New: our quarterly report on the S&P 500, read from 500 companies’ own filings and calls. Read the Q3 2026 report
Saturday, Nov 21, 2026

Anthropic, PBC

ANT

Technology Software Infrastructure

Anthropic, PBC’s revenue for the nine months to September 2026 was $33.2 billion, against $2.5 billion a year earlier, from its prospectus.

88.43 8.74 −8.99%

After hours trading: 87.80, down 0.71%, last trade Fri 8:00pm Eastern.

Anthropic, PBC (ANT) Business Profile

Updated · Covers results through prospectus 0001628280-26-074338 · How this is made

What it does

Anthropic is an AI research lab that develops Claude, a family of AI models and services for businesses, developers, institutions and individual users. Founded in 2021, it focuses on developing advanced AI alongside safety research.

Claude is available through three layers: the underlying models, a platform for developers, and applications built by Anthropic. An AI model processes information and produces outputs such as text, analysis or code. Anthropic’s platform lets customers use those models in their own software and workflows, while its applications provide tools that users can work with directly.

Claude Code assists software developers with tasks including writing, editing, testing and debugging code. Claude Cowork helps professionals research, draft, analyze and carry out other knowledge work. Claude Security can scan codebases, assess vulnerabilities and write fixes. Customers also use Claude to automate processes and add AI capabilities to products they sell.

Anthropic organizes its research, product development and sales efforts around enterprise use cases. Its customers range from large companies to AI-native businesses, government and research institutions, and prosumers. The company also uses Claude internally in research, engineering and operations, including to help develop subsequent versions of its models.

How it makes money

Anthropic earns revenue from customers using Claude models through its platform and applications. Its revenue includes both consumption-based and subscription-based sales. Customers can increase usage, move to higher subscription tiers and bring additional workflows onto the platform.

Claude Platform gives developers access to models and tools for building and scaling their own applications. Anthropic also sells access to applications it builds on the platform, including Claude Code for software development and Claude Cowork for knowledge work. Its Claude Enterprise plan provides governance, security and administration for organizational use.

The company sells directly and makes Claude available through AWS, GCP and Azure. Under its joint approach with these cloud providers, Anthropic establishes direct contractual relationships with customers and pays the cloud providers a fee. Systems integrators, consultants and data platforms also help customers adopt Claude.

Anthropic can change model prices as it releases new versions and offers contractual discounts. For the three months ended June 30, 2026 to the three months ended September 30, 2026, higher usage was partially offset by a decrease of approximately $600 million from price reductions introduced as part of new model versions and contractual discounts.

Revenue was $33.2 billion for the nine months ended September 30, 2026 and $17.0 billion for the three months ended September 30, 2026. Operating loss was $1.3 billion for the nine months ended September 30, 2026 and $567 million for the three months ended September 30, 2026.

Customers and how it sells

Anthropic serves organizations around the world, including enterprises, AI-native companies, government and research institutions, and individual prosumers. Its enterprise customers use Claude for tasks such as insurance underwriting, clinical reports, cybersecurity and investment portfolio analysis. Developers also build customer-facing products and internal tools on Claude.

Named enterprise customers include American International Group and Bristol Myers Squibb. Research institutions using Claude include the Garvan Institute of Medical Research, Gladstone Institutes and the University of California, San Francisco. Legora and Salesforce have built products on Claude.

As of September 30, 2026, Anthropic had more than 11,900 customers with $100,000 or more of revenue on a TTM basis. That group included every company in the Fortune 10, over 50% of the Fortune 500, and over 30% of the Global 2000. The company also had more than 1,900 customers with $1 million or more of revenue on a TTM basis and approximately 250 customers with $10 million or more of revenue on a TTM basis.

Customers can access Claude through Anthropic and through cloud distribution partnerships. Anthropic also works with systems integrators, including Accenture, Deloitte, KPMG and PwC, and with data platforms including Salesforce and Snowflake.

Suppliers and costs

Compute—the processing capacity used to train models and serve customers—is Anthropic’s most significant operating cost. The company trains and runs Claude across AWS Trainium, Google TPUs, NVIDIA GPUs and AMD GPUs. It manages capacity from multiple partners as a shared pool, directing it among research, model development, customer service and internal use.

Anthropic typically procures most of its compute under multi-year commitments 12 to 18 months before it becomes available. It also buys short-term compute when customer demand or research needs exceed long-term capacity; that capacity currently costs significantly more than comparable compute under multi-year commitments.

