Anthropic AI Timeline: Claude, MCP, and Cloud Funding

Founded by people who left OpenAI

Anthropic was founded in May 2021 by seven people who left OpenAI, including its former vice president of research, with $124 million and an explicit organizing idea: treat AI safety as a discipline to be researched, not a caveat to be added later. It incorporated as a public benefit corporation, and its own account of that structure says the company's purpose is the responsible development of advanced AI for the long-term benefit of humanity. Dario Amodei is chief executive and his sister Daniela Amodei is president; the board also includes Reed Hastings and Chris Liddell. [1][2]

The company sells access to its Claude model family, organized into tiers from Haiku through Opus, through a consumer chat product, an API, and agent tools including Claude Code, a command-line coding agent, and Cowork, a graphical interface aimed at non-engineers. Reporting through 2026 consistently frames Anthropic and OpenAI as the two AI companies large enough to be valued on their own, rather than as a feature of a larger company, which is the position the funding events on its timeline reflect. [3]

Why Anthropic mattered

Atlas interpretation: Anthropic's significance to AI history is less about a single technical breakthrough than about what its existence did to the field's structure. A group that had just built GPT-2 and GPT-3 at OpenAI left to build a competing frontier lab funded on a safety pitch, which meant the industry no longer had one company setting the pace and everyone else reacting. Claude launched the same afternoon as GPT-4, trained against a written constitution rather than human preference labels alone, and the two labs spent the next three years in a genuine two-horse race, which is most of what the model releases on the timeline document. [4]

Anthropic's other lasting technical contribution is the Model Context Protocol, released in November 2024: a common way for an AI application to reach external tools and data, with SDKs and example servers rather than a closed integration. Competing labs and tool vendors have since adopted it, which is the practical test of whether a protocol becomes an industry standard instead of a vendor feature. [5]

Funded by the cloud providers it competes with

Anthropic's growth has been funded largely by the cloud providers whose infrastructure it also depends on. Amazon committed an initial $1.25 billion in September 2023, with AWS as primary cloud provider and Trainium as a training chip, then added more through 2024 and 2026. Microsoft and Nvidia together put up to $15 billion into Anthropic in November 2025, with Anthropic committing $30 billion back to Azure in the same deal, making Claude the only frontier model sold on all three major clouds. By its May 2026 Series H, the largest private round on record, Anthropic was valued at $965 billion, eight weeks after OpenAI had set the previous record. [6][7]

Anthropic on the timeline

Three Anthropic events on the timeline are marked as turning points: the 2021 founding, the 2023 Claude launch, and Claude Code's February 2025 debut, shipped alongside Claude 3.7 Sonnet as a tool that could read a repository, run commands, and open pull requests on its own. That event marks coding agents leaving demos and entering ordinary workflows, and it is the product line Anthropic has invested in most visibly since, through Cowork and a series of managed-agent offerings. [8]

The fourth turning point, June 2026's export controls on frontier models, hit Anthropic directly: a Commerce Department directive suspended all foreign-national access to Claude Fable 5 and Claude Mythos 5, the first time access to a frontier model itself became a controlled export rather than the chips that trained it. Anthropic restored Fable 5 after Commerce cleared it on June 30. The same year also saw a federal supply-chain risk fight with the Pentagon, which a judge ruled unlawful in August 2026, and separate copyright suits from music publishers over training data, both of which the timeline marks as major events or smaller ones rather than turning points, because neither had, as of the date these sources were checked, reshaped the field the way the export-control decision did. [9][10]

Pricing is the other pattern. Claude Opus 4.5 arrived in November 2025 at a third of its predecessor's input token price while leading coding benchmarks, and Anthropic has kept shipping a new Opus point release roughly every two to three months since, a cadence that looks like a release schedule rather than a series of surprises once the smaller events on the timeline are lined up. [11]