Anthropic Growth Head: Only 2% of New Users Testing Pricing
Amol Avasare, Anthropic's Growth Head, stated that the pricing and feature changes related to Claude Code are currently being tested on a small scale among about 2% of new Pro users, with existing Pro and Max subscription users unaffected.
This clarification addresses previous concerns in the developer community about the "complete cancellation of low-cost AI programming capabilities." Feedback from English users and product observations indicate that Anthropic is conducting group experiments to test different pricing and feature combinations to assess the balance between user willingness to pay and computing costs.
Source: Public Information
ABAB AI Insight
This indicates that AI companies are entering a phase of "refined pricing experiments." Unlike the earlier unified subscriptions, platforms are beginning to dynamically adjust prices and feature boundaries through A/B testing to find the optimal balance between revenue and costs. This approach essentially treats users as real-time market signal sources rather than one-time pricing objects.
Testing only 2% of users shows that the company is managing public sentiment and revenue risks while obtaining real payment data. AI programming is a high-value scenario, and its price elasticity has not yet stabilized; companies need to gradually determine "how much users are willing to pay for productivity" through small-scale experiments.
From a competitive perspective, this also reflects that the industry has not yet formed a stable pricing consensus. On one side, OpenAI maintains low prices to expand its ecosystem, while on the other, Anthropic tests higher price ranges. These two paths essentially compete on different dimensions: the former focuses on scale and lock-in, while the latter focuses on unit economics and profit margins.
On a deeper level, this testing mechanism suggests that AI products will become increasingly "financialized," meaning that prices, features, and user tiers will be dynamically adjusted, similar to the advertising and cloud computing industries. In the future, the product forms users see may no longer be a unified version but rather a pricing system customized in real-time based on their behavior, value, and usage intensity.