Revenue for the first half of the year more than tripled compared to the entire previous year, yet the net loss reached a staggering $80.748 billion. On October 26, Anthropic, PBC (hereinafter referred to as 'Anthropic'), a U.S. artificial intelligence company, publicly submitted its registration documents for an initial public offering (IPO) to the U.S. Securities and Exchange Commission (SEC), disclosing this starkly contrasting financial performance. According to the prospectus, in the first half of 2026, the company achieved revenue of $16.286 billion, approximately 14 times that of the same period in the previous year, while its operating loss narrowed from $2.135 billion to $693 million.

The enormous net loss primarily stemmed from changes in the fair value of financing instruments and does not directly reflect the company's actual operating performance. As the number of enterprise customers increases and the demand for model usage expands, Anthropic's profitability is improving. However, facing competitors like OpenAI and Google, which continue to launch new models and adjust prices, along with substantial long-term computing power procurement contracts already signed, whether the company can sustain revenue growth and convert it into stable profits remains a critical challenge for this IPO.

01 Surge in Enterprise Customers: Half-Year Revenue More Than Triples Last Year's Total

Anthropic's revenue has sustained growth (continued to grow rapidly) over the past three years.

The prospectus shows that in 2023, Anthropic achieved revenue of $41 million, which grew to $386 million in 2024 and further to $4.585 billion in 2025.

Entering 2026, this growth momentum accelerated further. In the first half of the year, the company achieved revenue of $16.286 billion, approximately 14 times the $1.163 billion recorded in the same period of 2025 and more than triple the total revenue for the entire previous year. Among this, revenue in the first quarter reached $4.731 billion, with further growth to $11.555 billion in the second quarter.

The procurement and usage of artificial intelligence models by enterprise customers are significant drivers of Anthropic's revenue growth.

Unlike relying solely on fixed subscription fees, Anthropic offers both usage-based billing and subscription services. Enterprises can invoke the Claude model through application programming interfaces (APIs) or cloud computing platforms and pay fees based on actual usage.

The prospectus shows that in the first half of 2026, Anthropic's revenue from usage-based billing reached $13.064 billion, accounting for approximately 80% of total revenue, compared to $983 million in the same period of 2025. Subscription revenue grew from $180 million in the same period of the previous year to $3.222 billion, accounting for about 20% of total revenue.

Both types of revenue have achieved substantial growth, but usage-based billing contributed approximately 80% of the year-over-year revenue increase. This business primarily charges based on metrics such as the volume of data processed by the model, with higher fees typically incurred as clients invoke the model more frequently.

As enterprises apply Claude to more practical business scenarios, the number of Anthropic's major clients has also rapidly increased.

The prospectus disclosed that as of September 30, 2026, the number of clients contributing over $100,000 in revenue to Anthropic in the past 12 months exceeded 11,900, compared to just over 950 in the same period of 2025. Among them, over 1,900 clients contributed more than $1 million in revenue, with approximately 250 clients contributing over $10 million.

From the perspective of revenue scale and client numbers, Anthropic's commercialization has transitioned from early market expansion to a stage where enterprise customers are continuously increasing their procurement. Notably, code development is a key area where it has established a competitive advantage.

Anthropic's Claude Code assists developers in tasks such as code generation, modification, and debugging, gradually integrating into enterprise software development processes. Compared to general conversational applications, the demand for code development is more specific, allowing enterprises to evaluate the actual value of related tools based on development efficiency.

According to the 2025: The State of Generative AI in the Enterprise report released by U.S. venture capital firm Menlo Ventures in December 2025, based on a survey of AI usage among U.S. enterprises and market model estimates, Anthropic's market share in the enterprise large model API expenditure market reached 40%, higher than OpenAI's 27% and Google's 21%.

From a market perspective, Anthropic's estimated share increased from 12% in 2023 to 40% in 2025, while OpenAI's share declined from 50% to 27% during the same period. In the code development-related model market, Anthropic's estimated share reached 54%, compared to OpenAI's 21%. Menlo Ventures believes that the widespread adoption of Claude Code is one of the key factors driving Anthropic's increased market share.

Payment data from Ramp, a U.S. enterprise expense management platform, also reflects competitive changes between the two companies in the enterprise market.

Ramp's Ramp AI Index research released on October 14 showed that in September 2026, among its platform's enterprise clients, 43.8% paid fees to Anthropic, higher than OpenAI's 40.6%.

However, the growth dynamics differed between the two companies that month. Compared to August, Anthropic's enterprise paid adoption rate increased by 0.25 percentage points, while OpenAI's increased by 1.12 percentage points, narrowing the gap between the two.

Ramp Chief Economist Ara Kharazian pointed out in the report that as enterprise AI paid adoption rates increase, growth in new clients is slowing. Future market growth will rely more on existing clients expanding their usage rather than simply increasing the number of adopting enterprises.

