Anthropic’s IPO filing opens the books on what it costs to build frontier AI and who pays for it | HT Explainers

Anthropic, the company that makes the Claude chatbot, made its financial records public on Monday. Its filing gives the clearest picture yet of how companies that build the most advanced artificial intelligence (AI) models pay for them.

Anthropic filed a document called a prospectus with the US Securities and Exchange Commission (SEC), the US agency that oversees financial markets. Companies must publish this document before selling shares to the public. The 335-page filing prepares the way for Anthropic to list its shares on the Nasdaq stock exchange next month under the name “ANT”, the Associated Press reported. The company is aiming for the week of November 16, the Wall Street Journal reported, but the date could change.

The initial public offering (IPO), when a company first sells its shares to the public, could raise as much as $100 billion. That would make it the largest IPO ever, surpassing the $86 billion raised by SpaceX in June. Reports estimate that Anthropic could be worth between $1.5 trillion and $2 trillion when it lists its shares. Neither amount is final. According to the filing, shares will be offered not only in the US but also in Australia, Canada, Japan, Switzerland, the UK and parts of Europe.

Anthropic would be the first independent company focused on building frontier AI models to list its shares on the stock market. OpenAI, the company behind ChatGPT, has delayed its own listing until next year. Bloomberg described Anthropic's IPO as a test of whether investors are still willing to put money into AI companies after weeks of concern about high company values and heavy spending.

Much of this growth comes from businesses using Claude to write computer code and do office work. Nearly 12,000 organisations now spend at least $100,000 a year with Anthropic, up from fewer than 1,000 a year earlier, AFP reported. About 250 organisations spend $10 million or more, according to AP. Many customers pay based on how much they use the service. The amount is measured in tokens, which are small pieces of text, code or images that an AI model reads or produces. Individual users can also pay a fixed monthly fee.

Anthropic made its first operating profit in the April-June quarter, earning close to $1 billion. Operating profit means that the money a company earns is greater than the cost of running its business. However, across the full six months, Anthropic still recorded an operating loss of $693 million. That was lower than the $2.1 billion loss it reported for the same period a year earlier, according to the filing.

The filing says the charge was mainly caused by the increase in Anthropic's own company value. As the company becomes more valuable, the notes held by its investors also become more valuable. Accounting rules require Anthropic to record that increase as a loss. The charge reflects the rise in the company's value, rather than money it actually spent, AFP reported.

This type of charge has appeared every year as Anthropic's value has risen. It was $1 billion in 2023, $5.4 billion in 2024 and $34.3 billion in 2025. Anthropic recorded a loss of $13 billion in the first half of 2025, AP reported.

Revenue is growing faster than these costs. In the April-June quarter, revenue increased by 1,368% compared with the same quarter a year earlier, while spending on computing power and infrastructure increased by 300%, according to the filing.

Anthropic has also agreed to pay for much more computing capacity than it currently uses. Since the end of 2025, it has committed $417 billion to computing power and infrastructure through agreements lasting between one and 20 years, Bloomberg reported. Many of these payments are required even if Anthropic does not use all the capacity it has reserved. The company has agreed to spend at least $114.5 billion with Google through 2033, $110 billion with Amazon through 2036 and $44.9 billion with Microsoft, regardless of how much it uses. Anthropic estimates that it will spend another $125 billion to $150 billion on additional capacity using tensor processing units, or TPUs. These are chips designed by Google for AI. Google and other providers are expected to deliver this extra capacity between 2028 and 2031.

Against these future payment commitments, Anthropic had $131.7 billion in assets at the end of June, AP reported. Of that amount, $112.4 billion was cash and cash equivalents, while the rest was held in short-term investments, according to the filing.

Dario Amodei, Anthropic's chief executive, co-founded the company in 2021 with other former OpenAI employees. He has warned that getting demand wrong could cause the company to fail. In a podcast early this year, he said it was unusually difficult to predict demand for a technology that changes so quickly and to buy expensive computing capacity years in advance to meet that demand, Bloomberg reported. “If I’m just off by a year in that rate of growth, or if the growth rate is 5x a year instead of 10x a year, then you go bankrupt,” he said.

