Anthropic filed a public registration statement with the U.S. Securities and Exchange Commission on October 26, 2026, for a proposed initial public offering of Class A shares. The company behind Claude intends to list on the Nasdaq Global Select Market under the symbol ANT, but has not determined the number of shares or offering price range. (anthropic.com)

The filing is not a completed IPO. Anthropic’s announcement says the registration statement has not become effective and that securities cannot be sold before it does. The announcement therefore establishes a proposed offering, not a trading debut or a final fundraising amount. (anthropic.com)

The prospectus provides a detailed account of the business supporting Claude: sharply higher revenue, substantial infrastructure spending, expanding customer accounts and governance provisions that reserve powers outside the public shareholder base. Its financial statements and customer metrics cover different periods, an important distinction when assessing the company’s growth. (Source: supplied S-1 disclosures, pp. 22, 134–135, 149 and 247–248.)

Revenue growth and different measures of profitability

Anthropic reported $16.286 billion in revenue for the six months ended June 30, 2026, compared with $1.163 billion in the corresponding 2025 period. The latest interim financial statements are unaudited. For that first half, the company recorded an operating loss of $693 million and a much larger net loss of $80.748 billion. (S-1, pp. 22 and 148.)

The distinction between those losses matters. A $72.388 billion expense on the line covering fair-value changes to convertible notes, derivative liabilities and simple agreements for future equity, or SAFEs, dominated the net loss. That makes the headline net-loss figure a poor stand-alone description of the costs of running Claude: it incorporates substantial financing-related valuation charges. (S-1, p. 22.)

Anthropic also reported $1.618 billion in adjusted operating income, a 10% adjusted operating margin and $3.929 billion in operating cash inflow for the first half. These are different measures, not interchangeable descriptions of profitability. Its $974 million in second-quarter operating income under generally accepted accounting principles did not make the combined first half profitable on either the operating-income or net-income measure. (S-1, pp. 156–157 and 168.)

Compute and infrastructure expense reached $8.945 billion during the first half. Approximately 80% of revenue came from consumption and 20% from subscriptions. Together, those disclosures make usage and the cost of supplying it central to understanding the business, alongside the separate effects of financing arrangements on reported earnings. (S-1, pp. 138 and 149.)

Enterprise spending expands across products

The customer disclosures use a later measurement date: September 30, 2026. At that point, more than 11,900 customers generated at least $100,000 each in trailing-12-month revenue; more than 1,900 generated at least $1 million; and approximately 250 generated at least $10 million. These thresholds overlap and should not be added together as separate customer groups. (S-1, pp. 2 and 134.)

Among the $100,000-plus cohort, more than half used both the Claude Platform and at least one first-party application during the third quarter of 2026. That compared with approximately 30% in the fourth quarter of 2025. The applications discussed include Claude Code and Claude Cowork. The comparison documents broader product use; it does not establish that customers are unable to switch suppliers. (S-1, p. 134.)

Anthropic reported approximately 900% net revenue retention as of September 30. Despite its name, that is a revenue-cohort expansion measure, not the percentage of customers retained. It should not be read as a customer-renewal rate or a forecast that spending will keep expanding at the same pace. (S-1, pp. 2–3 and 135.)

Industry context does not establish customer returns

Broader industry research offers context, rather than independent confirmation of Anthropic’s figures. McKinsey’s August 25 survey found that 40% of respondents at organizations with more than $1 billion in annual revenue reported scaling AI agents, up from 27% the previous year. The survey collected responses from 1,719 participants across 97 countries. (mckinsey.com)

However, approximately one in five respondents said AI operating costs constrained usage, while 37% attributed some earnings-before-interest-and-tax impact to AI—essentially unchanged from the previous year. The distinction is relevant here: a supplier’s growing revenue does not, by itself, demonstrate financial returns for its customers. The survey measures broader reported experience, not outcomes specifically attributable to Claude. (mckinsey.com)

Public shares, reserved governance powers

The proposed governance structure separates ownership from several important control rights. Class F shares would carry 50.1% of voting power on matters submitted to stockholders generally, through the founder entity. The Long-Term Benefit Trust has separate board-election rights. Those provisions should not be collapsed into a single description of founder voting control. (S-1, cover and pp. 247–248; supplied governance summary.)

For prospective public shareholders, the practical issue is restricted influence over company decisions. Anthropic’s status as a public benefit corporation and its special governance arrangements are therefore part of the investment proposition, alongside revenue growth and financial performance—not incidental background to the offering. (Source: supplied S-1 corporate-information and governance disclosures.)

Concentration and operating risks remain material

The prospectus also identifies customer and distribution concentration. The five largest customers represented approximately 25% of revenue in the first nine months of 2026, while cloud marketplaces accounted for 48% in the first half. Those figures describe different periods and different forms of dependence; they should not be combined into a single concentration measure. (S-1, p. 56.)

Other disclosed risks include compute commitments, model failures and litigation. For organizations building products or workflows around Claude, the filing provides material for assessing the supplier’s finances, infrastructure dependencies and decision-making structure. It does not eliminate those risks. Nor do historical growth figures establish future results: the proposed offering remains incomplete, with its size and price still undetermined. (Source: supplied S-1 risk disclosures.) (anthropic.com)