30-second version: Anthropic is the AI safety company that makes Claude. Founded in 2021 by former OpenAI researchers led by Dario and Daniela Amodei, headquartered in San Francisco. Roughly 1,500 employees as of 2026, $7.3B+ in cumulative funding from Google, Amazon, and others, and a most-recent valuation reported around $61.5B. Anthropic does not have publicly traded stock — the only way to invest is indirectly through Google or Amazon, both major backers. The flagship product is the Claude model family (Haiku, Sonnet, Opus, and a frontier preview tier called Mythos), used by hundreds of millions of people across consumer, enterprise, and developer surfaces.
Best for: Anyone trying to understand who builds Claude, why they built it, and where the company is going.
You’ll get: Founding story, leadership team, model lineup, business model, key milestones, current status, and how Anthropic compares to OpenAI and Google.
Skip if: You already work in AI strategy. Daily AI fundamentals in our free Beginners in AI newsletter.
Anthropic is the company that makes Claude. It is also a deliberately strange company by Silicon Valley standards — founded specifically around the idea that the AI industry needs more careful actors in it, structured as a public benefit corporation, and willing to publicly explain why it does things differently from its competitors.
Here’s what the company is, who runs it, what it makes, and what to know if you’re tracking the broader AI industry.
The founding story
Anthropic was founded in 2021 by a group of researchers and operators who left OpenAI. The core founders are siblings Dario Amodei (CEO) and Daniela Amodei (President), both of whom were senior at OpenAI before departing. Dario was VP of Research; Daniela was VP of Operations. Several other senior OpenAI researchers came with them.
The stated reason for the split was a difference in views about AI safety — specifically how cautious to be as model capabilities scaled. The Amodei team believed safety needed to be a first-class research priority, not a side project. They started Anthropic to do that work alongside frontier model development, structured as a public benefit corporation so the safety mission was legally protected.
The first major model release was Claude in March 2023. It was positioned from launch as the “safer alternative to ChatGPT” — same chat interface, different research priorities under the hood. Three years later, Claude has grown into a model family used across consumer, enterprise, and developer products at scale.
The leadership team
- Dario Amodei — CEO and co-founder. Physics PhD, former research director at OpenAI, the public face of Anthropic on technical and policy questions. Known for being a careful communicator about AI risks and a serious technical leader.
- Daniela Amodei — President and co-founder. Runs operations, policy, and business. Has been the architect of Anthropic’s organizational culture and its public benefit corporation structure.
- Jared Kaplan — Chief Scientist. Theoretical physicist who co-authored the foundational scaling-laws paper that explained how language models improve with more compute and data. Co-founder.
- Tom Brown — Co-founder, formerly the lead engineer on GPT-3 at OpenAI.
- Chris Olah — Co-founder, interpretability research lead. One of the most respected researchers on understanding what’s actually happening inside neural networks.
The founding team is notable for being researchers first, businesspeople second. Most of them have continued to publish technical papers as Anthropic has scaled to a multi-billion-dollar company.
The Claude model family
Claude is structured as a family of models at different price-performance tiers. As of mid-2026:
- Claude Haiku (current: 4.5). The smallest, fastest, cheapest tier. Designed for high-volume tasks where speed and cost matter more than peak capability. See Haiku guide.
- Claude Sonnet (current: 4.6). The balanced workhorse. Strong reasoning, broad use cases, reasonable cost. The default tier for most consumer use.
- Claude Opus (current: 4.7). The top consumer tier. Maximum capability for difficult work. Slower and more expensive per query.
- Claude Mythos (preview). The frontier-tier model. Currently in preview through Anthropic’s Project Glasswing partnership program with AWS, Apple, Google, Microsoft, and Nvidia. Notable for its role in the May 2026 M5 macOS security disclosure.
Claude is available through claude.ai (the consumer web app), the Claude desktop and mobile apps, the Anthropic API for developers, and increasingly through partner platforms like AWS Bedrock and Google Vertex AI. The agent-style developer product Claude Code and the document-heavy workspace product Claude Cowork sit on top of the same underlying models.
The business model
Anthropic makes money in three places.
Consumer subscriptions. Claude Pro and higher tiers, for individuals. Roughly $17–20/month at the entry level, with higher tiers for power users and teams.
API usage. Developers pay per token (input and output) for API access. The pricing varies across the model tiers, with Haiku cheapest and Opus most expensive. This is the largest revenue driver for the company at current scale.
Enterprise contracts. Companies like Apple, Microsoft, Salesforce, and others have major enterprise agreements that include API capacity, contractual data-handling guarantees, custom support, and deployment infrastructure. Microsoft notably canceled internal Claude Code licenses in May 2026 but maintains the broader enterprise relationship.
Anthropic has consistently described itself as a research-first organization that commercializes to fund the research. The revenue trajectory has tracked closely with each new Claude release, with major step-ups in 2024 (Claude 3 family) and 2025–2026 (Claude 4 family + Cowork + Code).
