OpenAI Wins Coding Customers With Price Cuts, But Market Share Picture Is Mixed

The Information reports that OpenAI has captured a majority of coding work at several large enterprise customers by undercutting Anthropic on API pricing, though independent usage data shows a more contested picture.

Replit Switch Anchors the Shift

Coding platform Replit moved the bulk of its AI spending—over $100 million annually—to OpenAI after receiving a discount of more than 30%, according to The Information's reporting covered by AI Chat Daily [1]. Anthropic had previously given Replit a discount but declined to extend it unless Replit committed to higher spend, the report says. Replit CEO Amjad Masad confirmed the switch onstage October 16, citing "cost, partnership, and performance" and stating that in Replit's internal testing, OpenAI's Sol and Astra models "perform better at building vibe-coded applications from scratch" [1].

OpenAI is now the default model provider for Replit Agent; Claude models remain available but must be selected manually [1]. Masad also said Replit "remains model agnostic" and the company's documentation still lists seven Claude models, with a late-September blog post noting Claude Fable 5 and 5.1 running in production [1].

Rippling Data Shows a Sharp Reversal

At HR software company Rippling, The Information reports a dramatic flip: in May, Anthropic models handled 57% of coding work by hundreds of engineers while OpenAI handled 16%; those shares have since reversed to 52% for OpenAI and 18% for Anthropic [1]. The report also says OpenAI offered DataCamp a 40% discount to use Codex [1]. The Information's piece ran under the headline "Anthropic's Margin is OpenAI's Opportunity."

Independent Token Data Tells a Different Story

TokenMaxxer, an opt-in developer network that measures actual API token usage across tools including Claude Code and Codex, shows Anthropic still leading in overall token share. For the week ending October 25, Anthropic accounted for 66% of tokens versus 33% for OpenAI [3]. That week's total was 67.3 billion tokens across 22 Anthropic developers and 15 OpenAI developers on the network [3].

The two metrics differ: The Information cites "coding work" share at specific enterprise accounts, while TokenMaxxer measures token volume across its self-selected developer base. TokenMaxxer's own documentation notes that its sample describes developers who chose to share their usage, not the whole market, and that on a network this size one heavy user can move a week [3]. The Information's report did not include a statement from Anthropic on the Replit contract or the Rippling figures [1].

Enterprise Revenue Rising at Replit

Replit's shift coincides with a change in its customer mix. Masad said enterprise contracts grew from roughly 2% of revenue last year to about 15% this year, with more than half of revenue now coming from business customers [1]. He expects enterprise and government to exceed half of revenue in 2027 [1]. That trajectory makes Replit the kind of high-spend account both model providers want as a default.

What's Actually Demonstrated

• OpenAI offered Replit a >30% discount on a $100M+ annual contract; Anthropic would not extend its prior discount without a higher spend commitment [1].
• At Rippling, OpenAI's share of coding work rose from 16% to 52% while Anthropic's fell from 57% to 18%, per The Information [1].
• OpenAI offered DataCamp a 40% discount for Codex, per The Information [1].
• TokenMaxxer's live network shows Anthropic still ahead in token share (66% vs 33%) as of Oct 19–25 [3].

What remains unknown: whether the Rippling and Replit moves represent a broader trend or two high-profile wins; what Anthropic's retention and win rates look like across its full customer base; and whether price concessions are sustainable for either provider at scale. The Information's framing—that Anthropic's margin structure leaves it vulnerable to price competition—is a claim, not a demonstrated fact.