Anthropic generated $16.3 billion in revenue during the first half of 2026, up sharply from $1.1 billion a year earlier. But the Claude developer also disclosed an $80.7 billion net loss, more than $600 billion in computing commitments and mounting risks ahead of a potential IPO, raising questions about the cost of competing in the AI market.
Anthropic’s draft S-1 prospectus reviewed by The Wall Street Journal reveals a company experiencing extraordinary demand for its artificial intelligence products while taking on enormous financial obligations to maintain its growth.
The filing comes as the company prepares for a potential stock market debut as soon as November 2026. Its disclosures give investors a clearer picture of the economics behind Claude, including the cost of computing infrastructure and the financial risks that could affect its long-term growth.
The headline loss is particularly striking, but it needs context. Approximately $72 billion of the $80.7 billion net loss came from an accounting charge tied to the increased fair value of convertible bonds that could eventually turn into Anthropic shares. That amount was not a direct cash operating expense.
Nevertheless, the filing raises questions about whether Anthropic can sustain its rapid expansion without committing increasingly large amounts of capital. The company’s growth has been a major part of the AI investment story, but the new disclosures show that demand for its products comes with substantial financial and operational risks.
Anthropic’s $80.7 Billion Loss Is Largely an Accounting Charge
Anthropic’s first-half net loss widened from $13 billion in the same period of 2025 to $80.7 billion in 2026. The increase was driven largely by the revaluation of convertible financing instruments as the company’s valuation rose.
The distinction between accounting losses and operating performance matters. The charge does not mean Anthropic spent $72 billion in cash during the period. However, it also does not mean the company has become profitable under standard accounting measures.
The business continues to spend heavily to develop and operate its AI models. Anthropic paid nearly $9 billion for computing capacity and infrastructure in the first half of 2026, compared with $2.3 billion a year earlier.
Revenue is growing quickly, but the cost of serving customers and developing more capable models remains central to the company’s financial outlook. Anthropic’s reported third-quarter revenue reached $16.8 billion, up 45% from the second quarter, according to information shared with prospective investors. That followed a more than doubling of revenue between the first and second quarters.
The slower sequential growth rate has attracted attention from investors who are trying to determine whether the company can maintain its momentum. The earlier acceleration was extraordinary, but continued expansion at that pace would be difficult to sustain indefinitely.
Anthropic has also become increasingly dependent on substantial funding to secure the computing resources required by its business. Its draft prospectus points to more than $600 billion in future computing agreements with major technology and cloud providers, with another $150 billion in potential deals under negotiation. The total could reach $750 billion over the coming years, although the additional amount is not a finalized commitment.
Those figures are especially significant given Anthropic’s reported third-quarter revenue of $16.8 billion. Revenue growth helps explain the company’s demand for infrastructure, but the future economics will depend on whether sales and margins can generate sufficient returns to justify those obligations.
Computing Costs, Customer Concentration and AI Risks Complicate Anthropic’s IPO
The prospectus identifies computing access as a central risk. Anthropic relies on outside providers for much of the infrastructure needed to train and run Claude. Those suppliers include major cloud businesses that are also developing competing AI products, creating potential tension as demand for computing capacity increases.
The company is therefore exposed to more than the cost of its existing agreements. It must also secure enough capacity to support future demand, often years before the associated infrastructure can be fully used. Delays in data-center construction, electricity shortages or changes in supplier relationships could affect its ability to scale.
The proposed agreements with infrastructure providers illustrate the scale of that challenge. Anthropic has been exploring additional capacity through deals such as the reported second $45 billion computing agreement with Nscale. That potential transaction is separate from the broader commitments detailed in the prospectus, and the negotiations have not been finalized.
Customer concentration is another concern. The filing indicates that a relatively small number of customers account for a meaningful share of Anthropic’s revenue, with its two largest customers representing roughly one-fifth of sales. Losing a major customer or seeing it reduce usage could have an outsized effect on revenue.
Anthropic also disclosed risks associated with the behavior of increasingly capable AI models. Those include the possibility of unpredictable outputs or unauthorized behavior that could harm customers or damage the company’s reputation. The prospectus further points to political opposition to new data centers, which could constrain the infrastructure expansion the business needs.
These issues could influence how investors value the company. Earlier reports suggested a potential IPO valuation of around $2 trillion, but some investors have been considering a figure closer to $1.5 trillion following the latest financial disclosures, Reuters reported. Neither figure is a confirmed offering valuation, and the eventual pricing will depend on market conditions and the company’s final IPO plans.
Anthropic is also preparing to dedicate 5% of its IPO shares to a public benefit fund, according to the draft prospectus. The proposed structure adds another consideration for investors assessing the company’s governance and use of equity.
The wider AI market is watching closely. A successful IPO could help sustain investor confidence in companies building advanced AI systems, while a weaker reception could reinforce concerns that infrastructure spending is running ahead of the returns businesses can generate. Recent pressure on semiconductor stocks and AI infrastructure valuations shows that investors are increasingly examining the financial assumptions behind the AI boom.
Anthropic’s filing does not erase the evidence of strong demand for Claude. Instead, it sets that growth against the capital requirements and operational risks involved in competing at the frontier of AI. For prospective shareholders, the central question is whether the company can turn rapid sales growth into durable profits while meeting its enormous computing needs.
Until Anthropic provides more clarity on its final offering valuation, spending commitments and longer-term financial outlook, the draft prospectus presents a mixed picture: an AI business growing at remarkable speed, but one that will require significant resources to sustain that growth.