Anthropic weighs a new model to hold off OpenAI's Astra, as its IPO slips to November

Start with what the reporting actually establishes. According to Reuters, three sources said Anthropic is considering rolling out a new model to answer the competitive pressure from Astra, and one of the people familiar with the matter said the company is evaluating the next model's safety as part of its deliberations over a release. Reuters added that some of the discussions involve how to balance investment in releasing new models against efforts to strengthen profitability, against a backdrop of rising interest rates that have made investors more focused on the timing of expected profits. Anthropic declined to comment. Taken together, the news is not really about whether to ship a model. It is about a company whose brand is built on safety first, sitting in a position where it has to answer questions pointing in two directions at once.

Where the pressure comes from, Reuters gives three sets of numbers. The first is about the essay itself: published September 12, 3,800 words, painting a picture of swarms of AI agents taking over the internet and outpacing human control, an argument that OpenAI chief executive Sam Altman and SpaceX chief executive Elon Musk each publicly came out in support of. The second is about Astra's market performance: OpenAI released GPT-6 Astra on September 3, touting gains in computer use, software engineering, cybersecurity and professional work, with a strong response from enterprise users and developers; Astra accounts for about 13% of enterprise AI spending tracked by the corporate expense platform Ramp, against roughly 8% for Anthropic's Claude Fable. The third is about developer traffic: OpenRouter said its users spent more on OpenAI models than on Anthropic models last week, the first time OpenAI has led on that measure in more than two and a half years.

But the reporting also carries counter-evidence, and that side is just as specific. Some existing investors, and others who expect to invest in both Anthropic's and OpenAI's listings, told Reuters they do not believe Astra poses a significant threat to Anthropic, for two reasons: the size of its lead in the enterprise AI tools market, and the time it takes to unseat incumbent vendors at large companies. On the numbers, Anthropic's annualised revenue run rate topped $65 billion by the end of July, up from about $9 billion at the end of 2025, and the company is projecting 2028 revenue of roughly $190 billion to $200 billion. For comparison, OpenAI's annualised revenue run rate passed $40 billion in July. Investors also expect leadership among Anthropic, OpenAI and Alphabet's Google to shift repeatedly as each releases new generations of models, making any advantage potentially short-lived.

The real variable is placed further down in the reporting: the rise of open-source and open-weight models. Investors' judgement is that these will lower token costs and let companies build more of their own AI infrastructure rather than rely on providers such as Anthropic and OpenAI, putting pressure on the economics of the AI industry and turning a two-company race into a broader cost problem. There is also one concrete customer signal: Meta has been among Anthropic's largest customers, but people familiar with the matter said it is looking to reduce its use of Anthropic's models as it develops more AI capabilities internally. Meta did not immediately respond to a request for comment. On timing, the Wall Street Journal reported that Anthropic plans to stage its IPO in November, later than many investors expected, while Reuters said the listing could slip to after November's US midterm elections, with two people saying the elections are not expected to have a major impact; Reuters had previously reported that marketing was expected to begin in mid-October at the earliest. For contrast, OpenAI chief executive Altman confirmed on September 19 that the company would not go public in 2026, saying concerns around AI safety made this an ill-advised time for a listing.