Anthropic is having the best quarter in AI history and the worst customer service week imaginable, and both things are happening at the same time.

The Information reported this week that Anthropic has moved Claude Enterprise subscriptions from a flat fee of up to $200 per user per month to a usage-based model that charges for compute on top of a $20 monthly seat fee. Fredrik Filipsson, co-founder of Redress Compliance, told The Information the change could potentially triple costs for some enterprise customers.

The kicker: Anthropic says this change was first introduced in November 2025. Some customers only noticed it in recent weeks. An Anthropic spokesperson told Gizmodo that the pricing structure evolved in stages, with legacy plans now being forced onto the new model at their next renewal.

The Economics Made the Old Price Impossible

To be fair to Anthropic, the math on flat-rate AI pricing was never going to work. We wrote about this in April when Anthropic killed the flat rate for consumer AI agents. Claude Code and Claude Cowork are compute monsters. They run for extended periods, chewing through tokens at rates that make a $200 monthly cap look like a fire sale.

The company's own numbers prove the point. Weekly active users of Claude Code doubled between January and February. Revenue tripled to $30 billion annually. But compute costs scale with usage, and when your most popular products are the most expensive to run, you either raise prices or lose money on every power user.

Anthropic chose to raise prices. The new structure pools usage across organizations, so light users subsidize heavy users within the same org. On paper, that is a reasonable enterprise pricing model. In practice, CFOs are seeing bills that bear no resemblance to what they budgeted.

Then Came the Quality Complaints

The price hike landed at the worst possible time. Online communities are flooding with complaints that Claude's performance has degraded. Developers report longer response times, worse code quality, and more frequent refusals. Gizmodo noted "growing complaints online that Anthropic may have made recent tweaks to its models that worsened their performance."

Whether the model actually got worse is debatable. Opus 4.7 just launched with genuinely better benchmarks. But perception is reality in enterprise software. When your bill triples and the product feels worse on the same day, the narrative writes itself.

Anthropic Cannot Afford to Care

This is the contradiction at the heart of Anthropic's 2026. The company is worth $380 billion, preparing for an IPO, and burning through compute at a rate that requires constant revenue growth. It just signed a 3.5 gigawatt TPU deal with Google. It launched Claude Design, which crashed Figma's stock. It is winning the enterprise market for the first time.

None of that works on flat-rate pricing. Anthropic needs usage-based revenue to justify the infrastructure it is building. The company literally cannot afford to give enterprise customers unlimited tokens for $200 a month.

But the competitive landscape just got more dangerous. OpenAI launched a $100 ChatGPT Pro tier for coding. Cursor is raising $2 billion at a $50 billion valuation. Google gives away Gemini access on Enterprise Workspace plans. Every alternative is pricing aggressively while Anthropic is pricing up.

The Bet

Anthropic is betting that its product is so much better than the alternatives that enterprise customers will pay triple and complain about it on Reddit but never actually leave. Based on the Q1 enterprise spending data, that bet is working. Anthropic overtook OpenAI in enterprise AI spend for the first time this quarter.

The question is whether a company can maintain that position while simultaneously raising prices and fielding quality complaints. Apple did it for years with the iPhone. But Apple never had five well-funded competitors releasing comparable products every month.

First reported by The Information. Additional reporting by Gizmodo.