Anthropic publicly filed its IPO registration statement on October 26, 2026, and filed its first amendment on November 2 with financial statements through September 30. It plans to list on the Nasdaq Global Select Market under the ticker ANT, but the offering’s share count and price range remain undetermined; the amendment does not establish a completed public listing or a new valuation.

Its last disclosed primary financing was a $65B Series H announced on May 28, 2026, at a $965B post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. The company has raised approximately $125 billion in total funding to date, with Salesforce's stake valued at approximately $5 billion.

Anthropic closed a $30 billion Series G funding round on February 12, 2026, at a $380 billion post-money valuation. The round was led by GIC and Coatue and co-led by D. E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, and MGX, with participation from BlackRock-affiliated funds, Fidelity, General Catalyst, Sequoia, Temasek, and TPG.

Microsoft and Nvidia have also made strategic commitments to invest up to $5 billion and $10 billion, respectively, alongside Anthropic’s commitment to purchase Azure compute capacity.

Alongside the primary fundraising, Anthropic launched an employee tender offer in February 2026 sized at $5 billion to $6 billion, allowing current and former employees to sell shares at a valuation of approximately $350 billion.

In September 2025, Anthropic closed a $13 billion Series F funding round, valuing it at $183 billion, up from $61.5B in March 2025. The round was led by ICONIQ and co-led by Fidelity Management & Research and Lightspeed Venture Partners, with participation from Coatue, GIC, BlackRock- and Blackstone-affiliated funds, and the Qatar Investment Authority.

Anthropic was founded in 2021 by former OpenAI researchers led by Dario and Daniela Amodei. Its $124M Series A in May 2021 was led by Skype co-founder Jaan Tallinn, with participation from Dustin Moskovitz, Eric Schmidt, James McClave, and the Center for Emerging Risk Research, financing its initial research into reliable and steerable AI systems.

Anthropic develops Claude, a family of AI models available through its own applications and a developer platform. Users can ask Claude questions, give it documents to analyze, or delegate work that requires searching sources, editing files, running code, and interacting with other software.

The model lineup includes Opus 5.5 for complex coding and professional work and Sonnet 5.5 for faster, lower-cost tasks. Fable 5.1 provides another option for demanding, long-running projects. Mythos 5.1 uses the same underlying model as Fable 5.1 with different safeguards and restricted access through trusted programs for cybersecurity and life-sciences work.

Claude is available on the web, desktop, and mobile. A user can connect company files and applications, request an analysis, and turn the findings into a document, spreadsheet, presentation, or interactive artifact. Projects, memory, and reusable Skills supply context and instructions for recurring work, while connectors provide access to external information and actions.

Anthropic began merging Cowork and chat into one Claude experience in September 2026, initially rolling out to Pro and Max subscribers. Users can delegate longer tasks, check progress across devices, and schedule recurring work without choosing a separate agent interface.

Claude Docs and Slides add editable documents and presentations inside conversations, while Claude Design supports prototypes, visual layouts, and branded collateral. These capabilities let users develop the analysis and its presentation together, with shared context and outputs that can be edited, shared, or exported.

Integrations also bring Claude into existing work environments, including Microsoft 365, Google Workspace, browsers, and financial-data services. Teams can use these connections for tasks such as analyzing a workbook, researching a company, reviewing contracts, or preparing a customer briefing.

Developers give Claude Code a task, such as fixing a bug or implementing a feature, and the agent reads the repository, edits files, runs commands and tests, and produces changes for review. It operates through the terminal, editor integrations, and cloud sessions, supporting both interactive development and delegated work across larger codebases.

Skills package reusable instructions and scripts, while MCP connections expose external tools and data. These extend Claude Code into activities such as documentation, analysis, and file creation alongside software engineering.

Anthropic also offers Claude Security in public beta for Enterprise customers. It examines repositories for vulnerabilities and proposes fixes, extending the company’s coding capabilities into security review.

Developers access Claude through Anthropic’s API or cloud platforms including Amazon Bedrock, Google Cloud, and Microsoft Foundry. They can embed the models in their own products, define tools the models can call, and build agents around proprietary data and business systems.

Claude Managed Agents, available in beta, supplies the execution environment and agent loop for longer-running tasks. Developers configure the model, instructions, tools, and permissions, then launch sessions that can work with files, run code, browse the web, and maintain state.

Anthropic’s Model Context Protocol standardizes how applications expose tools and data to AI systems. MCP supplies integrations, while Skills supply reusable procedures and instructions for completing particular kinds of work.

Anthropic monetizes Claude through usage-based model access, individual and business subscriptions, and enterprise agreements. Claude Code and broader agent workflows connect these revenue streams, allowing a customer to begin with a subscription and increase spending as more work is delegated to the models. Consumption-based revenue contributed $26.3B, or 79% of revenue, in the first nine months of 2026, while subscriptions contributed $6.9B, or 21%.

