American customers of Anthropic are turning to lower-cost alternatives, abandoning its most powerful AI tools ahead of the company's anticipated largest IPO in history, raising questions about its high-cost model. According to payment group Ramp's spending data on 70,000 companies, cited by the Financial Times, two months after the launch of the most expensive model, Fable 5, spending on it accounted for only about 11% of total spending on Anthropic tools and has leveled off, breaking the convention of users defaulting to the strongest model.
High praise but poor sales, OpenAI takes advantage to rebound
Analysts and investors believe the main reason is that Fable is priced high and older models already meet most needs. Clements, a partner at Accel, said "most people don't need the most cutting-edge model," and the era of choosing only the most advanced model "is not sustainable." Fable 5 was once withdrawn in June due to national security intervention by the Trump administration government, and was approved to relaunch on July 1st. Although political uncertainty has become a secondary factor, slow adoption still adds variables to the IPO - the market expects its valuation could reach $2 trillion and it may list as early as next month.
Cost-effective open-weight models in regions such as China give customers more choices, no longer relying solely on US leading vendors; OpenAI has also recovered ground with GPT 5.6 released in July (priced significantly lower than Fable), with annualized revenue jumping 35% to over $40 billion this quarter. Its smaller Opus 5 also saw enterprise spending surpass Fable since late July due to its low price. Although Anthropic had an annualized revenue of $65 billion in July, which has nearly seven times since the beginning of the year and recorded its first operating profit in the second quarter, Ramp economists said that under the current situation where OpenAI's new model performs well and Fable's performance is average, predicting its future trend is almost impossible.
