Anthropic changes pricing for Claude as the era of unlimited AI employees ends, enterprise AI costs could triple

An era of “unlimited AI employees” built on Anthropic’s Claude model lasted only about three months. Anthropic has changed its pricing structure for enterprise customers, and for teams using Claude as a full-time unlimited employee, invoices can rise by as much as three times.

What changed in Anthropic’s enterprise pricing

Under the new model, Anthropic reduces the monthly seat fee for technical users such as developers to $20 per month, down from a prior range of $40 to $200.

However, the structure shifts the risk to customers. Enterprise customers must commit to and prepay an estimated monthly token consumption. If actual usage is lower than the commitment, customers still pay the full prepaid amount.

The token price remains, and Anthropic also removed traditional volume discounts. As a result, overall costs can increase even with the lower seat fee.

Example of how costs can rise

In one example, a 10-person software development team using Claude as a constant AI coworker previously paid a fixed amount for running Claude. Under the new approach, the team must pay a $200 monthly seat cost plus monthly anticipated token usage, prepaying the total.

If the forecast is too high, the team wastes money; if too low, it pays more later. The team also loses 10% to 15% discounts it previously enjoyed. Overall costs, according to the example, can triple.

Who is most affected

The change is particularly damaging for projects with variable usage, including seasonal or peak-demand workloads. Because customers must prepay a token estimate, fluctuations can lead to either overpayment or additional charges later.

Usage limits adjusted during peak hours

Anthropic also adjusted usage limits by reducing peak-hour service performance. A five-hour session limit applies during peak hours from 5:00 to 11:00 a.m. Pacific Time, which can mean users exhaust their five-hour session in less than five real hours.

At other times, a five-hour session can accomplish more tasks. The content states that about 7% of users will hit the session limit they previously did not.

Strategic rationale and likely customer response

From a strategic perspective, the pricing shift is described as optimizing for predictable revenue at the expense of customer flexibility. The change is expected to push some customers to consider alternative providers with more transparent volume-based discounts, or to optimize workflows to reduce token consumption.

A tech community expert quoted in the material said the “unlimited AI employees” period lasted only about three months, with heavy users seeing bills triple overnight. The expert also characterized the change as the end of the “free lunch.”

Recommended actions for procurement and engineering teams

The article outlines immediate steps for procurement and engineering teams, including:

• Recalculating total cost using a monthly token profile rather than seat counts.
• Seeking contract protections such as carryover credits and audit rights.
• Optimizing prompts and caches to reduce token spending if commitments are unavoidable.