Anthropic

Anthropic

Anthropic is an artificial intelligence research and product company that develops the Claude model and application family. It is organized as a public benefit corporation and describes its purpose as building reliable, interpretable, and steerable AI systems.

The company competes in consumer applications, developer APIs, coding, enterprise work, research, and model access through major cloud platforms. It should not be reduced to the "enterprise model" label used in Venture Step E092.

Company and governance

Anthropic's company page describes the organization as an AI safety and research company. It identifies Dario Amodei, Daniela Amodei, Yasmin Razavi, Jay Kreps, Reed Hastings, and Chris Liddell as board members at the time of review.

The page also describes a Long-Term Benefit Trust that participates in governance. The structure is relevant to company identity. It does not independently establish that every product decision achieves the stated safety mission.

flowchart TD
    A["Anthropic PBC"] --> B["Frontier research"]
    A --> C["Claude products and API"]
    A --> D["Cloud and partner distribution"]
    A --> E["Safety, policy, and governance"]
    D --> F["AWS Bedrock"]
    D --> G["Google Cloud"]
    D --> H["Microsoft Azure"]

Products and distribution

Anthropic translates research into Claude consumer plans, team and enterprise products, coding tools, and an API. Exact models, prices, regions, context windows, and product availability change quickly and should be checked in current documentation.

The company's 2026 financing announcement says Claude is available through AWS, Google Cloud, and Microsoft Azure. Its partner-network announcement describes an enterprise strategy involving consulting, implementation, and services partners.

Multi-cloud availability can reduce dependence on one procurement channel. It does not mean the products, model versions, terms, or features are identical across clouds.

Funding and reported scale

Anthropic's Series G announcement reports $30 billion of funding at a $380 billion post-money valuation. It also reports growth in customers with more than $100,000 of annualized revenue.

These are company disclosures. Post-money valuation is an implied transaction value, not cash, revenue, or profit. Annualized customer revenue is not the same as revenue recognized during a completed year.

In April 2026, Anthropic and Amazon announced an expanded compute collaboration. Anthropic said run-rate revenue had exceeded $30 billion, more than 100,000 customers used Claude through Amazon Bedrock, and the company would commit more than $100 billion over ten years to AWS technologies for up to five gigawatts of capacity.

Run-rate revenue, customer count, spending commitment, capacity ceiling, and deployment schedule are separate metrics. [[How to Analyze AI Lab Economics]] shows the correct ledger.

Strategic position

Anthropic's current position combines its own Claude product surface with distribution through clouds and partners. This can create access to organizations that already use AWS, Google Cloud, or Azure, while preserving a direct developer and end-user relationship.

The company still depends on external infrastructure, capital, and partner channels. The size of a compute agreement does not establish utilization, margin, or return. The number of cloud customers does not establish equal production depth or retention.

Evidence boundary

Public sources establish Anthropic's stated mission, governance, financing, distribution, partnerships, reported demand, and infrastructure plans. They do not provide a full audited financial record, independent market share, comparative model superiority, customer retention, or unit economics.

E092 preserves Dalton Anderson's late-2025 view that Anthropic had a clearer enterprise position than some rivals. This profile replaces that broad characterization with a current, source-bounded company record.

This profile reflects official Anthropic sources reviewed on July 27, 2026. AI assistance was used for research organization, drafting, and validation. Publication remains unauthorized.