Revenues in the billions: Anthropic aims to go public with a valuation of trillions

Anthropic's planned IPO will be a stress test for the AI ​​industry. The US company is currently targeting a valuation of up to one trillion US dollars. Originally there was talk of twice as high a sum - but the euphoria has been muted.

The US AI company Anthropic with the chatbot Claude is heading for one of the largest IPOs in history. The updated prospectus filed with the US Securities and Exchange Commission (SEC) shows that the company is seeking an aggregate valuation of up to $1 trillion.

Initially, shares will be offered at a price of 70 to 78 dollars (around 62.40 to 69.53 euros). The Anthropic share is to be listed on the New York technology exchange Nasdaq under the ticker symbol ANT. Anthropic itself wants to sell up to 801.4 million shares, the previous owner wants to sell 23.6 million shares. Company boss Dario Amodei is targeting gross revenue of up to 64.4 billion dollars (57.6 billion euros) - slightly less than the space and technology company SpaceX at its record debut in June. Elon Musk's group had raised $75 billion with its basic offer, including $86.2 billion including over-allotments.

In the previous months there had even been speculation on the market that Anthropic would be valued at up to two trillion dollars. The reason for the valuation discount is the noticeable cooling of the global AI euphoria on the financial markets, which has recently been reinforced by a downturn in the industry as well as severe turbulence and allegations surrounding the chip giant Nvidia.

The move onto the trading floor is flanked by a billion-dollar deal with large industry partners, some of whom are also in competition with the start-up. Several tech companies have committed to buying their own shares worth a further $26.0 billion directly from the company in parallel with the IPO. However, these papers are not linked to any right to have a say, so they do not confer any voting rights. The backers include Nvidia, Microsoft, Google, Advanced Micro Devices (AMD), Broadcom, Micron Technology and the chip manufacturer TSMC.

The figures submitted to the stock exchange supervisory authority show rapid business development. Since the commercial launch in the first quarter of 2023, quarterly sales have climbed from zero to $17.0 billion in the third quarter of 2026. In addition to the Claude language model, the main drivers are software assistants such as Claude Code and Claude Cowork, which carry out development and office tasks independently.

At the same time, the document shows immense obligations to secure computing capacity. Anthropic estimates the long-term purchase and rental agreements for computer chips and data centers at around $900.9 billion. To distribute the computing workload, the company uses systems from Amazon Web Services, Google, Nvidia and AMD.

The distribution of power within the group is particularly unusual. Anthropic is legally organized as a company with a public interest obligation. In addition to profit, management must also consider ethical standards and the safety of the technology. An independent trust board determines the majority of seats on the supervisory board through special rights.

The co-founders around CEO Dario Amodei and company president Daniela Amodei are also guaranteed a fixed voting majority of a good 50 percent in all other decisions through special shares. Anyone who buys the shares regularly on the stock exchange has full entitlement to financial profits, but practically no say in the company's management. Trading on the stock exchange is scheduled to start as soon as the US authorities have finally released the documents.

In the market for cutting-edge AI, Anthropic is in cut-throat competition with several well-funded rivals. The main competitor is the Microsoft-backed industry heavyweight OpenAI, which dominates the end customer business with ChatGPT. In addition, the competitor Google not only has an AI with Gemini, but also its own cloud infrastructure and the Android smartphone platform. In addition, there is the Facebook parent company Meta, which primarily relies on open systems with freely accessible Llama models.

While Anthropic lags behind the market leaders' user numbers among end users, the company has established a strong position among business customers and programmers. With applications for automatic software development and a focus on legally compliant, controllable systems, Claude is recording above-average growth, especially with corporate customers via interfaces.

However, the development of AI systems has not been without setbacks and has fueled debate about the responsible use of the technology. In the summer, an incident caused a stir in which test versions of the Claude model broke out of an isolated test environment during security tests, accidentally gained access to the open Internet and independently attacked vulnerabilities in company networks and accessed data. Critics see this as evidence that even companies with strong security promises find it difficult to fully control the risks of highly autonomous AI agents in practice.