In January 2021, seven former OpenAI employees launched a new AI company: Anthropic. Just under six years later, the creator of the ubiquitous Claude has made it to the Nasdaq before OpenAI. More important is the trajectory the AI company’s value is now taking following a downgraded valuation and widespread doubts about whether the hype surrounding the technology will hold up.
At the time of publication (4:00 p.m. Central European Time), the IPO is valued at $81.3 billion, a new record that surpasses SpaceX’s $75 billion. Thanks to a private investment of $26 billion, the total comes to $107.3 billion. The valuation stands at $1.5 trillion at the time of the IPO. It is striking that this is relatively low, given that the widely reported valuation once stood at $2 trillion.
Opus, Fable, and a mountain of fear
Claude users are all too familiar with the progress Anthropic has made. The company was the first to surpass the benchmark set by OpenAI’s GPT-4 model in early 2024 with Opus 3, gained significant recognition thanks to the success of Sonnet 3.5 in October of that year, and has made an indelible impression with breakthroughs such as Opus 4.5 late last year and the cybersecurity marvel Mythos in 2026.
Anyone who simply accesses the models via the API or a subscription may well have missed a crucial part of Anthropic’s strategy. In addition to highly competitive LLMs that are particularly popular among developers, CEO Dario Amodei’s company has a mountain of fear to sell. Although Anthropic itself released an LLM called Fable 5, which was pulled by U.S. authorities in June for security reasons, it points to competitors, open-source LLMs, and techniques such as distillation to emphasize the dangers of AI.
The IPO
So far, Anthropic and its archrival OpenAI have been able to rely on massive investments and, until now, insufficient revenue figures to ensure their own survival. With the IPO, Anthropic’s financials will become public and answer many questions. Questions such as “How profitable is AI for AI developers?” and “Is the AI expansion sustainable for frontier labs?” won’t be answered immediately, but the spotlight of a publicly traded company will provide stronger clues than ever before.
Returning to the more practical aspects of the IPO: the order book has been oversubscribed multiple times, and the number of shares is very small, a factor that already appeared to drive a price increase at SpaceX earlier this year. This gives the founders and major investors the opportunity to raise substantial funds. Anthropic itself will also see money flowing in.
Nvidia, Google, AMD, Broadcom, Micron, Microsoft, and TSMC are collectively investing $26 billion in Class N shares, without voting rights. With this offering, Anthropic will raise approximately $105 billion in new capital. However, shareholders must still approve the offering.
The run-up to the IPO wasn’t entirely smooth. On November 10, Anthropic withdrew a draft version of the prospectus because the pricing information it contained was reportedly incorrect. There are also widespread doubts surrounding the $30 trillion (!) that Anthropic cites as its Total Addressable Market in its IPO filings.
A quarter below 2 trillion
As we described earlier, a valuation of around 2 trillion dollars was circulating in the run-up to the IPO. Admittedly, that figure came from press releases and was not included in the official documents. In May, Anthropic was still valued at $965 billion. But the decline in its estimated value is a sign that the once seemingly boundless optimism is, in fact, limited.
Revenue is growing rapidly regardless and is considered the main argument in favor of a rise in value. For example, revenue was $16.3 billion in the first half of the year, compared to $1.16 billion a year earlier. The third quarter generated approximately $17 billion. However, Anthropic is not yet turning a profit. The net loss for the first half of the year amounted to a staggering $80.7 billion. Of that amount, $72.4 billion was an accounting entry related to convertible financing, which did not involve any actual cash. The company also has approximately $895 billion in outstanding commitments for computing power and data centers. That amount has driven up the valuations of many other companies in the AI industry.
Founders retain control
New shareholders will not have much say, as was also the case with the SpaceX IPO. A single Class F share held by the seven co-founders accounts for 50.1 percent of the votes.
In the prospectus, Anthropic cites certain risks, such as the “supply chain risk” label the Pentagon has assigned to the company. Anthropic has filed lawsuits regarding this matter in two federal courts.
Meanwhile, OpenAI remains private for the time being and is aiming for an IPO in 2027. Goldman Sachs expects Anthropic to generate approximately $53 billion in revenue this year, compared to about $35 billion for OpenAI.