Anthropic multi-line “preparing” for IPO: introducing Accenture (ACN.US) to conduct security assessments, with annualized revenue of $100 billion

The Zhitong Finance App learned that Anthropic is simultaneously promoting two things: sending itself to Wall Street and finding an external endorser for the word “safety” before going public. According to people familiar with the matter, the AI company's annual revenue is expected to exceed 100 billion US dollars and is preparing for the November listing transaction as soon as possible; at the same time, Anthropic will bring in evaluators from consulting firm ACN.US (ACN.US) and directly embed them into its operating system to conduct safety tests on the latest models. The first one shows how fast this business makes money, and the latter one shows that it wants to finish answering the question “who will oversee me” before ringing the bell.

Looking at these two pieces of news together, Anthropic's preparation for listing is roughly divided into three parts: proof of revenue, capital arrangements, and paving the way for reputation and compliance. The first two pieces have clear numbers and timelines, while the third block is the hardest to quantify, and the easiest to question.

IPO: Listed in November as soon as possible, the roadshow may start in mid-October

According to the New York Times, Anthropic's revenue run rate (run rate, an indicator of annual revenue in a short period of time) has soared since this year as enterprise customers are increasingly willing to use its Claude artificial intelligence software to simplify programming and other work tasks. The report also mentioned that while advancing the IPO plan, the company is also facing a significant rise in public concerns that AI may cause catastrophic damage. Anthropic declined to comment.

Anthropic's revenue operating rate reached $65 billion at the end of July, compared to only $9 billion at the end of 2025. The sharp acceleration in this round of revenue provided strong support for the company's listing plans. Both Anthropic and OpenAI have confidentially submitted listing documents, and Anthropic is expected to complete their Wall Street debut within the next few weeks — the two companies' race for open market funding has begun. Earlier this month, it was reported that Anthropic will launch roadshow marketing for its IPO as early as mid-October and complete listing a few days before the November US midterm elections.

There have been more specific specifications before regarding the size of this IPO. According to reports, Anthropic plans to raise up to 100 billion US dollars through the IPO, with a target valuation of about 2 trillion US dollars. Nvidia is negotiating to invest up to 10 billion US dollars as a cornerstone investor; the company secretly submitted S-1 documents in June and selected NASDAQ as the listing site. What needs to be distinguished is that $2 trillion is the target valuation circulating in the market, not the final issuance price.

According to a social media post, Anthropic also hosted a venture capital forum this week to discuss the progress of its AI work, which was attended by several investors. One of the questions in the campaign's public vote was how to determine whether a company has implemented a stronger form of AI known as “general artificial intelligence” (AGI).

Revenue operating rate: From $9 billion, $65 billion to $100 billion

The figure of 100 billion US dollars is a further extrapolation of the “New York Times” over its previous statement. According to reports, Anthropic's operating rate jumped from 9 billion US dollars at the end of 2025 to 65 billion US dollars at the end of July; the “New York Times” recently gave an annual level of “over 100 billion US dollars this year.”

Anthropic told investors that the company's second-quarter revenue was over $11.5 billion, with an annualized revenue operating rate of over $65 billion, which is roughly in line. For enterprise customers, Claude's selling points focus on programming efficiency and automation of everyday office tasks — a major source of increasing operating rates.

Security preparations: Accenture evaluators are in, and the two sides plan to invest at least $1 billion

Anthropic said in a blog post published on Friday that Accenture's evaluators will launch Red Team tests on the lab's latest model, test its safety protection mechanisms, and conduct evaluations to ensure that its technology is “aligned” — that is, behavior is consistent with human goals. According to the blog post, the authority enjoyed by these evaluators within the company is similar to that of Anthropic's own employees.

In a 3,800-word long article published last Saturday, Amoudi responded to growing concerns about the security risks of cutting-edge AI models, called for a slowdown in development, and stated that new security measures will be implemented. Third-party evaluators are one of them.

According to previous public reports, Amoudi's reason for this appeal is that a mechanism known as “recursive self-improvement” is beginning to appear across the industry — that is, AI participation in improving itself, if unrestrained, may exceed the scope of human ability to understand and control these systems. This also explains why he included an “embedded independent evaluation” into the plan rather than just publishing a security promise.

Anthropic said that Accenture's embedded evaluators will closely track how the AI model evolves during training and observe the undisclosed decision-making processes behind Anthropic's development technology. The company stated: “From this position, embedded evaluators can assess how a company is operating, verify that it has fulfilled its safety commitments, and identify blind spots. They can also report events to give the public a fuller and more informed understanding of the benefits and risks.”

The two companies are expected to invest at least $1 billion in this area to build related capabilities. Anthropic and Accenture have previously collaborated: last year, they reached a multi-year partnership and formed a team of 30,000 people to receive Claude training to help customers deploy AI systems on a large scale; in March of this year, Accenture also launched Cyber.AI, a security solution driven by Anthropic's Claude model.

Dispute: How independent is this “third party”

The “third party” nature of the collaboration may be called into question — Accenture's close relationship with large AI startups can hardly be viewed as unrelated outsiders. Earlier this year, the consulting giant anticipated that its business associated with key AI partners such as Anthropic, OpenAI, and Palantir would more than double over the next year. In other words, the party that was invited to rate is also the beneficiary of this relationship. This is likely to be questioned over and over again after the manuscript is published.

The original text did not explain how the evaluation findings would be disclosed to the outside world, nor did it indicate the size and work cycle of the evaluators.

Why now: resignations, calls to slow down, and “AI makes AI”

Anxiety about AI's existential risks heated up last week due to a high-profile termination incident. Anthropic employee Jacob Coxon (Jacob Coxon) posted a resignation post on social media accusing the AI company of “gambling our lives.” According to previous public reports, Coxon had previously conducted pre-training research at OpenAI and Anthropic, and posted a resignation statement on social networking platform X on the evening of September 8, which was followed up by various media outlets.

Over the next few days, AI company leaders, including Amoudi and OpenAI CEO Sam Altman (Sam Altman), urged slowing down the technology's progress to deal with its increasingly difficult risks — but they were divided on exactly how to handle it. Amoudi previously suggested that the company should open up to embedded independent evaluators to oversee the technology development process. Accenture's collaboration is a sample implementation of this proposition.

The division within the industry has been traced before: in July of this year, more than 1,100 employees from leading AI companies, including Anthropic and OpenAI, signed a petition requesting the US government to support a mechanism that can “deliberately set the pace” for AI development.

A report published by Anthropic last Thursday (9/10) provides another side: 26% of the company's AI R&D work is driven by its Claude chatbot. This number itself is also a go-to-market narrative — it points both to efficiency and to the hidden concern that “AI is speeding up the creation of stronger AI.”