$16.3 billion revenue, $417 billion compute commitments: Inside Anthropic’s IPO numbers

Anthropic’s planned initial public offering (IPO) will bring greater public attention to the costs and profits of advanced artificial intelligence (AI). The company’s latest filing shows that its revenue is growing quickly, but it is also facing large losses, long-term commitments to computing infrastructure and warnings about the possible dangers of its technology.

Anthropic, the company behind Claude, has filed its registration statement with the US Securities and Exchange Commission (SEC). It plans to list its shares on the Nasdaq Global Select Market under the ticker symbol ‘ANT’. Anthropic first filed its paperwork privately in June, allowing regulators to review its finances before the information became public. Goldman Sachs and Morgan Stanley are leading the offering, along with 20 other banks. The filing does not yet say how many shares the company will sell.

The financial details highlight a major question for AI companies entering the stock market: Can the growing use of their AI models lead to lasting profits while the cost of computing infrastructure keeps rising?

Anthropic reported revenue of $16.3 billion in the first six months of 2026, compared with $1.2 billion during the same period last year. However, its net loss grew to $80.7 billion from $13 billion, according to the filing. This reported loss includes large accounting charges that do not involve cash payments and are linked to the rising value of the company. So, the loss does not mean Anthropic spent an equal amount of cash.

Much of Anthropic’s revenue growth has come from businesses using Claude to write software and handle workplace tasks. Its net revenue retention rate, which measures how much revenue a company earns from its existing customers over time, is about 900%, according to the filing. Anthropic also said that 79% of its revenue came from customers outside its five biggest accounts in the third quarter. This suggests that the business does not depend only on a small number of major customers.

Payments based on how much customers use Anthropic’s AI models have become an increasingly important source of revenue. In 2024, this type of revenue was $288 million, or 75% of total revenue. In 2025, it rose to $3.8 billion, or 83%. From January to June 2026, it reached $13.1 billion, making up 80% of revenue.

These figures show that businesses are using AI models more often. They also highlight the link between customer demand and the computing infrastructure needed to provide these services.

Computing commitments emerge as a key financial risk

Anthropic’s filing shows how much it spends on the computing power and infrastructure needed to build and run advanced AI models. The company spent $8.9 billion on computing power and infrastructure in the first half of 2026, up 283% from the same period a year earlier.

Since the end of 2025, Anthropic has signed additional computing agreements with minimum payment commitments of about $417 billion. These contracts are generally expected to begin in 2026 or 2027 and may last as long as 20 years.

Even so, infrastructure spending has become a smaller share of Anthropic’s revenue as its business has grown. Computing and infrastructure costs were more than 10 times its revenue in 2023. This fell to 653% in 2024 and 160% in 2025. In the first half of 2026, these costs were 55% of revenue. In the second quarter, they fell further to 48%.

This improvement suggests that revenue is growing faster than infrastructure spending. However, the size and length of Anthropic’s contracts could create problems if customer demand is lower than expected. The company has warned that it will still have to make many of these payments, even if it does not use all the computing power it has agreed to pay for.

The filing also shows that Anthropic had total debt of $6.2 billion as of June 30, 2026. Its revolving credit facility, which lets it borrow money when needed and repay it over time, has increased from $3.25 billion to as much as $15 billion. Anthropic said it may borrow more money to meet its future funding needs.

These financial details come as people raise broader questions about whether income from AI services will be enough to justify the huge investments in data centres and computing infrastructure. For investors in the stock market, Anthropic’s results will offer a closer look at how a leading AI developer manages business growth, infrastructure costs and its need for funding.

AI safety and governance feature prominently in the filing

Along with its financial risks, Anthropic has given considerable attention to the dangers its technology could create. Its prospectus warns that advanced AI systems might behave in unexpected ways, act against their intended goals or operate on their own. This could include trying to avoid being shut down or hiding and changing information to influence people. Anthropic warns that without proper safety measures and human oversight, these systems could pose catastrophic or existential risks to humanity.

The company also warned that its public views on AI safety and regulation, along with its work with policymakers, could lead to closer scrutiny, political opposition or unusually strong regulatory action.

These warnings come as Anthropic is involved in disputes with the US government over limits on military use of its technology. The Trump administration has taken action against the company after it refused to remove certain restrictions on how the Pentagon could use its technology. The US Department of Defense later labelled Anthropic a national security risk and barred it from serving as a supplier.

Anthropic’s ownership and decision-making structure will also matter to potential investors. The company is registered as a public benefit corporation. This means it must consider its goal of developing AI safely alongside its financial duties to shareholders. Its seven co-founders, led by chief executive Dario Amodei and president Daniela Amodei, are expected to keep just over half of the voting rights after the listing. This would limit the ability of outside investors to outvote the founders on company decisions.

The co-founders have promised to donate 80% of their shares to charity. However, the filing says these promises are not legally binding.

The IPO is expected to be one of the biggest stock offerings linked to the AI industry. According to reports cited in the supplied information, Anthropic is aiming for a possible offering of up to $100 billion, with a company valuation of $1.5 trillion to $2 trillion. These figures are reported targets, not final terms confirmed in the registration statement. The listing date could also change.

Anthropic was founded in 2021 by Dario Amodei and a group of researchers who left OpenAI. Its growth has sped up in 2026, helped by demand for Claude Code and its other AI models. More businesses are using AI to develop software and manage their internal work.

The planned listing would bring a major developer of advanced AI models onto the stock market before its rival OpenAI, which has reportedly moved its own planned stock market debut to 2027. After SpaceX’s record-setting IPO earlier this year, Anthropic’s offering will also test whether investors are willing to back companies whose future growth depends heavily on large infrastructure investments that may take many years to pay off.

For investors, the filing shows both the opportunities and the risks of the AI boom: a business growing quickly among enterprise customers, alongside financial, operational and safety risks linked to increasingly powerful AI systems. The final valuation, offering terms and market reaction will help determine how investors judge this balance.