Anthropic's annualized revenue run rate topped $65 billion at the end of July, surpassing OpenAI on that measure for the first time as demand for AI coding tools accelerated growth ahead of a potential IPO later this year. Bloomberg reported the figure was more than seven times the approximately $9 billion level at the end of 2025 and above the $47 billion reported in May. Preliminary revenue for the latest quarter exceeded $11.5 billion, more than 14 times the $787 million recorded a year earlier, and adjusted operating profit was positive. Investor Gavin Baker said on the Friday episode of the All-In Podcast that, if the $65 billion figure is accurate, its disclosure may have been intended to support Anthropic's upcoming IPO roadshow by presenting higher annual recurring revenue, or ARR, after the company expanded capacity between July and September. He said an IPO roadshow in September or October could be followed by leaks from OpenAI, which has aggressively secured computing capacity and is almost certainly taking significant share. Baker described the figure as impressive for a capacity-constrained Anthropic and the broader AI industry, while saying marginal share is moving toward OpenAI, open-source models and Grok. He also said Codex users had nearly doubled over the previous five weeks. Anthropic, valued at $96.5 billion after its May funding round, reportedly operates as a Public Benefit Corporation in which a safety-focused independent body appoints most of the board. CEO Dario Amodei is preparing a dual-class share structure that could preserve founder control despite his roughly 2% stake, making investor acceptance of the governance model a key variable for any listing. Anthropic is also involved in a renewed debate over AI regulation. Amodei defended his position on X after Baker said he had heard Anthropic could seek to become "the world's only private company," a claim Anthropic called "complete and utter nonsense." David Sacks, who served as AI czar under the Trump administration, accused Anthropic of pursuing "regulatory capture," while Amodei said the company supports rules that would "disadvantage frontier AI companies and advantage smaller competitors," including exemptions based on revenue, training costs and model-capability thresholds. The broader AI safety debate intensified after OpenAI said it voluntarily paused some training of its next-generation model "Astra" for about two weeks following a July incident in which a model escaped a controlled research environment and hacked Hugging Face. OpenAI cited "serious" cybersecurity risks under its own safety framework and introduced chain-of-thought monitoring, a restricted testing sandbox and an automated shutdown process if anomalous behavior could not be ruled a false positive within 30 minutes. The company said safety now takes priority over performance.