Anthropic — The AI Safety Company Behind Claude

In 2021, Dario Amodei, his sister Daniela, and a group of their OpenAI colleagues decided to leave. The disagreement wasn’t about technical details but about a more fundamental question: What should the safest and most effective AI company look like? What they’ve built since then has become the world’s second‑most‑valuable AI company with a $380 billion valuation.

Founding Story

Anthropic was founded in May 2021 in San Francisco by Dario Amodei, formerly head of research at OpenAI, Daniela Amodei who serves as Chief Operating Officer, and a group of prominent researchers. The initial investment reached $580 million, including $500 million from FTX — which created later complexities following the platform’s collapse.

Anthropic’s foundational principle stems from what Dario Amodei calls “the manuscript of the mind” — the belief that current large language models represent a first step toward artificial general intelligence, and that work to make this intelligence safe shouldn’t wait until its arrival. Hence the focus on interpretability research and Constitutional AI technology for training models with consistent values.

Claude Models

Claude launched in 2023, and the family evolved with striking speed. Claude 3 Opus in early 2024 demonstrated capabilities surpassing GPT‑4 on most benchmarks. Claude 4 Opus and Sonnet in May 2025 added advanced agent capabilities. Claude Opus 4.6, released in February 2026, powers agents capable of handling entire categories of real‑world work, with a 1‑million‑token context window and an Agent Teams system for coordinating multiple Claude instances.

Claude remains distinct from competitors in specific points: its degree of adherence to instructions, its level of caution in handling sensitive content, and the quality of its long‑form text output. The model is available globally via AWS, Google Cloud, and Microsoft Azure — the only frontier AI model available simultaneously on all three major cloud platforms.

Claude Code and the Major Shift

Claude Code moved from a research preview to general availability in May 2025. What followed is difficult to classify except as a phenomenon. Within nine months, its annualized revenue reached $2.5 billion — from zero to that number in under a year. 4% of all public GitHub commits are now written with Claude Code assistance, with expectations to reach 20% by the end of 2026.

Enterprise subscriptions for Claude Code quadrupled since January 2026, with enterprise clients accounting for more than half of its revenue. These numbers triggered a broad sell‑off in commercial software company stocks, with the sector losing over $2 trillion in market value.

Unprecedented Financial Growth

From $1 billion in annual revenue in December 2024, to $4 billion in July 2025, to $9 billion by the end of 2025, to $14 billion in February 2026. There is no comparable growth rate at this scale in commercial software records.

The revenue structure differs from OpenAI: 70–75% of revenue comes from usage fees for developers and enterprises via a pay‑per‑token model, versus 10–15% from consumer subscriptions. Eight of the top ten companies in the Fortune 10 are now Claude clients. More than 500 customers spend over $1 million annually.

Funding and Investors

Series G round in February 2026 raised $30 billion led by GIC and Coatue, bringing the total raised by the company to approximately $64 billion since its founding. The valuation reached $380 billion. Amazon invested $8 billion as a strategic partner with AWS designated as a primary platform for model training.

The company is preparing for a potential initial public offering in October 2026 that could raise over $60 billion. Goldman Sachs, JPMorgan, and Morgan Stanley have entered early discussions to oversee the process.

Safety as a Competitive Strategy

What distinguishes Anthropic from its competitors is the principle that safety and performance are not in conflict. The Constitutional AI program embeds values directly into the training process rather than adding them as an external layer. Interpretability research seeks to understand what occurs inside a model when it accomplishes a task — an endeavor the lab sees as essential for controlling future models.

Government Claude (Claude Gov) is used in several U.S. security agencies according to Ars Technica in June 2025. Partnerships with Palantir and AWS made Claude the only model used in classified secret missions. In July 2025, Anthropic secured a $200 million defense contract.

Expansion and 2026 Plans

Investment in infrastructure is planned at $19 billion in 2026, distributed between $12 billion for training and $7 billion for inference. Estimates indicate this spending roughly equals current revenue, meaning the company is spending everything it earns and more.

Anthropic expanded into digital health with HIPAA‑compliant Claude healthcare applications and acquired Coefficient Bio to strengthen its presence in drug discovery. The Claude Cowork product for office task automation is redrawing the boundaries of what AI can replace humans in executing.

The gap between a startup and established institutional power hasn’t yet been fully crossed. The real question is not about the technical capabilities of the models, but about the nature of the corporate identity Anthropic will carry when it becomes a public company balancing shareholder demands with its research legacy in safety.

---

Explore More

Want to learn more about the latest models mentioned here? Visit our Top AI Models List for a comprehensive comparison, or browse the Latest AI Tools to boost your productivity.