OpenAI pulled in nearly $6 billion in revenue during the first quarter, with its Codex coding tool emerging as a major growth driver, according to The Information. That’s a staggering pace for a company that closed 2025 at roughly $13 billion in annual revenue, and it signals OpenAI is on track to blow past most analyst projections for the year.
The Information reports that Codex, OpenAI’s coding agent, is doing a lot of the heavy lifting behind the quarter’s numbers. Developer tooling has quietly become one of the most lucrative slices of the AI market, and OpenAI just put a number on it.
What’s actually happening
- Revenue: ~$6 billion in Q1 alone
- Key driver: Codex, OpenAI’s coding assistant and agent platform
- Run rate implication: If the pace holds, OpenAI is staring down a $24B+ annualized figure, roughly double its 2025 exit rate
- Context: ChatGPT subscriptions and API usage remain the base, but Codex is now a material line item
Why this matters
The AI coding category went from “interesting feature” to “primary revenue engine” in about 18 months. Cursor, Anthropic’s Claude Code, GitHub Copilot, and now OpenAI’s Codex are all printing real money. What stands out here is that Codex isn’t just a chatbot bolted onto an IDE. It’s an agent that plans, edits, and ships code across multiple files, and enterprises are paying enterprise prices for it.
This also reframes the competitive picture. Anthropic has been eating share in the coding space with Claude Code and the Sonnet family, and many engineering teams quietly switched over the last year. OpenAI’s Q1 numbers suggest Codex is clawing some of that back, or at least growing the overall pie fast enough that both can win.
What to watch next
- Margins: Revenue is one thing, but inference costs on agentic coding tools are brutal. Profitability remains the open question.
- Enterprise deals: Expect OpenAI to push Codex into Fortune 500 IT budgets, where Microsoft’s GitHub Copilot already has a beachhead.
- Pricing pressure: With Anthropic, Google, and a wave of startups in the same lane, the per-seat pricing on coding agents could compress fast.
- Model release cadence: A bigger Codex model or a new agent SKU would likely follow numbers this strong.
For practitioners, the takeaway is simple. Coding agents are no longer experimental. They’re a budget line, and the vendors selling them are now operating at the scale of mid-tier enterprise software companies. Expect procurement teams to start treating them that way: with RFPs, security reviews, and multi-vendor evaluations.
The full breakdown is in The Information’s original report.