OpenAI is winning coding customers from Anthropic with deep discounts, report says

What was reported

OpenAI is winning business customers away from Anthropic by undercutting it on price, according to a report by The Information published on 24 October 2026 under the headline Anthropic's Margin is OpenAI's Opportunity. The piece, by Laura Bratton and Stephanie Palazzolo, is paywalled, and AI Lookout has not read the original. This article relies on detailed summaries from several outlets that did.

The figures quoted are striking. The Decoder's summary headlines discounts of up to 90 percent and names Rippling, Replit and DataCamp as companies shifting more of their AI spending to OpenAI. The common thread is coding: the work that made Anthropic's Claude models the default for many developer tools over the past year is now where OpenAI is pushing hardest on price.

Anthropic has not commented publicly on the specific deals. AI Chat Daily notes that the reports did not include a statement from Anthropic on the Replit contract or the Rippling figures, so the numbers below are those of customers and unnamed sources as relayed by The Information.

Rippling flipped its coding work

The clearest example is Rippling, the HR software company. Its chief executive, Parker Conrad, told The Information that in May, Anthropic models handled 57 percent of the coding work done by hundreds of Rippling engineers, while OpenAI models handled 16 percent. A Subvolts summary of the report says the split has since reversed, with OpenAI at 52 percent and Anthropic at 18 percent.

Conrad gave cost as the reason. Rippling wanted to cut spending, and OpenAI proved cheaper, he said. The summaries do not say where the rest of the work goes.

That swing happened in roughly five months, which says a lot about how easily coding workloads can move. Many coding agents let teams switch the underlying model with a setting, so the main lock-in is habit and the quality of results, not a long integration project. When the quality gap narrows, price decides.

Replit and a $100 million contract

Replit, the browser-based app builder, is the biggest deal in the report. According to AI Chat Daily's account, a source told The Information that OpenAI offered Replit more than 30 percent off a contract worth over $100 million a year, and Replit moved most of its AI spending from Anthropic to OpenAI. Anthropic had declined to extend a discount Replit previously received unless Replit committed to spending more.

This fills in the background to a public back-and-forth earlier in the month. On 16 October, Replit chief executive Amjad Masad said on stage that the company was no longer working with Anthropic, citing cost, the partnership and performance, and then clarified that Replit had made OpenAI its default model but kept Claude. Replit says it remains model agnostic, and its documentation still lists Claude models that customers can select.

The default is what matters commercially. AI Chat Daily points out that most Replit customers keep the default setting, and that Replit had recently been one of Anthropic's five largest customers. Masad also said enterprise contracts grew from about 2 percent of Replit's revenue last year to about 15 percent this year, so the cost of the models underneath is becoming a bigger part of its margins.

DataCamp and a bank

The report's other examples are less detailed and the summaries do not fully agree. The Decoder says DataCamp chief executive Jonathan Cornelissen told The Information that OpenAI's pricing cut his costs by 80 to 90 percent. AI Chat Daily instead says OpenAI offered DataCamp a 40 percent discount to use Codex, citing coverage of the report. Both could be true if the larger saving comes from moving to cheaper models as well as the discount, but the summaries do not say.

The Decoder also describes a US bank with more than $7 billion in assets that had its ChatGPT seats doubled to 500 employees at no extra charge. That works out to less than $7 per user, compared with the usual $20 or more. The bank is not named.

Together, the examples show OpenAI using different levers for different buyers: contract discounts for high-volume API customers, cheaper models for cost-sensitive teams, and free extra seats to deepen use of ChatGPT inside companies.

How the discount terms differ

The Information has been reporting on contract terms for weeks. A 16 October summary by The Clarity of an earlier report says software managers put Anthropic's enterprise discount at roughly 15 percent off list price, and that it ends once a customer uses up the tokens in its contract. At that point the customer must renegotiate or pay list price, about 18 percent more per token.

OpenAI is described as more lenient. According to a licensing adviser quoted in that report, it gives customers the rest of the current month plus one more month to agree a new contract before list prices apply. The report also says both labs have started adding share-of-wallet clauses to discount deals, which require large customers to send most of their AI budget to one platform. CodeRabbit chief executive Harjot Gill said OpenAI had become his main AI supplier, where six months earlier it was Anthropic.

Why OpenAI can afford it

The Decoder argues that OpenAI is partly buying market share at the expense of its own margins, but that it has also achieved genuine price cuts through its Sol and Luna model lines, thanks to more efficient models and infrastructure. Lower list prices plus targeted discounts make a hard combination to match.

Anthropic is not standing still on pricing, but its recent moves have gone in both directions. The Clarity summary mentions a cheaper Opus model, while fast mode for Claude Opus 5.5 left research preview at $8 and $40 per million tokens, double the standard rate, a premium aimed at customers who value speed over cost.

Both companies also face pressure from China. The Decoder notes that new models from DeepSeek and Kimi deliver comparable performance for a fraction of the cost, which caps how much either American lab can charge for routine coding work.

What it means for Anthropic and for buyers

Anthropic is in a strong position overall. It recently told investors that third-quarter revenue reached $16.8 billion ahead of a planned listing. But The Decoder cites an investor memo saying its growth slowed from the second quarter to the third, and The Clarity summary says Exponential View, citing leaked investor figures, put its adjusted gross margin at 85 percent. That gap between price and cost is exactly what OpenAI is attacking.

For buyers, the lesson is that list prices are only a starting point. Large customers are getting very different terms, and the threat of moving work to a rival appears to be the main source of leverage. Teams that keep their tools model-agnostic, as Replit and Rippling did, can capture those savings. Teams locked into one provider through share-of-wallet clauses may find that the discount comes with less freedom later.

The caveat is sourcing. The figures come from a single paywalled report and from executives with an interest in being seen to drive hard bargains. Until OpenAI or Anthropic discuss pricing in their own filings, the scale of the discounting remains a reported claim rather than a confirmed fact.