Two of these three companies are among Anthropic's biggest investors. Amazon has invested about $18 billion and has agreed to invest up to another $15 billion if certain conditions are met. Google has invested about $23 billion and agreed to invest another $20 billion if performance targets are reached, Bloomberg reported, citing the filing. Anthropic also rents computing power from all three companies under the agreements described above.

All three companies also sell access to Claude. Businesses can buy it through their existing Amazon, Google or Microsoft cloud accounts. The cloud provider charges the customer, keeps a fee and sends the remaining money to Anthropic. These cloud marketplaces accounted for 48% of Anthropic's revenue in the first half of 2026, up from 11% in 2023.

These fees are becoming expensive. Anthropic's spending on sales, marketing and partnerships rose to $3.2 billion from $120 million a year earlier. Most of the increase came from an additional $2.7 billion in fees paid to partners, including cloud companies, Bloomberg reported.

All three companies also develop or sell AI models that compete with Claude. The filing warns that cloud providers can see Anthropic's prices and business terms and might promote their own AI models instead. It also says these business relationships have attracted the attention of antitrust authorities, which investigate whether companies are harming competition. Critics have previously called such arrangements “circular deals” because the same companies invest money in Anthropic and then receive money back from it, Bloomberg noted. Amazon and Google did not immediately respond to Bloomberg's requests for comment.

The first deal of this kind, called Big Sky, raised $35 billion, with Broadcom backing most of the amount. On Monday, Broadcom announced another deal worth $63 billion. Earlier this month, Bloomberg reported that banks were preparing a $42 billion senior portion of a $60 billion chip-financing package intended to benefit Anthropic and other companies. Blackstone, a US investment firm, was leading an $18 billion junior portion. Junior debt is repaid only after the senior debt has been paid. Broadcom and Blackstone declined to comment to Bloomberg.

Anthropic also borrows money directly. As of June 30, it owed Amazon $6.2 billion through convertible notes. It has a $15 billion credit line from banks led by Morgan Stanley, which means it can borrow up to that amount under the agreement. It can also obtain up to another $77 billion through convertible notes from Amazon, Broadcom and Google, Bloomberg reported. The value of leases it had signed but had not yet started paying for, mostly for computing power and infrastructure, rose to $321 billion at the end of June from about $50 billion in December. The filing warns that Anthropic may need to raise money earlier than expected. Anthropic declined to comment to Bloomberg.

Other technology companies were also seeking large amounts of money that week. Meta was trying to raise as much as $95 billion, while SpaceX signed an equipment-leasing deal worth about $40 billion, Bloomberg reported.

The company will publish its results for July-September before its stock market listing.

Anthropic also depends heavily on a small number of customers. Its five biggest customers accounted for about 25% of its revenue this year, AP reported. Nearly half of its revenue also came through the cloud marketplaces of three companies that compete with Anthropic.

Its relationship with the US government is another risk. In February, the Trump administration accused Anthropic of threatening national security and labelled it a supply chain risk. The dispute concerned limits Anthropic had placed on military use of its technology. Anthropic sued the Pentagon in March. A judge in California ruled in Anthropic's favour, while a judge in Washington, DC, sided with the government in a separate, more limited case. Both sides are appealing the decisions.

According to AP, the filing lists risks across 93 pages. These include the possibility of losing important employees and the danger that AI models could “pose catastrophic or existential risks to humanity”.

A successful IPO is expected to encourage other AI companies to sell shares to the public. “Never have IPOs represented so much to the venture capital industry writ large,” Ravi Viswanathan, founder and managing partner of NewView Capital, a US venture capital firm, told Bloomberg.

The investment case for Anthropic depends on whether the money it earns from AI in the coming years will cover what it is spending now. Aswath Damodaran, a finance professor at New York University's Stern School of Business who specialises in valuing companies, told Bloomberg that nobody, including Amodei, knows for certain how much revenue Anthropic will earn five years from now. “They are just as lost as you or me,” he said.