Anthropic’s reported revenue run rate tops $65B as IPO race heats up

Anthropic’s business momentum is accelerating ahead of a widely expected public-market debut, according to new reporting from Axios that cites figures previously reported by Bloomberg. The August 17 report says the AI lab’s annualized revenue run rate surpassed 65 billion dollars at the end of July, a striking figure even in an industry where growth expectations have been stretched by demand for coding agents, enterprise assistants and frontier model APIs.

The scale of the reported quarter is the main reason investors are paying attention. Axios says Anthropic generated more than 11.5 billion dollars in preliminary revenue in the second quarter, based on documents seen by Bloomberg. That would be more than 14 times the same quarter last year and more than double the 4.73 billion dollars reported for the first quarter. The 65 billion dollar annualized run rate also implies a sevenfold increase from the end of last year.

The comparison with OpenAI is unavoidable but should be read carefully. Axios reports that OpenAI’s latest revenue run rate reached 40 billion dollars, citing an internal message shared by co-founder Greg Brockman the previous week. The article also notes that the two companies may not calculate revenue metrics the same way. That caveat matters because private AI companies can classify usage, commitments, credits and enterprise arrangements differently, making clean comparisons harder than a headline number suggests.

The revenue surge is closely tied to the IPO race. Axios says Anthropic is meeting potential new investors ahead of a planned listing that could come in September or October, with Morgan Stanley, Goldman Sachs and JPMorgan working on the offering. Public markets would give Anthropic access to a larger capital base at a moment when frontier AI companies need enormous spending capacity for chips, data centers, model training, inference and sales expansion.

Enterprise adoption appears to be the center of the story. Anthropic has built much of its commercial momentum around Claude in business workflows, especially software development and knowledge work where accuracy, long-context reasoning and tool use can justify premium pricing. Axios cites PitchBook analyst Harrison Rolfes arguing that a more expensive model can still cost less per successfully completed task if it produces correct answers more often and reduces the need to rerun work or add human review.

The numbers also point to a new competitive axis: efficiency, not just raw intelligence. Leading labs are trying to lower inference costs through specialized infrastructure, model routing, partnerships with inference providers and internal chip efforts. If a company can deliver a higher completion rate per dollar, it can expand margins while still giving customers a reason to pay for premium models. That is especially important as business users shift from casual prompts to delegated tasks that consume more tokens and require more reliability.

There is still reason for caution. Run rate annualizes a shorter period and can overstate durability if demand, pricing or capacity changes. Anthropic did not immediately comment to TechCrunch on similar reporting, and the underlying figures remain private-company data filtered through sources and documents. Even so, the report captures the larger moment: generative AI is becoming a revenue market, not only a research race. For enterprise buyers, the question is whether paid AI systems reliably save time, reduce errors and fit into governed workflows. For investors, the question is whether today’s explosive growth can support the infrastructure spending and valuations now being attached to the leading labs.