3 charts on Anthropic’s historic IPO filing
As Anthropic prepares to sell stock to the public, the S-1 reveals important information for investors on revenue growth, profitability, and adoption.
For three years, the stock market has been driven by the historic artificial intelligence revolution. Now, Anthropic—one of the central players in this boom—is ready to sell stock to the public. However, what will likely be the largest IPO in history comes as investors question AI stock valuations. In particular, there’s uncertainty around Anthropic and other key AI players’ ability to generate revenue that can support their massive infrastructure investments and financial obligations.
While Anthropic’s Form S-1, filed with the Securities and Exchange Commission on Oct. 26, did not provide information on how many shares the company expects to sell, or at what price (that will be revealed in coming weeks), the firm offered key information for investors considering whether to buy its stock at the IPO. Among the important metrics: revenue growth, profitability including free cash flow, and enterprise adoption.
Here’s a look at three critical pieces of information from the S-1 for investors to focus on.
Anthropic’s revenue growth and client base
Anthropic’s S-1 filing states that the company has more than 11,900 customers with $100,000 or more in trailing 12-month revenue as of Sept. 30, up from more than 950 a year prior.
Morningstar senior equity analyst Malik Ahmed Khan says the S-1 filing “shows real adoption” of the firm’s AI assistant, Claude, “which is growing fast.” He explains: “It’s pretty clear, revenue growth has been sky-high. This is the fastest scaleup of any technology business that we’ve seen.” He adds that enterprise software customers who spend more tend to be stickier. That said, “we don’t have retention data for Anthropic yet on any of its customer cohorts.”
PitchBook senior analyst Harrison Rolfes points to another key stat: The number of enterprise customers spending $100 million per year “grew more than twelvefold” in the year to September.
In addition, the S-1 shows Anthropic sports 900% net revenue retention. That metric compares what customers spent in the third quarter with what they spent a year earlier, when the company brought in about $1.4 billion a quarter. However, Rolfes says that against a base that small, the net retention percentage will come down sharply as the comparison catches up with Anthropic’s current scale.
That said, Rolfes points to a more durable measure of Anthropic’s business. Among publicly traded Fortune 500 companies spending at least $100,000 a year with Anthropic, spending averaged less than 0.1% of their total expenses in the third quarter, according to the S-1. Even the top 10 Fortune 500 customers spend less than 1% of their total expenses on Anthropic, he explains—more than eight times the average share. If the typical large customer moves even partway toward that level, there’s a lot of revenue left in accounts Anthropic already has, he says.
Anthropic’s compute efficiency
Anthropic’s revenue per dollar spent on compute has also been growing, according to the S-1. “AI labs like Anthropic have more leverage when scaling on compute versus scaling on headcount,” Khan says. “Quarterly revenue grew to $12 billion from $800 million, and headcount did not multiply at that rate. Revenue per employee has been increasing substantially.” He states that revenue per employee has grown to over $6.6 million, compared with $1.5 million as of September 2025.
Khan says that for Anthropic to continue growing revenue, it needs to bring more compute online and keep increasing its efficiency in scaling compute. “It’s been very interesting and positive for Anthropic so far.”
PitchBook’s Rolfes notes that part of Anthropic’s growing compute efficiency comes from timing. “Compute is contracted 12–18 months ahead and expensed as it comes online,” he says. “Management says rising use of short-term compute will compress margins in the near term. Anthropic’s arrangement with SpaceXAI alone runs about $3.1 billion per month once fully delivered, which is expected in December 2026.” By Rolfes’ calculation, “That works out to roughly $9.3 billion a quarter, against $5.6 billion of total compute expense in the second quarter.”
Rolfes says second-quarter 2026 compute efficiency should be viewed more like a margin high-water mark than a run rate: “Third-quarter revenue is blank in this filing, and the amendment that fills it in is the first test of that call.”
Operating efficiency and margin expansion
The S-1 showed that Anthropic was profitable on a GAAP basis for the first half 2026, Khan explains. “It’s quite incredible.” On an adjusted basis, the company delivered an adjusted operating margin of 19%.
Anthropic is also free cash flow positive, Khan says: “Anthropic has raised about $140 billion to date, and they still have more than $130 billion left. They’ve burned only about $9 billion so far.” Overall, Khan says the strongest data revealed in Anthropic’s S-1 concerns rapid revenue growth, compute efficiency, and operating efficiency in terms of revenue per employee.
Khan remains skeptical about upcoming compute costs. “A lot of Anthropic’s compute costs still haven’t technically come online,” he says. “The company has multiyear commitments with Google, Amazon, and Broadcom … and there is a world in which after compute costs come online, revenue disappoint. Anthropic would still be liable to pay these billion-dollar commitments back.”
Khan says that if revenue can’t support those commitments, Anthropic would be pushed back into operating losses. “The key risk is that revenue growth doesn’t turn out to be durable,” he explains. Morningstar’s view is that AI labs such as Anthropic and OpenAI are building durable business models which “should prevent such an outcome … but that is the biggest uncertainty.”
