3 Charts on Anthropic’s Historic IPO Filing

For the past three years, the stock market has been shaped by the huge growth of artificial intelligence (AI). Now, Anthropic, one of the main companies in this boom, is getting ready to sell shares to the public. This could be the largest initial public offering (IPO) in history. But investors are questioning whether AI stocks are worth their high prices. In particular, they are unsure whether Anthropic and other major AI companies can earn enough money to cover their huge spending on computing equipment and their financial commitments.

Anthropic filed its Form S-1 with the Securities and Exchange Commission on Oct. 26. The filing did not say how many shares the company plans to sell or what each share will cost. Those details are expected in the coming weeks. But the filing did give investors important information to help them decide whether to buy shares in the IPO. Key measures include revenue growth, profits and free cash flow, which is money left after paying for operating needs and investments, as well as how widely businesses use Anthropic's products.

Here are three important areas investors should examine in the S-1 filing.

Anthropic's Revenue Growth and Client Base

Anthropic's S-1 filing says that, as of Sept. 30, the company had more than 11,900 customers that each brought in at least $100,000 in revenue over the previous 12 months. A year earlier, it had more than 950 such customers.

Malik Ahmed Khan, a senior equity analyst at Morningstar, says the filing shows that Anthropic's AI assistant, Claude, is being widely adopted and growing quickly. He says the company's revenue growth has been extremely fast, making it the fastest-growing technology business of its kind that he has seen. He also explains that business customers who spend more money on software tend to keep using it. However, he says Anthropic has not yet provided data showing how many customers in each group continue to use its products over time.

Harrison Rolfes, a senior analyst at PitchBook, points to another important figure: The number of business customers spending $100 million a year grew more than twelvefold in the year leading up to September.

The S-1 also reports that Anthropic's net revenue retention rate is 900%. This measure compares how much customers spent in the third quarter with how much those same customers spent a year earlier, when Anthropic was bringing in about $1.4 billion per quarter. Rolfes warns that this percentage will fall sharply as the comparison catches up with Anthropic's much larger business today, because the starting revenue figure was so small.

Still, Rolfes points to another measure that may give a clearer picture of Anthropic's future growth. According to the S-1, publicly traded companies in the Fortune 500 that spent at least $100,000 a year with Anthropic spent, on average, less than 0.1% of their total expenses on the company in the third quarter. Even Anthropic's 10 biggest Fortune 500 customers spent less than 1% of their total expenses on it. That share was more than eight times the average. Rolfes says Anthropic could earn much more from its existing customers if typical large customers increased their spending even partway toward the level of its biggest customers.

Anthropic's Compute Efficiency

The S-1 shows that Anthropic has been earning more revenue for each dollar it spends on computing power, also called compute. Khan says AI companies such as Anthropic can often grow their computing power more effectively than they can grow their workforce. Quarterly revenue rose from $800 million to $12 billion, while the number of employees did not grow nearly as quickly. As a result, revenue per employee has increased substantially. It has risen to more than $6.6 million, compared with $1.5 million as of September 2025.

Khan says Anthropic must bring more computing power online and keep finding ways to use it more efficiently if it wants revenue to keep growing. So far, he says, the results have been very interesting and positive for the company.

Rolfes says part of Anthropic's improved computing efficiency comes from the timing of its costs. The company signs contracts for computing power 12–18 months before it uses it, and records the expense as the computing power becomes available. He says management expects greater use of short-term computing resources to reduce profit margins in the near future. Once fully delivered, the company's arrangement with SpaceXAI alone is expected to cost about $3.1 billion per month. Full delivery is expected in December 2026. Rolfes calculates that this would equal about $9.3 billion per quarter, compared with total compute expenses of $5.6 billion in the second quarter.

Rolfes says Anthropic's computing efficiency in the second quarter of 2026 should be seen as a possible high point for its profit margins, rather than a level it can expect to maintain. The filing does not include third-quarter revenue, he notes. The next version of the filing, which will add that figure, will provide the first test of this view.

Operating Efficiency and Margin Expansion

Khan explains that the S-1 showed Anthropic made a profit under generally accepted accounting principles (GAAP) in the first half of 2026. He says this is remarkable. Using adjusted figures, the company reported an adjusted operating profit margin of 19%.

Khan also says Anthropic is generating positive free cash flow, meaning more cash is coming in than is going out after relevant spending. The company has raised about $140 billion so far and still has more than $130 billion left. It has spent only about $9 billion more than it has brought in, he says. Overall, Khan says the strongest findings in the S-1 are Anthropic's rapid revenue growth, its more efficient use of computing power, and its higher revenue per employee.

Khan remains concerned about future computing costs. He says much of Anthropic's contracted computing power has not yet become available, so the related costs have not fully appeared in its financial results. The company has multiyear commitments with Google, Amazon, and Broadcom. If those costs arrive but revenue falls short of expectations, Anthropic would still have to pay the billions of dollars it has committed to spend.

Khan says that if Anthropic's revenue cannot cover these commitments, the company could start losing money from its operations again. He explains that the main risk is that its revenue growth may not last. Morningstar believes that AI companies such as Anthropic and OpenAI are building businesses that can remain strong over time, which should help prevent this outcome. However, he says this remains the biggest uncertainty.