Anthropic released its initial public offering prospectus two days ago, providing a detailed view of the AI firm’s financials, business model and governance. The document shows a company that has built a fast‑growing revenue engine around its Claude family of models, yet continues to post substantial losses and faces a set of investor‑focused risks.
Business model and revenue growth
Anthropic describes itself as a research lab that monetises artificial‑intelligence models, most notably Claude. The firm sells access to these models directly to users and also offers a suite of products that embed Claude into broader workflows. The Claude Platform enables businesses and developers to integrate the models into custom applications, while ready‑to‑use tools such as Claude Code and Claude Cowork target software development and professional research tasks respectively.
This layered approach is intended to create a feedback loop: as customers adopt Claude for more functions, they generate higher usage, which in turn provides Anthropic with data to refine its models. The company reports that roughly 80 % of its revenue comes from usage‑based fees measured in tokens – the units of text or data the model processes – while the remaining 20 % is derived from subscription plans that grant recurring access.
Revenue numbers illustrate the speed of adoption. For the fiscal year 2025 (January–December 2025), Anthropic recorded $4.5 billion in revenue, a rise of more than 1,000 % compared with the prior year. In the first half of fiscal year 2026, revenue surged to $16 billion, an increase of about 1,300 % over the same period a year earlier. The prospectus notes that a handful of large customers contributed a quarter of total revenue in the most recent full year, and that cloud marketplaces supplied nearly half of the first‑half‑2026 revenue.
Financial performance and cash position
Despite the headline‑grabbing revenue growth, Anthropic’s profitability remains elusive. In FY25 the company posted an operating loss of $8 billion and a net loss of $41 billion. The first half of FY26 showed an operating loss that narrowed to $693 million, but the net loss ballooned to $80 billion – almost double the full‑year net loss recorded a year earlier.
The prospectus attributes the widening net loss in part to a non‑cash accounting adjustment. Over time, investors have entered contracts that can convert into equity when Anthropic’s valuation rises. Accounting rules require the firm to record the increase in the fair value of these potential shares as an expense. In the first half of FY26, this adjustment added $72 billion to the net loss figure.
Running large AI models is capital intensive. Anthropic spent close to $9 billion on computing infrastructure during the first half of 2026, and approximately $3 billion on employee compensation, including researchers, as well as fees paid to cloud providers and other partners. The filing states that the company holds a cash balance exceeding $130 billion, a figure that underscores the scale of its financing needs even as it raises additional capital through the IPO.
Governance and risk factors
Anthropic is organized as a Public Benefit Corporation (PBC), a structure that limits the influence of ordinary shareholders. The founders – seven in total, including CEO Dario Amodei and President Daniela Amodei – created an entity called Founder LLC that controls 50.1 % of voting power on matters presented to shareholders. Regular shareholders collectively hold the remaining 49.9 % of votes, meaning that control over strategic decisions remains firmly in the hands of the founders and the trust that backs the PBC.
The prospectus highlights three primary risks. First, the limited voting power could constrain investors’ ability to affect the company’s direction, even if they hold a substantial share of equity. Second, revenue concentration poses a vulnerability: two customers accounted for 25 % of FY25 revenue, and reliance on cloud marketplaces – some owned by major investors such as Amazon, Google and Microsoft – contributed nearly half of H1‑FY26 revenue. Third, several of Anthropic’s biggest investors are also key suppliers of computing power and chips. This dual role creates a potential conflict of interest, as those investors might prioritize their own AI initiatives when compute capacity is scarce.
Anthropic acknowledges that it has at times requested more computing capacity than partners were willing to provide, underscoring the dependency on supplier‑investors. The company argues that the PBC structure allows it to focus on long‑term AI safety rather than short‑term financial pressures, but the prospectus makes clear that this governance model and the intertwined relationships with cloud providers are material considerations for prospective shareholders.
In summary, Anthropic’s prospectus paints a picture of a company that has achieved explosive top‑line growth while still wrestling with deep losses, significant cash outlays for infrastructure, and a governance model that curtails ordinary shareholder influence. Investors will need to weigh the potential of a rapidly expanding AI platform against the accounting adjustments, loss profile and structural risks outlined in the filing.
Be First to Comment