Anthropic Moves to Cut Off Discounts When Customers Hit Their Cap

Anthropic is flexing its muscle with enterprise customers by taking a hard line on discounts once those customers use up all the tokens they purchased. That’s created an opening for OpenAI to take a more flexible approach.

While both Anthropic and archrival OpenAI provide discounts to customers that commit to spending millions of dollars a year on their products, Anthropic has taken the unusual step of ending customers’ discounts once they hit the usage limits in their contracts, according to managers from three software firms that buy AI from both companies. Customers that reach the cap must negotiate a new agreement or face higher prices, these people said.

“The second you hit the cap, you get phone calls and emails from Anthropic sales teams, saying ‘Hey, we noticed you hit your cap—if we address this today with a new agreement, you can avoid overages,’” one of these people said. “Otherwise you’re on your own.”

Amazon, Microsoft, and Google typically allow discounted pricing to remain in effect for spending that exceeds the committed level during the term of the agreement, according to Jeff Muscarella, chief innovation officer at NPI Financial, a company that helps Fortune 500 companies negotiate software agreements

OpenAI, too, seems to take a softer touch than Anthropic with enterprise customers, according to software licensing consultants who work with both companies’ customers. For instance, when its customers reach their spending commitment in discounted deals, OpenAI gives them the rest of the calendar month, plus an extra month, to negotiate a new deal before reverting to publicly listed prices, said a person close to OpenAI.

A spokesperson for Anthropic did not have a comment.

Anthropic’s discounts on its models can be around 15% off listed prices, according to managers from the three software firms that buy AI from both companies. That’s a meaningful markdown for the more than 100 firms that spent more than $10 million each with Anthropic in the 12 months ending in June, and the more than 1,000 firms that spent more than $1 million apiece during that period.

Among Anthropic’s biggest customers are firms such as Meta Platforms and Cursor, which have used Anthropic models for internal uses or to power their products. Many other enterprises pay for Claude Code or Claude Cowork, both of which have gotten more expensive as Anthropic has changed their pricing. (Anthropic told at least one business partner it was generating a gross profit margin of around 75% on the sale of its existing models.)

In recent months, OpenAI has been particularly focused on catching up to Anthropic, which zoomed far ahead of OpenAI in revenue this year. OpenAI has publicly cut the price of new models 50% for customers that access them through intermediaries such as OpenRouter and Vercel. When OpenAI on Tuesday launched its latest AI models, CEO Sam Altman tweeted that they are “half the price per token, and even less per task” compared to their predecessors.

“Right now, OpenAI is trying to win enterprise business and [is] more aggressive than Anthropic in discounting,” said Fredrik Filipsson, co-founder of Redress Compliance, a firm that helps companies negotiate software licensing agreements.

The quality of OpenAI’s models has also improved, lending weight to its efforts with large AI buyers like CodeRabbit, which sells a code-review service. Six months ago, CodeRabbit CEO Harjot Gill said Anthropic was the majority supplier of models powering the service. Now OpenAI, which has shown a keener interest in discounting its models than in the past, is CodeRabbit’s main supplier, he said. He declined to discuss the discount but said his three-year-old company spends tens of millions of dollars a year on AI.

The moves come at a delicate time for OpenAI’s sales team after the executive who spearheaded the AI firm’s enterprise push abruptly left in August. The unit is undergoing big changes. OpenAI is merging separate roles that previously focused on working with customers before and after signing contracts. Now the same people who create demonstrations of OpenAI products for customers must also act as so-called customer success managers after contracts are finalized, according to an employee.

As competition heats up, OpenAI and Anthropic are now adding language to their discounting agreements that asks large customers to direct a majority of their AI spending to them, known as a “share of wallet” clause, said Gill and a manager from another software provider that negotiates contracts with the AI labs.

Such language is symbolic, however: OpenAI and Anthropic have no way to verify that customers are doing so, according to the licensing consultants.

It isn’t clear whether OpenAI’s softer tactics are slowing Anthropic down, though OpenRouter said the discounting prompted more of its customers to pick OpenAI models for their work. Anthropic expanded its revenue lead over OpenAI between May and August, though OpenAI continues growing quickly relative to many other businesses.

As paranoia rises over whether the two AI leaders could learn from their customers’ intellectual property, large firms involved in sensitive corporate work, such as Palantir Technologies, Nvidia and Booz Allen Hamilton, have started demanding new guarantees or reducing or eliminating use of the AI firms’ most advanced models. Still, the two firms offer the best-performing models for automating white-collar work and research, according to a variety of evaluations, making them hard to ignore for businesses that want to automate customer service, legal work and other tasks—or to develop and sell AI apps that perform such work.