The expansion of enterprise clients has driven rapid revenue growth for Anthropic, but providing model services also requires continuous investment in computing resources. After the revenue scale continues to expand, the company's profit performance presents a different picture.

02 Net Loss Exceeds $80 Billion, but Operating Loss Narrows Significantly

The prospectus shows that from 2023 to 2025, Anthropic reported net losses of $1.531 billion, $8.309 billion, and $41.965 billion, respectively. In the first half of 2026, the net loss further expanded to $80.748 billion, far exceeding the same period (same-period) revenue of $16.286 billion.

The primary cause of the enormous loss was not model research and development or daily operational expenses but rather accounting valuation changes in financing instruments.

In the first half of 2026, Anthropic recognized a loss of $72.388 billion due to changes in the fair value of financing instruments such as convertible notes, derivative liabilities, and Simple Agreements for Future Equity (SAFEs), accounting for nearly 90% of the same period (same-period) net loss. Additionally, the company recognized other net expenses of $8.757 billion, primarily related to the initial fair value recognition and subsequent valuation changes of financing-related agreements.

These losses are closely tied to Anthropic's financing structure. The company previously raised funds through convertible notes and other means, with the value of some instruments adjusting in response to changes in the company's valuation. When Anthropic's valuation increases, the fair value of corresponding financial liabilities may also rise, resulting in losses on the income statement. However, this does not mean the company actually paid an equivalent amount of cash during the period.

Beyond the enormous book losses, Anthropic's operating profit showed significant improvement.

In the first half of 2025, Anthropic reported an operating loss of $2.135 billion, with an operating loss rate of approximately 183.6%. In the same period of 2026, the operating loss narrowed to $693 million, with the operating loss rate dropping to 4.3%.

From a quarterly perspective, in the first quarter of this year, Anthropic still recorded an operating loss of $1.667 billion, but in the second quarter, it achieved an operating profit of $974 million, marking the first time the company reported a positive quarterly operating profit during the period covered by the prospectus.

This change primarily resulted from revenue growth outpacing the increase in operating expenses.

The prospectus shows that in the first half of 2026, Anthropic's total operating expenses reached $16.979 billion, compared to $3.298 billion in the same period of 2025, representing approximately a 4.1-fold increase. However, revenue grew approximately 13-fold during the same period. Revenue increased by $15.123 billion, while operating expenses increased by $13.681 billion, with the difference corresponding to a $1.442 billion year-over-year reduction in operating losses.

Among these, computing power and infrastructure remained Anthropic's largest expense items.

In the first half of 2025, the company's computing power and infrastructure expenses were $2.334 billion, increasing to $8.945 billion in the same period of 2026, a 283% year-over-year increase, even surpassing the full-year total of $7.326 billion in 2025.

However, as revenue expanded rapidly, the proportion of these expenses relative to revenue decreased from approximately 200.7% in the first half of 2025 to 54.9%.

Anthropic explained in the prospectus that the growth in computing power expenses was primarily driven by the sequential delivery of computing resources under previously signed cloud service contracts. These expenses are typically recognized when computing power becomes available for use, rather than at the time of contract signing. As revenue grows, the proportion of pre-locked computing power costs in revenue correspondingly declines.

According to the functional classification of operating expenses in Note 8 of the prospectus, Anthropic's gross profit margin also showed significant improvement.

In the first half of 2026, the company's revenue cost was $2.380 billion, resulting in a gross profit margin of approximately 85.4%, compared to about 12.4% in the same period of 2025. During the same period, Anthropic's research and development expenses reached $9.741 billion.

Revenue costs primarily include computing resources and related expenses directly used to serve clients, while research and development expenses cover investments in model research and development. These two categories, classified differently from the computing power and infrastructure expenses listed by nature, cannot be double-counted.

In addition to computing power, personnel and sales expenses also grew rapidly.

In the first half of 2026, Anthropic's personnel expenses increased from $538 million in the same period of the previous year to $3.152 billion, with approximately $2.2 billion attributable to equity incentive expenses. Other operating and administrative expenses grew from $306 million to $1.714 billion.

More significant growth was observed in sales, marketing, and partnership expenses, which increased from $120 million in the same period of the previous year to $3.168 billion. The company attributed approximately $2.7 billion of this year-over-year increase to growth in channel partner expenses in the prospectus.

This change is related to Anthropic's sales model. In the first half of 2026, approximately 48% of the company's revenue was generated through cloud service provider marketplaces. These platforms help Anthropic reach enterprise clients, but the company is required to pay collaboration fees as per the agreements.

In terms of cash flow, in the first half of 2026, Anthropic's net cash flow from operating activities reached $3.929 billion, compared to a net outflow of $1.513 billion in the same period of the previous year.