Developers pay for input and output tokens, with rates varying by model. Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, compared with $4 and $20 for Opus 5.5 and $10 and $50 for Fable 5.1. Prompt caching reduces charges for reused inputs, while batch processing offers discounted rates for work that does not require an immediate response.

An agent can make many model calls while reading a repository, editing files, running tests, and checking its work. Spending can therefore grow with the number, duration, and complexity of delegated tasks, without requiring an equivalent increase in users. Lower token prices make more tasks economical, but revenue growth depends on usage expanding enough to offset those reductions. Anthropic estimated that price reductions lowered revenue by approximately $600M from Q2 to Q3 2026, illustrating the usage growth needed to offset cheaper model access.

Claude Managed Agents adds a charge of $0.08 per active session-hour on top of standard token fees. This lets Anthropic sell the infrastructure for executing agents alongside the underlying model calls.

Claude Pro costs $20 per month when billed monthly, while Max plans cost $100 or $200 per month for higher usage allowances. Team offers standard seats at $20 per user per month with annual billing or $25 monthly, and premium seats at $100 annually billed or $125 monthly.

Enterprise pricing combines seats with model usage. The self-service offering starts at $20 per seat per month with annual billing and a 20-seat minimum, with usage charged at API rates. Security controls, administration, and data-handling requirements support larger organizational deployments.

Included usage creates an entry point for adoption, while paid usage beyond plan limits and higher-capacity plans create expansion revenue. Claude Code and other bundled capabilities also give customers more ways to consume their allowances and reasons to retain the subscription.

Anthropic sells directly and through cloud platforms, software companies, and consulting firms. Cloud distribution lets enterprises purchase Claude through existing procurement relationships, while Claude and Claude Code give Anthropic direct relationships with end users.

The company’s multi-year $200M Snowflake partnership combines model availability with joint enterprise sales. Deloitte has made Claude available to more than 470,000 employees, and PwC is deploying Claude internally and packaging it into client engagements. These firms can turn model capabilities into industry-specific implementations for buyers that need integration and operational support.

Anthropic committed an initial $100M to its Claude Partner Network in 2026 for training, technical support, and joint market development. The distribution model broadens its reach, but cloud partners and application developers can also sell competing models or products.

Anthropic must fund both model development and the computing capacity used to serve customers. More efficient models, cached inputs, and better utilization can reduce costs, while longer agent tasks increase the computing consumed per user.

The November 2 IPO amendment disclosed approximately $895.4B in aggregate purchase commitments and lease agreements, including short-term commitments and leases not yet commenced, entered into as of and after September 30, 2026. Approximately $447.5B is expected to be incurred before 2031, with substantially all of the remainder between 2031 and 2045; $69.8B is due in the 12 months following September 30. These are future contractual spending commitments, rather than expenses already incurred.

As of September 30, Anthropic reported approximately $146.6B in liquidity, comprising $95.6B in cash and cash equivalents, $36.3B in short-term investments, and $14.8B of available credit facility capacity. Its remaining performance obligations were $13.8B, up from $10.8B at June 30; these represent contracted revenue yet to be recognized, rather than cash available to fund its infrastructure commitments.

These contracts secure capacity ahead of demand while creating a large fixed spending burden. Anthropic must grow paid usage, manage serving costs, and secure financing across the period between committing to infrastructure and generating revenue from it.

Anthropic competes most directly with OpenAI in foundation models, developer APIs, and coding agents, with Claude and Claude Code competing against GPT models and Codex. Google competes across models, cloud distribution, and productivity software, while Chinese open-weight labs give developers lower-cost alternatives they can customize and run through other providers.

OpenAI is the San Francisco-based AI lab that Dario and Daniela Amodei and other researchers left to found Anthropic in 2021. ChatGPT gives OpenAI the largest direct consumer distribution in AI, with more than 1 billion weekly active users as of August 2026, while Anthropic has focused on enterprise customers and model access through cloud providers.

Claude Code gave Anthropic a head start in coding agents when it launched in February 2025. OpenAI followed with Codex CLI in April and its cloud coding agent in May, and has increasingly emphasized the Codex desktop app while Claude Code built around the command line. Both companies are now bringing coding agents together with broader work products: Claude Code, Cowork, and Claude at Anthropic, and Codex, ChatGPT Work, and ChatGPT at OpenAI.

Companies often use both OpenAI and Anthropic, shifting workloads between them as new releases change which model performs best for a particular task. Leadership in coding or reasoning can draw usage toward one provider until the next release shifts it back, making model performance, pricing, and workflow integration ongoing competitive pressure points.

