OpenAI Cuts Enterprise Coding Prices to Win Anthropic Accounts Before November IPO

The market for enterprise coding assistants is now being decided partly at the negotiating table rather than only in model benchmarks. Large software companies, banks and training providers are comparing Claude, ChatGPT and Codex on price, default settings and procurement terms as much as on raw model quality. That shift has turned ordinary contract renewals into strategic contests between OpenAI and Anthropic.

OpenAI has moved aggressively to win those contests. The Information reported on October 24, 2026 that OpenAI is converting some of Anthropic's largest coding customers by undercutting Claude on price and accepting lower margins to take share before Anthropic's planned November IPO. The report, headlined Anthropic's Margin is OpenAI's Opportunity, describes a deliberate campaign rather than a set of isolated discounts.

The most visible case is Replit. According to The Information, OpenAI offered more than 30 percent off a contract worth more than $100 million a year after Anthropic declined to extend Replit's previous discount unless the coding platform committed to higher spending. Replit moved most of its AI budget from Anthropic to OpenAI in recent months, and the change required only hours of code work and a few days of product level adjustments.

Rippling shows how quickly usage can follow price. CEO Parker Conrad said Anthropic's models handled 57 percent of its engineers' coding work in May 2026 while OpenAI's handled 16 percent. Those shares have since reversed, with OpenAI at 52 percent and Anthropic at 18 percent. Conrad attributed the shift to cost, making Rippling one of the clearest public examples of a large enterprise account changing its default model mix.

Key Facts

The Information reported on October 24, 2026 that the Replit contract was worth more than $100 million a year and that OpenAI's discount exceeded 30 percent. The report says Anthropic had refused to keep Replit's earlier discount without a higher spending commitment, a decision that opened the door for OpenAI. Replit CEO and cofounder Amjad Masad told TechCrunch Disrupt in San Francisco on October 16 that Replit is no longer working with Anthropic, describing the move as a decision based on cost, partnership and performance.

The split is not absolute. AI Weekly reported on October 24, 2026 that a person familiar with the deal said the Anthropic partnership is still in place and that Replit's documentation still lists seven Claude models. OpenAI is now the default model provider for Replit Agent, so customers who want Claude must select it. AI Weekly noted that default settings carry outsized commercial weight because most users keep defaults, which means Anthropic can lose bulk volume even while Claude remains available.

At Rippling, the numbers are starker. The Information reported on October 24, 2026 that OpenAI's share of coding work rose from 16 percent in May 2026 to 52 percent, while Anthropic's fell from 57 percent to 18 percent. The Decoder reported on October 25, 2026 that Rippling CEO Parker Conrad linked the reversal to cost and that the discount strategy is pulling spending toward OpenAI's Sol and Luna models.

DataCamp provides another example. Sina Finance reported on October 25, 2026 that the enterprise training provider, which has about 200 staff, received a 40 percent discount on models used for internal product development, conditional on also using Codex rather than OpenCode or Claude Code. CEO Jonathan Cornelissen said DataCamp expects overall costs to fall 80 to 90 percent. The Information reported on October 24, 2026 that OpenAI's pricing cut DataCamp's costs by 80 to 90 percent.

A US bank with more than $7 billion in assets shows how seat pricing works in practice. Sina Finance reported on October 25, 2026 that the bank doubled its ChatGPT seats to 500 employees while keeping its annual contract at roughly $40,000, which works out to under $7 per user per month instead of the usual $20 or more. IT ticketing startup Atomicwork plans to spend about $2 million on Claude Code in the fiscal year ending January but only about $300,000 on Codex, up from almost nothing, after receiving free credits from investor Khosla Ventures.

Analysis

The pattern is consistent: OpenAI is using price to change default behavior in accounts where coding work is large, recurring and measurable. Replit, Rippling, DataCamp and the bank are not small experiments. They are accounts with enough volume to justify steep discounts and enough internal visibility for a shift to become a public reference. OpenAI is partly buying market share at the expense of its own margins, as The Decoder reported on October 25, 2026, but it has also achieved real price cuts through its Sol and Luna model lines, more efficient models and infrastructure.

Anthropic's position is more complicated than a simple loss of accounts. Its growth slowed from the second to the third quarter, though it still expanded about 45 percent sequentially, according to The Information's October 24, 2026 report. The figures attached to the report put Anthropic's Q3 2026 revenue at $16.8 billion, with an IPO planned for November. Sina Finance reported on October 25, 2026 that Anthropic's Q2 compute spending was $5.6 billion against $11.5 billion of revenue, a reminder that serving frontier models is expensive even when revenue is growing.

What this really means is that OpenAI is attacking the renewal moment, not just the model leaderboard. Anthropic's discount rule, which cancels discounts once a customer hits its contract cap, is what cost it the Replit order, according to Sina Finance's October 25, 2026 account. A pricing policy designed to protect margin can become a sales liability when a rival is willing to accept lower margin for a strategic account. The Replit episode also shows that switching costs are lower than they once were: Replit needed hours to change code and a few days for product level adjustments, faster than its original contract negotiations with Anthropic.

The bigger picture here is that enterprise AI pricing is becoming a procurement fight with public benchmarks of its own. The Decoder reported on October 25, 2026 that both OpenAI and Anthropic face price pressure from Chinese models from DeepSeek and Kimi that deliver comparable performance for a fraction of the cost. On OpenRouter, OpenAI led a month ago, but Anthropic retook the top spot after releasing Claude Opus 5.5 in late September and last week drew about twice OpenAI's usage, according to Sina Finance's October 25, 2026 report. Usage leadership and enterprise contract leadership are not the same thing, and OpenAI's discounts are aimed at the second.

Why It Matters

The immediate stakes are financial. Anthropic is preparing a November IPO, and investors will look at enterprise retention, pricing power and revenue quality, not only model benchmarks. A public example such as Rippling, where Anthropic's share of coding work fell from 57 percent to 18 percent, is more damaging than a lost logo because it shows a change in daily developer behavior. Replit matters for the same reason: a contract worth more than $100 million a year is large enough to move revenue expectations and to serve as a reference for other platform companies.

The strategic stakes are broader. If default model settings continue to shift toward OpenAI, Anthropic can keep a presence in accounts while losing the bulk of usage. AI Weekly reported on October 24, 2026 that Replit's documentation still lists seven Claude models and that the Anthropic partnership is still in place, but OpenAI is the default for Replit Agent. That arrangement gives Anthropic technical availability without guaranteed volume, a weaker position than the headline word partnership might suggest.

For buyers, the discounts are real but conditional. DataCamp's 40 percent model discount came with a requirement to use Codex, and the bank's lower per seat price came with a larger seat count. Those terms show that OpenAI is pricing for commitment and expansion, not simply cutting list prices for everyone. The risk for customers is lock in through defaults and workflow changes, while the benefit is lower costs at a moment when AI budgets are under scrutiny.

Next Up

The next test is whether Anthropic's November IPO changes the pricing dynamic. A public company has to defend margins and show durable enterprise demand, which may make it harder to match OpenAI's discounts account by account. The Information reported on October 24, 2026 that Anthropic's growth slowed from the second to the third quarter, so the IPO story will depend on whether large customers stay and whether Claude Opus 5.5 can convert usage interest into contracted enterprise volume.

OpenAI, meanwhile, must show that share gained through discounts becomes profitable share after renewals. The Replit, Rippling and DataCamp cases will be watched as templates. If defaults stick and expansion follows, OpenAI will have bought a durable position in enterprise coding. If customers treat the discounts as temporary, the price war will continue into the next contract cycle, with DeepSeek and Kimi adding pressure from below.