logs. The script just adds those up and applies the published rates. I also hit the usage limit several times along the way but what is important is the outcome of that script: the same usage would have cost $7,470 on the API but I paid $200 . Questions I have: Is Anthropic simply eating a $7.2k gap on a single heavy user? Or is API list pricing so far above their real serving cost that the gap is nowhere near $7.2k in the first place? And a separate one: is this a move to gather users now and shift to an API-driven model later, or is it sustainable as it is? I genuinely don't know which of these dominates, and the answer changes whether subscriptions look sustainable or look like a phase. And if there's a flaw in how I'm reasoning about this, say so. Maybe I'm looking through the wrong lens here.