Anthropic develops artificial intelligence models, including Claude, and generates revenue by selling access to them. However, its business extends beyond the models themselves, encompassing a range of products and services designed to integrate its technology into the daily operations of individuals, businesses and developers.
The company describes itself as a “research lab”. Its business model begins with developing AI models and selling access to them. The Claude Platform allows businesses and developers to use these models to build their own products and workflows. Anthropic also offers ready-to-use applications, including Claude Code, which supports software development, and Claude Cowork, which handles professional tasks such as researching and analysing information.
Together, these offerings are designed to encourage customers to incorporate Claude into their everyday operations, build products around it and delegate entire workflows to AI. As customers identify additional uses for Anthropic’s products, their spending can increase. Greater adoption also provides the company with more opportunities to understand how its products are being used and improve them, creating a self-reinforcing growth cycle.
Improved products can attract new users and encourage existing customers to use Claude for more tasks. That, in turn, can increase spending and provide Anthropic with additional resources to develop its products further.
Anthropic generates revenue through two primary channels: usage-based payments and subscriptions. Under the usage-based model, customers pay according to how much they use Claude, measured in tokens. A token is a unit of text or other information processed by an AI model. The more information Claude processes and generates, the more tokens it consumes. Under the subscription model, individuals and businesses pay recurring fees for specified levels of access and usage.
In FY25, covering January to December 2025, Anthropic generated $4.5 billion in revenue, more than 1,000% higher than in the previous year. Revenue increased to $16 billion in the first half of FY26, representing growth of around 1,300% compared with the same period a year earlier. Usage-based payments currently account for roughly 80% of its revenue, while subscriptions contribute the remaining 20%.
However, this rapid revenue growth has not yet translated into consistent profitability. Anthropic reported an operating loss of $8 billion and a net loss of $41 billion in FY25. During the first half of FY26, its operating loss narrowed to $693 million, but its net loss ballooned to $80 billion, almost twice the loss recorded throughout FY25 in just six months.
One major reason is the cost of operating advanced AI models. Anthropic spent nearly $9 billion on computing and infrastructure in the first half of 2026 alone. It also spent around $3 billion each on employees, including researchers, and fees paid to partners such as cloud service providers. These expenses add to the financial burden of maintaining operations while expanding the business.
The company is also raising money to support its ambitious growth plans. Despite holding more than $130 billion in cash, Anthropic needs additional funding to finance its expansion and cannot rely on investor money alone to continue growing.
However, operating expenses are not the only reason for its substantial net loss. An accounting adjustment has also played a significant role.
Over the years, investors have provided funding through contracts that can eventually convert into shares. When these contracts were signed, Anthropic's valuation was considerably lower. As the company's valuation increased, the shares investors could eventually receive became more valuable. Accounting rules require Anthropic to record that increase in value as an expense, even though it does not involve an actual cash outflow.
This adjustment contributed $72 billion to Anthropic's net loss in the first half of FY26 alone.
The operating figures present a different picture. Anthropic's operating loss in the first half of FY26 was substantially lower than the $2 billion it reported in the first half of FY25, suggesting that its underlying operating economics are improving despite the enormous net loss.
Whether Anthropic justifies the valuation it is reportedly seeking remains difficult to determine from the prospectus alone, as the document does not disclose the IPO size or price range. The valuation debate has also been examined separately in an analysis of whether Anthropic is overvalued.
Beyond its financial performance, the company's IPO prospectus devotes more than 90 pages to risk factors. Three concerns stand out: limited shareholder control, revenue concentration and dependence on investors that also supply critical technology infrastructure.
The first risk concerns corporate governance. Buying Anthropic shares will not necessarily give investors substantial influence over how the company operates, regardless of how many shares they own. Anthropic is incorporated as a Public Benefit Corporation (PBC), with a governance structure that limits the control ordinary shareholders might otherwise expect.
The company's founders and a separate trust will retain most of the control. Anthropic has seven co-founders, including CEO Dario Amodei and President Daniela Amodei. They have established an entity called Founder LLC, which will hold 50.1% of the voting power on matters put to shareholders generally. Holders of regular shares will have the remaining 49.9%.
Anthropic says this arrangement is intended to protect its long-term focus on responsible AI development and safety research from pressure to deliver short-term financial results.
For example, the company may decide to spend billions of dollars on AI safety research, while investors may prefer greater spending on products that generate revenue more quickly. Retaining voting control would allow the founders to pursue their strategy even if some shareholders disagree.
The structure could support the company's stated mission if its founders make sound decisions and remain committed to it. However, it also gives investors fewer ways to challenge management if the company encounters problems.
The second major risk is revenue concentration. Two customers accounted for a quarter of Anthropic's revenue last year. In addition, nearly half of its revenue in the first half of FY26 came through cloud marketplaces.
Some of these marketplaces belong to major investors, including Amazon, Google and Microsoft. These companies also provide the computing power Anthropic needs to operate its AI models. Consequently, Anthropic depends on businesses that have a financial interest in its success but also have their own commercial priorities.
The third risk arises from the overlap between Anthropic's investors and its suppliers. Amazon, Google, Microsoft and Broadcom have financial ties with the company while also supplying computing capacity and chips required to run its AI models. This creates potential conflicts of interest.
Amazon, Google and Microsoft also develop their own AI models. If computing capacity becomes scarce, these companies could prioritise their own products over Anthropic's requirements. Anthropic has acknowledged that it has requested more computing capacity than some of its partners were willing to provide.
There is also a financial relationship between the investments these companies make in Anthropic and the payments they receive from it. Some of the money invested in the company can flow back to its investors through payments for computing capacity and chip leases. This makes it more difficult for investors to determine how much of Anthropic's growth reflects genuine customer demand.
If Anthropic encounters financial or operational difficulties, its suppliers could prioritise protecting their own investments and contractual interests rather than the interests of public shareholders.
Anthropic's prospectus therefore presents a business with rapidly expanding revenue and signs of improving operating performance, but also substantial costs, a significant accounting-driven net loss and structural risks that prospective investors will need to assess. Its governance arrangements, dependence on a concentrated customer base and reliance on investors that also supply critical infrastructure are central considerations as the company moves towards a potential public listing.
About The Author
Comment List