OpenAI is also competing for developers through aggressive API price cuts. GPT-5.6 Luna ranked No. 3 on OpenRouter by tokens processed for the week ending August 18, 2026, compared with Claude Opus 5 at No. 9, though OpenRouter cautions that token volume does not directly equate to requests, users, or spending. Anthropic has differentiated on trust: after OpenAI began testing ads in ChatGPT's free and Go tiers, Anthropic ran a Super Bowl LX campaign pledging that Claude will remain ad-free.

Google competes with Anthropic in frontier models through Gemini, developed by its combined DeepMind unit. The Gemini app surpassed 1 billion monthly active users in August 2026, and tools like Antigravity extend Google's competition into coding agents.

Google's main advantages are distribution and compute. It can build Gemini into Search, Workspace, Android, and Chrome, and it trains and serves models on its own TPUs and data centers, which can support cheaper inference. Claude is offered through Google Cloud, where it competes directly with Gemini on the same platform.

The two companies also partner. Google is an investor in Anthropic and one of its largest compute suppliers, with a hosting commitment of at least approximately $121.3B disclosed in the amended prospectus and a separate agreement to lease Google chips through Broadcom.

Moonshot AI and DeepSeek compete for developer workloads with open-weight models that customers can customize and run through independent inference providers. Moonshot's 2.8T-parameter Kimi K3, released in July 2026, shows that Chinese labs are competing on large-scale agentic and reasoning capabilities as well as smaller, cheaper models.

Open weights also let application companies build specialized models around their own workflows rather than relying entirely on Claude. Cursor's Composer 2.5 is built on Moonshot's Kimi K2.5, and US inference providers such as Fireworks offer Kimi K3 with US-only endpoints and zero data retention. These options give coding platforms more control over inference costs and model behavior while keeping the customer relationship inside their own product, increasing pressure on both Anthropic's API business and Claude Code.

Anthropic’s expansion opportunities lie in applying its coding and agent capabilities to broader professional work and scientific research. Both extend the company beyond software-development budgets, but require domain-specific integrations, reliable outputs, and workflows that customers can use repeatedly.

Claude’s ability to research, manipulate files, and work across connected applications gives Anthropic a path into finance, legal, marketing, healthcare administration, and operations. The opportunity is to handle a larger share of a completed assignment, from gathering evidence through producing a deliverable or updating a business system.

For a financial analyst, that could mean collecting company data, updating a model, and preparing an investment presentation. For a legal team, it could mean finding relevant agreements, checking them against an internal playbook, and preparing revisions. Healthcare workflows such as prior authorization and claims appeals similarly combine document review with work inside existing systems.

Anthropic can monetize this through additional enterprise seats and model consumption as customers delegate more work. Consulting partners provide a route into larger implementations, including legacy-system modernization and the redesign of finance, HR, and insurance processes. PwC’s reported reduction of an underwriting workflow from 10 weeks to 10 days illustrates the kind of operational spending these deployments can address.

Scientific research creates a distinct opportunity to sell Claude as a system for conducting computational work, generating hypotheses, and coordinating specialized tools. Pharmaceutical companies and research institutions could use these capabilities across biological discovery, experimental analysis, and drug-development workflows.

Anthropic’s beta science workbench connects models with scientific databases, analysis tools, and customers’ computing infrastructure. Its internal life-sciences laboratory provides a testing ground for the same approach. In September 2026, Anthropic reported that approximately 950 Claude agents consumed 210 million tokens over 21 hours while searching biological datasets, contributing to the identification of an enzyme system whose function remains under investigation.

The commercial opportunity includes subscriptions and model usage for research campaigns that can involve many parallel agents and substantial computation. Anthropic has not established a separately disclosed, commercially significant scientific-research revenue stream.

Dario Amodei’s essay Machines of Loving Grace describes a broader ambition for AI systems that perform the work of researchers, including designing experiments and interpreting results. That vision supports the direction of the opportunity, while physical experiments, data quality, and clinical validation continue to determine how quickly computational findings become useful discoveries.

Compute commitments and financing: Anthropic disclosed approximately $895.4B in aggregate purchase commitments and lease agreements, including $69.8B due in the 12 months following September 30, 2026, against reported liquidity of $146.6B that includes available credit capacity. Capacity delays, higher-cost short-term compute or weaker demand could leave substantial payments unsupported by customer usage and increase its need for financing.

Customer concentration and enforcement: The five largest customers generated approximately 25% of revenue in the first nine months of 2026, and major buyers can change models or reduce spending. Anthropic also estimated that customers violating its usage policy or commercial terms contributed 5–7% of revenue over that period, so enforcement can remove paid usage as well as protect the platform.

Model liability and restrictions on use: The IPO prospectus warns that models may act in unintended or misaligned ways, including resisting shutdown or concealing information, and that catastrophic liabilities could exceed insurance coverage. Harmful behavior can lead to regulatory restrictions and political backlash, while stronger safeguards can limit deployments or delay releases.