Anthropic's 407 page IPO prospectus has two positives Wall Street is zeroing in on

Key Points

• Anthropic's IPO will be an important test for investors betting on artificial intelligence.

• The company reportedly aims to raise $100 billion at a valuation of more than $2 trillion.

• The S-1 filing explains the company's finances, spending commitments, reliance on major customers, risks, and other important details.

After months of waiting and several delays, Anthropic has finally released its S-1 registration statement. This document gives the public a closer look at one of the companies building advanced large language models (LLMs), which are AI systems that can understand and generate language.

The company behind the Claude family of LLMs plans to list its shares on the Nasdaq stock exchange under the ticker symbol ANT. Anthropic's IPO will be an important test for the next stage of investment in artificial intelligence stocks. Reports say the company is preparing to raise $100 billion at a valuation of up to $2 trillion.

Here are five important numbers from Anthropic's S-1 filing.

$11.6 Billion

Anthropic made $11.6 billion in revenue in the second quarter of 2026, which ended on June 30. That was much more than the $787 million it made in the second quarter of 2025 and the roughly $4.7 billion it made in the previous quarter.

In the first six months of 2026, Anthropic made nearly $16.3 billion in revenue. That was 12.5 times as much as it made during the same period in 2025.

Most of Claude's revenue comes from customers paying based on how much they use it. Companies add Claude to their own products and work systems. Individual customers and businesses can also pay for subscriptions that give them access to Claude models.

$637 billion

One reason AI is so expensive is that it needs a lot of computing equipment and physical space to process huge amounts of data. AI companies need data centers filled with graphics processing units (GPUs) and other equipment. They also need special systems, such as liquid cooling, to keep chips from getting too hot.

Anthropic's S-1 lists possible spending commitments for these resources, although the company is not fully locked into every listed amount. The following are the specific spending commitments it named for computing power and GPUs, covering the period from now through 2036:

• $161.2 billion for five years of access to computing capacity powered by Tensor Processing Units (TPUs) through Broadcom. This agreement cannot be canceled unless a default occurs.

• $120.5 billion to access Nvidia GPUs owned by Space Exploration Technologies Corp. The agreement lasts until 2029, but either party can cancel it with 90 days' notice.

• $114.5 billion in promised spending on cloud computing with Google from 2026 through 2033, regardless of how much computing power Anthropic uses.

• $110 billion in promised spending on cloud computing with Amazon from 2026 through 2036, regardless of how much computing power Anthropic uses.

• $44.9 billion in promised spending on cloud computing with Microsoft from 2026 through 2033.

• $44.6 billion to access Nvidia GPUs from Nscale Global Holdings from 2026 through 2033. The agreement cannot be canceled except in certain limited situations.

• $41.3 billion to access Nvidia GPUs from Lambda from 2026 through 2034. The agreement cannot be canceled except in certain limited situations.

These figures cover only the specific deals Anthropic named. Some agreements, such as the one involving SpaceX, can be canceled with fairly little notice. Anthropic's total spending could also end up being much higher.

For example, Anthropic said it had agreed to buy additional TPU computing capacity from companies such as Google between 2028 and 2031. This is expected to cost between $125 billion and $150 billion. The S-1 also mentions other deals but provides fewer details about them.

79%

In the third quarter of 2026, which ended on September 30, customers outside Anthropic's five biggest customers accounted for 79% of its revenue. Also, no single customer accounted for more than 10% of the company's revenue during the nine months ending September 30.

A major concern for many AI companies is that too much of their revenue comes from just a few customers. This makes them more vulnerable if one of those customers stops buying from them or runs into problems.

I expect investors to view Anthropic's figures positively because its revenue comes from a wide range of customers. As of September 30, the company said it had more than 11,900 customers who had each spent over $100,000 with it during the previous 12 months.

85%

Anthropic reported a gross margin of 85% for the first half of 2026, based on the cost of revenue it reported. Gross margin is the share of revenue left after subtracting the direct costs of providing a product or service. An 85% margin is very high and suggests that Anthropic may have strong power to set prices. By comparison, Nvidia's gross margin has been in the mid-70% range.

Even so, investors will likely ask whether Anthropic's reported cost of revenue, which mainly includes computing and infrastructure costs, accounts for all the expenses required to run its AI models.

For example, the models need regular training, which is expensive. Anthropic reported about $9.7 billion in research and development costs for the first half of 2026. These costs mainly covered model development, research, and computing work done inside the company.

Anthropic also pays business partners, including Amazon, Google, and Microsoft, to help sell its models through their online marketplaces. These payments are included in sales and marketing costs, which totaled $3.8 billion in the first half of 2026.

Research and development costs and sales and marketing costs are not included in the company's cost of revenue.

94

The number 94 refers to the pages devoted to risk factors in Anthropic's prospectus. Investors have noticed that this is a large section of an S-1 filing that runs for more than 400 pages in total.

This is especially important because people are increasingly worried about the wider safety risks of AI. Recent hacking incidents involving AI agents have made these concerns more noticeable.

Not all 94 pages focus on AI safety. However, the S-1 warns that failing to put proper protections, or safeguards, around its models could damage Anthropic's reputation, brand, or financial condition. It also identifies the risk that its models could behave in ways that are unintended, do not match their intended goals, or act on their own.

*Stock Advisor returns as of October 26, 2026.

Bram Berkowitz does not own any of the stocks mentioned. The Motley Fool owns shares in and recommends Amazon, Broadcom, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.