Compute and infrastructure expenses were $9.9 billion for the three months ended September 30, 2026. Anthropic has also entered into purchase commitments and lease agreements with aggregate committed spend of approximately $900.9 billion as of and subsequent to September 30, 2026, substantially all related to compute and infrastructure. Approximately $70.7 billion of these commitments are due in the 12 months following September 30, 2026.

Other expenses include personnel, sales and marketing, channel-partner fees, data services and professional services. Anthropic expects increasing investment in owned or leased compute and infrastructure, including commitments requiring payments before capacity becomes operational.

Competitive position

Anthropic competes with OpenAI and Google in developing frontier AI models for software developers and enterprises. These rivals have financial resources and existing customer bases, and can integrate their models into applications sold to businesses. Anthropic also competes with other model providers, including specialized AI labs and organizations releasing open-weight or open-source models.

Anthropic emphasizes model capability, safety and reliability for enterprise customers. It integrates frontier research with safety work and uses Claude to assist in research, training and testing for subsequent model generations. It also uses a multi-chip architecture to direct workloads across different hardware.

Customers can build Claude into their applications and encode workflows and institutional knowledge into Claude-powered agents. Anthropic uses customer feedback from production deployments to improve its models, platform and applications. Distribution partnerships place Claude inside customers’ existing technology stacks.

Our reading: Workflow integration and products built on Claude can create a meaningful competitive advantage because changing providers may require customers to revisit embedded processes and applications. Our reading: Research and compute flexibility can support that advantage, but their durability is less certain because competitors also have substantial resources and alternative models are available.

Competitor Where they compete Threat (our reading)
OpenAI Frontier models for developers and enterprises High
Google Frontier models for developers and enterprises High

Growth opportunities

Anthropic’s growth strategy combines attracting new customers with expanding use among existing ones. Customers often begin with one use case and then deploy Claude across more departments, applications and workflows. The company reported Net Revenue Retention of approximately 900% as of September 30, 2026.

Its revenue was $17.0 billion for the three months ended September 30, 2026, up from $1.4 billion for the three months ended September 30, 2025. From September 30, 2025 to September 30, 2026, customers with $100,000 or more of revenue on a TTM basis grew from more than 950 to more than 11,900.

Anthropic expects advancing model capabilities to open additional uses across coding, knowledge work and other tasks. It is pursuing cybersecurity through Claude Security and its OSS Scanner, and is extending Claude into robotics. It also launched Ode with Anthropic with private equity partners to bring Claude-based services to mid-sized companies.

The company believes knowledge workers’ work represented more than $40 trillion in economic value globally in 2025. It sees an opportunity both in broader use of current models and in new applications made possible by future models. Anthropic expects continued substantial compute investment to support research, development and customer demand.

Structural risks

  • Compute commitments and funding: Anthropic must predict capacity needs well in advance. If requirements differ materially from projections, it could face insufficient capacity or be committed to excess compute that remains underused. It may need additional capital to meet obligations and support growth.

  • Infrastructure supply: Specialized AI chips and memory are produced by a limited number of manufacturers. Data centers also require substantial power, time, capital and expertise. Some infrastructure partners also offer competing AI services, creating potential conflicts over access, cost or workload priority.

  • Model safety and performance: Anthropic’s models have behaved and may behave in unintended, misaligned or autonomous ways, including taking harmful actions. Dissatisfaction with performance, reliability, safety or accuracy can affect demand.

  • Competition and customer retention: Customers may develop their own AI capabilities, reduce their use of Anthropic’s services or move to alternative providers. Anthropic derives substantially all revenue from a limited number of products and applications.

  • Regulation and trade controls: Evolving AI, privacy and trade rules can affect model development and deployment. Export restrictions have previously affected Anthropic’s models and could also restrict access to compute or international markets.

Bottom line

Our reading

Anthropic combines an AI research lab with a platform and applications business, with revenue tied to subscriptions and customer consumption. Its opportunity rests on wider enterprise adoption and deeper integration into customers’ work. (item1; item7)

Our reading: The business’s rapid expansion must be weighed against its operating losses, large infrastructure commitments and the need to secure compute ahead of demand. Embedded workflows may help retain customers, but competitors and changing model economics remain important challenges. (item1; item1a; item7)