Edited by Paul Derrington, Chief Editor
TL;DR
The Information reported on October 24 (“Anthropic’s Margin is OpenAI’s Opportunity”) that OpenAI has been offering large discounts to business customers who use AI models to write software, and that some have shifted work from Anthropic. According to the report, Replit moved most of its AI spending to OpenAI on a contract worth more than $100 million a year after a discount above 30 percent, and at Rippling OpenAI’s share of coding work rose from 16 to 52 percent while Anthropic’s fell from 57 to 18 percent. Neither OpenAI nor Anthropic has confirmed the reported figures, and these are individual customers’ deals and estimates, not a published price list.

What the report says
The Information reported on October 24 that OpenAI has been offering large discounts to business customers who use AI models to write software, in a piece titled “Anthropic’s Margin is OpenAI’s Opportunity.” The report is paywalled; this briefing relies on The Information’s public excerpt syndicated by Benzinga and on summaries by The Decoder, Techpresso, AI Weekly and others. Those summaries attribute the figures to customer executives and a person familiar with the deals. Neither OpenAI nor Anthropic has confirmed them publicly, and no company statement on the specific offers was found.
This is a follow-up with new numbers, not a repeat of the mid-October split: this briefing covered Replit chief executive Amjad Masad saying onstage at TechCrunch Disrupt that Replit was no longer working with Anthropic. What is new here is what the switch reportedly cost, and which other accounts reportedly shifted work.
The accounts that moved
Replit is the largest named deal. According to the report as summarized by The Decoder and AI Weekly, it shifted most of its AI spending from Anthropic to OpenAI under an annual contract worth more than $100 million, after OpenAI offered more than 30 percent off. Neither company has published the terms, and the summaries do not say whether the $100 million is measured before or after the discount. AI Weekly reports that Anthropic had declined to extend Replit’s earlier discount unless it committed to higher spending. The split itself is more nuanced than a headline suggests. TechCrunch reported that Masad later said Replit “remains model agnostic”, that users can still choose Claude, and that OpenAI is now its preferred partner for Replit Agent and the default; a person familiar with the Replit-Anthropic relationship told TechCrunch the partnership was still in place.
Rippling, the HR software company, reported a reversal in five months. Per The Information’s excerpt, chief executive Parker Conrad said Anthropic models handled 57 percent of the coding work of his hundreds of engineers in May and OpenAI models handled 16 percent; today OpenAI handles 52 percent and Anthropic 18 percent. Each pair is a share of Rippling’s coding work (the shares do not add to 100 percent, and the report does not say who handles the rest), not a market-wide figure. Conrad said the company sought to cut costs and OpenAI proved cheaper. DataCamp, an online-learning company, was offered a 40 percent discount, reportedly on models for its internal product development and on the condition it use OpenAI’s Codex coding agent. Chief executive Jonathan Cornelissen put the possible cost reduction at 80 to 90 percent. Outlets differ on what that figure means: The Decoder and Techpresso say OpenAI’s pricing cut his costs by that much, while ALFW and others say he estimated costs could fall that far if the 40 percent terms were extended to the models inside the products DataCamp sells. The paywalled report could not be checked, so treat the figure as an executive’s estimate. The report also describes a US bank with more than $7 billion in assets whose ChatGPT seats were doubled to 500 employees at no extra charge, so it pays under $7 per user against a usual $20 or more; the bank is not named.
Why price may matter
Techpresso reads the story as meaning that when two frontier models are close in quality, price decides the contract; the report itself does not establish that. According to The Decoder’s summary, OpenAI is partly buying market share at the expense of its own margins, and it has also cut prices through its Sol and Luna model lines, which The Decoder says are built on more efficient models and infrastructure. The Decoder is the only source for that last point here, and it was not independently checked.
The Decoder, citing an investor memo, says Anthropic’s growth slowed from the second quarter to the third, and other outlets report that Anthropic is preparing a public listing. Both labs also face price pressure from below, with lower-cost Chinese models from DeepSeek and Kimi delivering comparable performance for a fraction of the cost, according to The Decoder.
Read the numbers carefully
Some headlines say “up to 90 percent”; that is not a price list. DataCamp’s 80 to 90 percent is one executive’s figure (reported as a cost cut by some outlets and as an estimate by others), the only discount reported as granted there is 40 percent, and the Replit and Rippling figures are a negotiated contract and workload shares, not published rates. What the reporting describes is three named customers shifting some or most coding work toward OpenAI, with cost cited as a reason by Rippling and Replit.
What to watch: whether Anthropic matches on price to hold its remaining large accounts, and whether OpenAI's discounts survive its own margin pressure. If you buy AI coding tools for a team, one practical takeaway is to compare quotes from both vendors before renewing; the reporting describes large customers negotiating on price.