ANTHROPIC

Dario Amodei was VP of Research at OpenAI. His sister Daniela Amodei was VP of Operations.

They were two of the most senior people at the company. In 2020-2021, they grew increasingly concerned that OpenAI was prioritizing commercialization over safety research.

The board crisis that would eventually happen in 2023 was already brewing beneath the surface — the tension between "move fast and ship products" and "slow down and do the safety work" was real.

In early 2021, Dario and Daniela left OpenAI and took a group of key researchers with them. They founded Anthropic as a public benefit corporation — a structure that legally requires the company to consider its impact on society, not just shareholder returns.

The name comes from "anthropic principle" in physics — the idea that the universe's fundamental parameters seem fine-tuned for human existence.

The founding thesis was simple: AI was going to become incredibly powerful whether anyone wanted it to or not. The safest path was to have a safety-focused lab at the frontier of capabilities, not watching from the sidelines.

Build the most powerful AI you can, but build it with safety baked into every layer.

Anthropic makes money through API access and subscriptions, similar to OpenAI. The Claude API charges developers per token for input and output.

Claude Pro costs $20/month for individuals with priority access and higher usage limits. Claude Team is $25-30/user/month for businesses.

Claude Enterprise offers custom pricing with enhanced security, admin controls, and longer context windows. Amazon Web Services resells Claude through Amazon Bedrock and Google Cloud offers it through Vertex AI, both generating revenue-sharing income for Anthropic.

Claude is the flagship AI assistant — available via web app, mobile app, and API. Claude excels at long-document analysis, coding, writing, and reasoning.

Claude's context window handles up to 200,000 tokens (roughly 500 pages) — dramatically more than most competitors. The Claude API lets developers build applications powered by Claude.

Claude for Enterprise provides businesses with a private, secure deployment. Constitutional AI is Anthropic's research framework for training AI systems to be helpful, harmless, and honest — the safety methodology that differentiates Claude from competitors.

Anthropic grew through a deliberate "safety as a brand" strategy. While OpenAI chased consumer virality with ChatGPT, Anthropic positioned Claude as the thoughtful, reliable, safety-conscious alternative.

Developers who found ChatGPT inconsistent or who worried about data privacy gravitated to Claude.

The enterprise partnerships were the real growth engine. Amazon invested $4 billion and made Claude the featured AI on Amazon Bedrock.

Google invested $2 billion and integrated Claude into Google Cloud. These partnerships gave Anthropic instant distribution to millions of enterprise developers without building a sales team.

Claude's strength in specific use cases drove adoption. Claude became known as the best AI for long-document analysis, nuanced writing, and careful reasoning.

Law firms, financial analysts, researchers, and enterprise customers who needed accuracy over speed chose Claude. The reputation for quality over flash built a loyal and growing user base.

The funding arms race is existential. Training frontier AI models costs billions.

Anthropic has raised $13.7 billion and needs to keep raising because each generation of Claude costs more to train. If a funding round fails or investors lose confidence, Anthropic can't compete at the frontier.

The company is in a spending war with OpenAI (backed by Microsoft) and Google (with DeepMind) — two of the richest companies in history.

Being second in consumer awareness hurts. ChatGPT is a household name.

Claude is not. Most non-technical people have never heard of Anthropic.

This matters because consumer brand recognition drives enterprise adoption — CIOs buy what they've heard of. Anthropic has to fight for mindshare against a competitor with a massive head start in public awareness.

The safety-capabilities tension is real. Anthropic's entire brand is built on being the "safe" AI company.

But to stay competitive, they must build increasingly powerful models. Every capability improvement creates new risks.

If Anthropic ships something that causes harm, the reputational damage is catastrophic because safety is their core promise. If they move too slowly, they become irrelevant.

Google ($2B+), Amazon ($4B), Spark Capital, Salesforce, Menlo Ventures, SK Telecom, Lightspeed