Replit is no longer working with Anthropic, CEO says

Replit CEO Amjad Masad said the company is no longer working with Anthropic, citing cost, partnership considerations and performance, and said ChatGPT is now better; however, people familiar with the partnership disputed that it had ended, and Replit Agent was cited as using Anthropic models in September. Masad said enterprise contracts rose from 2% of revenue last year to 15% this year, while B2B accounts for more than 50%; he did not disclose ARR.

Analysis

The investable signal is not a clean model-vendor win: the account of a terminated relationship conflicts with evidence of recent Anthropic model use and an unnamed source saying the partnership remains active. Treat this as a possible change in model mix or commercial terms—not proof of a full switch—until Replit clarifies scope and timing.

If Replit is shifting workloads, the second-order effect is stronger evidence that AI coding platforms can route among models on quality and unit cost, weakening any single model vendor’s distribution advantage. That could pressure model providers to discount inference or offer better tooling, while benefiting platforms that can substitute models without degrading user experience. OpenAI may gain a reference account, but one private customer’s decision does not establish a broad performance lead. Anthropic’s potential loss is likewise limited absent usage or revenue data.

The enterprise mix claim is directionally constructive for Replit’s business quality, but not enough to infer durable recurring revenue, margins, or scale: the company withheld ARR, and the stated revenue shares do not disclose contract size, retention, or inference economics. Over 1–3 months, verify whether Anthropic remains in production and whether enterprise bookings convert to retained revenue. Over 6–18 months, the key structural test is whether security, identity, sandboxing, and orchestration make coding agents acceptable to larger buyers—or whether reliability and governance costs erase model-cost savings. No clear public-equity trade follows directly; Microsoft, Amazon, and Alphabet have only indirect exposure through cloud and developer ecosystems.

Key Decisions for Investors

• Do not trade this as a confirmed Anthropic-to-OpenAI displacement. Seek clarification on whether the change applies to all production workloads, selected models, or commercial terms before assigning vendor-level impact.
• Put Replit’s enterprise conversion on a watchlist: look for disclosed ARR, net retention, contract duration, and gross-margin or inference-cost trends. Rising enterprise revenue share without retention or unit-economics evidence is not yet a quality upgrade.
• For public exposure, keep Microsoft, Amazon, and Alphabet as indirect beneficiaries to monitor rather than initiate a position on this report alone; the mechanism is demand for cloud and developer infrastructure, not a demonstrated near-term earnings catalyst.
• Falsify the model-routing thesis if Replit confirms Anthropic remains a material production provider, or if subsequent product evidence shows users cannot switch models without worse coding outcomes. Reassess enterprise optimism if bookings fail to convert or security/reliability requirements raise delivery costs.