Big Brand History
Case file 04

The History of Anthropic: Why a Group of OpenAI Staff Left to Build a Rival

Published · 7:39 video · about 4 min read

Chapters (11)
  1. Cold open
  2. Part 1: The believers
  3. Part 2: The walkout
  4. Part 3: A different kind of company
  5. Part 4: The chatbot they didn't release
  6. Part 5: Big Tech money
  7. Part 6: Claude goes to work
  8. Part 7: Saying no to the Pentagon
  9. Part 8: Overtaking OpenAI
  10. Part 9: Three lessons
  11. Close
On your phone? Portrait chapter clips from this case file are coming soon to TikTok, X and Facebook. Follow us there so you don't miss them.

The story in full

At the end of 2020, a group of senior researchers walked out of OpenAI, the lab that would go on to build ChatGPT, and started a rival from scratch. Six years on, that rival was preparing to sell its shares on the stock market. On 10 November 2026 it set a price range that could value it at more than a trillion dollars, and the Financial Times reported that its bankers were preparing to value it at $1.5 trillion or more. Its name is Anthropic. Its chatbot is Claude.

The believers

Dario Amodei joined OpenAI in 2016 and rose to vice president of research; his sister Daniela was vice president of safety and policy. In January 2020 a team led by Jared Kaplan and Sam McCandlish, with Amodei advising, published a paper on scaling laws, showing that language models got better in a smooth, predictable way as they grew and were given more data and computing power. That May came GPT-3. Amodei later said a group inside OpenAI believed two things more strongly than most: that more computing power keeps making these models better, and that scale alone was not enough without safety.

The walkout

In late 2020, Amodei and six senior colleagues resigned, and seven more OpenAI staff soon followed. "We left OpenAI because of concerns around the direction," Daniela Amodei later said. Dario was more diplomatic: "It all comes down to trust," he told TIME. The seven co-founders were Dario and Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Chris Olah.

A different kind of company

Anthropic was founded in early 2021 as a public benefit corporation, with a stated purpose of the responsible development of advanced AI for the long-term benefit of humanity. That May it announced $124 million in funding led by the Skype co-founder Jaan Tallinn. In April 2022 came a $580 million round led by Sam Bankman-Fried of the crypto exchange FTX. By the end of that year FTX had collapsed, and in 2024 the FTX estate sold most of its Anthropic stake for $884 million to help repay customers.

The chatbot they didn't release

By the summer of 2022, Anthropic had trained a chatbot called Claude. According to TIME, Amodei decided not to release it, worried about starting a race to build bigger and more dangerous AI. About three months later, OpenAI launched ChatGPT. "I suspect it was the right thing to do," Amodei said, "but it's not totally clear-cut." In December 2022 Anthropic published Constitutional AI, a training method in which the model critiques and revises its own answers against a written list of principles. Claude launched to the public on 14 March 2023.

Big Tech money

Anthropic needed computing power, and lots of it. In 2023 Google invested $300 million, then agreed to invest up to $2 billion more. Amazon invested $4 billion between September 2023 and March 2024 and another $4 billion in November 2024, becoming Anthropic's main cloud and training partner. In September 2023 Anthropic published its Responsible Scaling Policy and revealed a Long-Term Benefit Trust with the power to pick a growing share of its board.

Claude goes to work

In March 2024 came Claude 3. But the real breakthrough was coding: in February 2025 Anthropic released Claude Code, which lets developers hand Claude whole programming jobs from the command line. At the start of 2025 Anthropic's run-rate revenue was about $1 billion a year; by August it was more than $5 billion. Growth brought lawsuits. Authors claimed Anthropic had downloaded their books from pirate websites, and Anthropic agreed to a $1.5 billion settlement, given final approval in July 2026. Reddit also sued, claiming its site was scraped without permission; Anthropic says it disagrees.

Saying no to the Pentagon

The Pentagon wanted to use Claude for any lawful purpose. Anthropic refused to drop two limits: no mass surveillance of Americans and no fully autonomous weapons. On 27 February 2026 President Trump announced a ban on Anthropic's products in the federal government, and the Defence Secretary labelled the company a supply-chain risk. Anthropic sued. In August a federal judge in San Francisco ruled for Anthropic; in September an appeals court in Washington ruled two to one for the Pentagon under a different law. Anthropic has asked the full court to rehear the case.

Overtaking OpenAI

In February 2026 Anthropic raised $30 billion at a $380 billion valuation, and in May $65 billion more at $965 billion, ahead of OpenAI's $852 billion at the time. On 26 October 2026 it filed to list on the Nasdaq under the ticker ANT. Its run-rate revenue reached about $81 billion by the end of October, though it also reported a nine-month net loss of $104 billion, much of it from accounting charges. Figures are accurate as of 11 November 2026, and the IPO price and valuation may change.

Key dates

  • January 2020: the scaling-laws paper is published.
  • Late 2020: the Amodeis and colleagues leave OpenAI.
  • Early 2021: Anthropic is founded as a public benefit corporation.
  • 14 March 2023: Claude launches to the public.
  • February 2025: Claude Code is released.
  • 26 October 2026: Anthropic files to list on the Nasdaq.

Three lessons

  1. Your biggest rival might be sitting in your own team. Anthropic was built by people who helped make GPT-3.
  2. Saying no has a price. Holding back Claude and refusing the Pentagon both cost Anthropic, at least in the short run.
  3. Follow the paying customer. Anthropic sold mainly to businesses and developers, and coding became its rocket fuel.

Accurate as of 11 November 2026. Figures such as valuations and revenue move over time. This is an independent editorial history, not affiliated with or endorsed by any company mentioned. The full source list is in the YouTube description and pinned comment.

More case files

The History of Twitter: How twttr Became X thumbnail
Case file 01

The History of Twitter: How twttr Became X

· 9:59

How a side project at a failing podcast company became Twitter, then X: from "just setting up my twttr" in 2006 to Elon Musk's $44 billion deal and the rebrand.