Anthropic close to going public: here’s what you need to know
Anthropic has filed the paperwork needed to list its shares on the Nasdaq stock exchange. Its official filing reports a net loss of $80.7 billion for the first half of 2026, but $72.4 billion of that loss comes from accounting rules rather than actual cash spending. In the second quarter, the company made an operating profit of more than $1 billion. At the same time, Anthropic has committed to spending between $542 billion and $567 billion on computing power and infrastructure.
On October 26, Anthropic filed an S-1 registration statement to list its shares on the Nasdaq under the stock symbol ANT. This appears to put it ahead of OpenAI in the race to become the first major AI model developer to go public. OpenAI is the company behind Claude's rival AI products, and the race for the first major IPO had already started to take shape when Anthropic first filed private registration documents with the US Securities and Exchange Commission (SEC).
Anthropic's planned stock market launch is now expected to value the company at around $1.5 trillion. Just one week ago, estimates put its value at $2 trillion. That estimate was lowered after Anthropic shared its financial figures with investors. Even so, this would be a major increase in value. In its latest funding round in May, Anthropic was seeking $30 billion in investment at a company value of $965 billion.
At first, the numbers look worrying. For the first half of 2026, Anthropic reported a net loss of $80.7 billion. However, $72.4 billion of this was an accounting loss, not money the company actually spent. Amazon and Google have helped fund Anthropic in recent years through convertible loans and warrants. These are financial agreements that can turn into shares or give their holders the right to buy shares. As Anthropic's estimated value has risen sharply ahead of its stock market launch, the value of these agreements on its balance sheet has also increased. The company must record the difference as a loss. But this accounting loss does not use up any of Anthropic's cash.
From losses to quarterly profit
The company's day-to-day business tells a different story. In the second quarter, Anthropic earned $11.6 billion in revenue and made an operating profit of more than $1 billion. Its total revenue for the first half of the year was $16.3 billion, compared with $1.2 billion during the same period a year earlier. However, the company still reported an operating loss of $693 million for the full six months. This means the results from the first quarter still weighed heavily on the total. By May, it was already clear that Anthropic was close to making a profit, with expected revenue of $10.9 billion. The company is now shifting its focus away from the expensive work of training AI models and toward running them for large numbers of customers. This includes customers who send requests to its models through APIs, which allow software to communicate with other software.
One especially striking figure is Net Dollar Retention (NDR). In the traditional software-as-a-service (SaaS) industry, an NDR of 120 percent is considered excellent. This means existing customers, without counting new customers, spend 20 percent more each year. Anthropic's S-1 filing reports an NDR of nearly 900 percent. In other words, business customers have increased their spending on Claude APIs to roughly nine times what it was a year earlier. This also raises questions about how easy it is for customers to leave. System prompts, retrieval-augmented generation (RAG) systems, and fine-tuning are often designed to work with the specific features of one AI model. Moving to another provider can therefore be expensive.
Hundreds of billions of dollars committed to computing power
This growth also comes with major financial commitments. Anthropic has signed contracts to buy hundreds of billions of dollars' worth of computing power from AWS, Google Cloud, and, to a smaller extent, Azure. Some of these contracts run until 2036. Anthropic must pay for the agreed capacity even if demand for its AI models turns out to be lower than expected.
The biggest commitments are with Google, worth at least $114.5 billion through July 2033, and Amazon, worth at least $110 billion through April 2036. Microsoft also has a commitment of $44.9 billion through May 2033. Together, these agreements create $265.6 billion in fixed minimum payments. There are billions more in other commitments. The filing also includes plans to buy additional Google tensor processing units (TPUs), special chips used for AI computing, between 2028 and 2031. These will cost another $125 billion to $150 billion. Two other deals are each worth more than $100 billion: an agreement with Broadcom for TPU capacity worth $161.2 billion, and a flexible contract with SpaceX worth up to $120.5 billion that can be cancelled.
These arrangements also matter to business customers buying Claude. They can pay for Claude capacity through existing agreements with AWS or Google, or sign a contract directly with Anthropic. Buying through AWS or Google makes the customer more dependent on that cloud provider. A direct contract means the customer pays according to Anthropic's pricing structure, which is designed to cover the company's own cloud computing costs.
A trust with special powers of control
Anthropic's leadership and control structure is different from that of many other publicly traded technology companies. The company was founded because of concerns about safety risks at OpenAI. Its S-1 filing confirms that the Long-Term Benefit Trust (LTBT) has significant power. Through a separate type of share called Class T, the trust appoints most members of the board of directors. The board has special voting powers and can step in if it believes an AI model could harm society or is unsafe.
At the same time, the founders hold a separate, special Class F share. This share automatically gives them 50.1 percent of all general voting rights. This arrangement allows the trust to oversee AI safety while the founders keep control of the company's business decisions. In theory, the trust can decide to replace a model with a less capable version, take it offline, or turn off certain features. Standard service-level agreements (SLAs), which describe the service customers can expect, do not cover these kinds of changes.
Anthropic has not yet said how many shares it will issue or how much each share will cost. However, this latest step suggests that its stock market launch is very close. According to Bloomberg, the launch could happen in November.