The prospectus shows that this change, in addition to improved operating performance, was also influenced by factors such as increased accounts payable, reduced prepaid expenses, and growth in deferred revenue. During the same period, the company's accounts receivable also increased significantly with business expansion.

Overall, revenue growth outpacing operating expense growth has significantly narrowed Anthropic's operating losses and improved its operating cash flow. However, with competitors continuously launching new models and changes in customer demand and pricing, whether this operational improvement can be sustained remains a critical question.

03 Intensifying Model Competition: Long-Term Computing Power Investment Remains a Challenge

The prospectus lists OpenAI and Google as Anthropic's primary competitors in the developer and enterprise client markets. The company also notes that some competitors possess greater financial strength, technical resources, and more abundant computing power supplies.

Among them, OpenAI is continuously expanding its enterprise client base and code development market, while Google owns the Gemini model, a cloud computing platform, and self-developed computing chips. For Anthropic, competition extends beyond model capabilities and product pricing to include customer channels, computing resources, and sustained funding for research and development.

This competition is already reflected in enterprises' behavior when switching model suppliers.

Research released by Ramp on October 6 showed that as of September 2026, the proportion of enterprises in its sample switching AI model suppliers, calculated on a three-month moving average, reached 8%, a new high for this statistic.

The report pointed out that after Anthropic launched Claude Opus 4 and Sonnet 4 in May 2025, it attracted more enterprises to adopt its models. Subsequently, the frequency of supplier switches gradually decreased, but in June 2026, OpenAI launched new models and lowered prices, once again altering the competitive landscape.

Further research released by Ramp on October 14 showed that for the week ending October 11, Token usage in its Token expenditure management sample increased by 52.4% compared to the week of September 13, while spending increased by only 23%.

In this sample, the growth in the number of Tokens used by enterprises significantly outpaced actual spending. Changes in model pricing, usage structure, and computing efficiency mean that large model companies may not necessarily achieve revenue growth proportional to usage increases.

The ability of enterprises to adjust suppliers based on model performance and pricing requires Anthropic to continuously launch competitive new products.

The prospectus disclosed that in the third quarter of 2026, over 80% of the company's revenue came from model versions released within the previous six months. Anthropic stated that after launching new models, most clients gradually transition to more powerful versions.

The continuous launch of new models requires sustained investment in research, development, and training resources, while computing power procurement typically needs to be arranged in advance.

The prospectus shows that as of June 30, 2026, Anthropic's future fixed or minimum procurement commitments were approximately $265.6 billion, with contract terms ranging from 1 to 20 years. The company separately disclosed that the minimum amount of its computing power procurement agreement with Amazon Web Services was at least $110 billion, while its related procurement commitments with Google Cloud were approximately $114.5 billion. These agreements fall under different disclosure scopes, and their amounts cannot be simply added together.

Long-term procurement can secure the computing resources needed for future model research and development and client usage, but some contracts include minimum payment obligations that do not decrease synchronously with reductions in client usage.

In contrast, most of Anthropic's revenue comes from model services billed based on actual usage, and some enterprise clients have no long-term procurement commitments, allowing them to adjust suppliers based on model performance, pricing, and their own needs.

If market competition causes clients to switch to other models or if model price declines slow revenue growth, Anthropic will still need to bear some of the pre-agreed computing power expenses.

This competitive and cooperative relationship is particularly evident among major cloud service providers.

Amazon, Google, and Microsoft not only provide computing resources for Anthropic but also sell Claude services through their respective cloud platforms while competing with Anthropic in certain AI businesses.

Take Google as an example. It not only possesses the Gemini model and cloud computing infrastructure, enabling it to provide computing power to Anthropic, but also directly sells its own AI models to corporate clients.

For Anthropic, cloud platforms serve as a crucial channel for expanding its customer base and acquiring computing power, while the relevant procurement prices and cooperation fees directly impact the company's operating costs.

To improve the supply of computing resources and long-term costs, Anthropic is advancing the 'Anthropic Direct' infrastructure model. By directly leasing computing chips and deploying dedicated computing resources, it aims to enhance its ability to allocate computing power.

This model can reduce some reliance on traditional cloud service procurement methods, but it also means the company needs to bear equipment leasing and other infrastructure costs, and arrange corresponding funds in advance.

From expanding its corporate client base to achieving positive operating profits, Anthropic has demonstrated improved profitability brought about by scaled operations. However, faced with competitors such as OpenAI and Google continuously launching new models, adjusting prices, and the gradual execution of long-term computing power contracts, the company still needs to maintain sufficient customer demand and improve the efficiency of computing resource utilization.

For Anthropic, which is about to enter the capital market, revenue growth is already supported by actual performance. Whether this revenue can be translated into stable profits amid continuous model competition and computing power investments will be a more critical test in the next stage.