OpenAI July revenue run rate topped second-quarter level, CFO told staff

OpenAI July revenue momentum is becoming a key talking point inside the company as it prepares investors and employees for a bigger test in public markets. CFO Sarah Friar told staff on Wednesday that OpenAI’s annualized recurring revenue in July was above its second-quarter level, CNBC reported, citing a partial transcript of the internal meeting.

Friar and OpenAI board chair Bret Taylor used the meeting to stress that demand remains strong while competition grows from Anthropic, Google and lower-cost open-weight AI models. According to CNBC, Friar said the company’s second quarter was strong, while Taylor addressed Anthropic’s recent gains and OpenAI’s push to catch up in coding tools.

The message matters for retail investors because OpenAI and Anthropic have both confidentially filed for initial public offerings with the Securities and Exchange Commission, according to CNBC. Neither company has said when it expects to list shares, but the numbers being discussed now are shaping expectations for what could become one of the most closely watched AI market debuts.

What does annualized recurring revenue mean?

Annualized recurring revenue, often shortened to ARR, estimates yearly revenue by taking the current pace of repeatable sales and stretching it over 12 months. It is a common software metric, but it is still a run-rate snapshot rather than a guarantee that the company will actually book that amount over a full year.

CNBC reported that Friar and Taylor linked OpenAI’s recent growth to several products: the GPT-5.6 model series, ChatGPT Work for enterprise customers and Codex, its AI coding product. Enterprise customers matter because businesses can produce larger and steadier contracts than individual consumers, while developers can drive heavy usage through coding tools.

OpenAI is also trying to support a valuation CNBC pegged at $852 billion while funding an expensive buildout of computing infrastructure. In February, CNBC reported that OpenAI told investors it was targeting about $600 billion in total compute spending by 2030.

That spending plan may include outside support. CNBC reported this week that OpenAI is in talks with Nvidia for a backstop of as much as $250 billion tied to plans to lease a large AI data center in Ohio.

How is Anthropic challenging OpenAI?

Anthropic has become one of OpenAI’s most serious rivals, especially among enterprise and developer customers. CNBC reported that Taylor told employees Anthropic had a strong start to the year and said OpenAI had work to do in the coding market, though he described Codex’s growth as encouraging.

Taylor also told employees that some users who had used Claude Code heavily faced high bills and began looking for another option, according to CNBC’s partial transcript.

Anthropic said in May that its revenue run rate had passed $47 billion, up from roughly $10 billion in revenue for all of 2025. The Information reported in March that OpenAI had recently passed $25 billion in annualized revenue, citing a person familiar with the matter.

Competition is also coming from cheaper AI models that are easier to access. CNBC reported that China’s Moonshot AI introduced Kimi K3 earlier this month and claimed the model narrowed the gap with top U.S. systems while beating OpenAI and Anthropic’s strongest models on some benchmarks.

For investors watching the AI sector, the core issue is whether revenue growth can keep pace with the spending needed to train, run and distribute advanced models. OpenAI’s internal message, as reported by CNBC, is that demand is still growing. The public-market question will be how durable that growth looks once investors can scrutinize the numbers in an IPO filing.

This story draws on original reporting from CNBC.