Funding and valuation
Anthropic has raised more than $7.3 billion in cumulative funding across multiple rounds. The two largest strategic investors are:
- Amazon. Multi-billion-dollar investment with AWS as the preferred cloud provider for Claude. AWS Bedrock hosts Claude as a primary offering.
- Google. Substantial investment alongside Google Cloud distribution partnerships.
Other investors include Spark Capital, Salesforce Ventures, Sound Ventures, and a long list of strategic and financial backers. The most recent reported valuation is around $61.5 billion as of late 2025 / early 2026, making Anthropic one of the most valuable private companies in tech.
There is no Anthropic stock. The company is private. Retail investors who want exposure to Anthropic typically do so indirectly through Google (Alphabet) or Amazon stock, both major backers. Anthropic has not publicly discussed IPO plans as of mid-2026.
Constitutional AI and safety research
One of Anthropic’s signature research contributions is Constitutional AI — a training technique where the model is taught to follow a written set of principles (a “constitution”) rather than being trained primarily on human-labeled examples of good and bad behavior. The technique reduces the labor cost of safety training and makes the model’s values more inspectable.
Anthropic has also been one of the most active publishers in interpretability research — the field that tries to understand what’s actually happening inside large language models. Chris Olah and the team have produced widely-cited papers on circuits, features, and the structure of neural network activations.
The May 2026 launch of Project Glasswing — giving major infrastructure companies (AWS, Apple, Google, Microsoft, Nvidia) early access to Claude Mythos for defensive security research — is the most visible public expression of the safety mission. The premise: frontier models can find vulnerabilities faster than ever, and defenders should have the model first.
Anthropic vs OpenAI vs Google DeepMind
The three frontier AI labs differ in real ways.
- OpenAI is the largest by user count and revenue. ChatGPT is the most-used consumer AI product. OpenAI is structured as a capped-profit company with Microsoft as the dominant partner. Product strategy is breadth-first — image, voice, search, agents, ads, finance all in one app.
- Google DeepMind is the research-rich incumbent. Gemini integrates deeply with Google Workspace and Search. Distribution advantage is unmatched. Has historically lagged on shipping speed but has been catching up.
- Anthropic is the smaller, more focused safety-first lab. Smaller user base than OpenAI but stronger position in enterprise (especially legal, financial services, and developer tools). Product strategy is depth-first — fewer features but each one carefully built. Constitutional AI and interpretability research is a public differentiator.
Honest read: Anthropic isn’t trying to be OpenAI. The deliberate slower-and-safer positioning is a strategic bet that the AI market will reward careful actors over time, especially in regulated industries.
Key milestones
- 2021: Anthropic founded by Amodei team and other OpenAI veterans.
- 2022: Initial research output, including Constitutional AI papers. Anthropic operates without a consumer product through 2022.
- March 2023: Claude launches publicly.
- 2024: Claude 3 model family launches (Haiku, Sonnet, Opus). Major scale-up in employee count and revenue.
- 2025: Claude 4 family. Major enterprise wins. Multi-billion-dollar revenue run rate.
- February 2026: Claude Cowork launches. Initial Claude for Legal preview.
- May 2026: Project Glasswing announced. Mythos preview model. Claude for Legal ships 12 open-source plugins. The Mythos/Apple M5 disclosure becomes the most-discussed AI security event of the year.
What to know if you’re tracking Anthropic in 2026
- The legal vertical is where Anthropic is winning hardest. Lawyers are reportedly the top profession on Cowork. The May 2026 Claude for Legal launch (Apache 2.0 plugins on GitHub) cements this.
- The plugin ecosystem is the strategic bet. Open-source vertical plugins for legal, financial services, healthcare, real estate are the obvious next-wave categories.
- Apple is the most consequential partner relationship. Project Glasswing membership, plus the rumored iOS 27 Extensions framework (June 8 WWDC) that would let Apple Intelligence route to Claude.
- Mythos GA is the next major model release. Currently preview-only. Timing for general availability has not been publicly announced.
- Microsoft is a complicated partner. The May 2026 internal Claude Code license cancellation was real, but the broader commercial relationship continues. Worth watching how this evolves.
FAQ
Who founded Anthropic?
Dario and Daniela Amodei (siblings, both former senior OpenAI executives) founded Anthropic in 2021 along with several other OpenAI veterans including Jared Kaplan, Tom Brown, and Chris Olah. Dario is CEO; Daniela is President.
Can I buy Anthropic stock?
No. Anthropic is privately held. There is no public stock. Retail exposure to Anthropic is typically through Google (Alphabet) or Amazon, both of which are major investors. Anthropic has not announced IPO plans as of mid-2026.
How is Anthropic different from OpenAI?
The founding team left OpenAI in 2021 over differences in safety philosophy. Anthropic positions itself as the safety-first frontier lab. In practice, Anthropic ships fewer features than OpenAI but with deeper polish. Claude is known for better writing quality and rule-following; ChatGPT is known for broader feature surface. See our full comparison.
How big is Anthropic in 2026?
Roughly 1,500 employees as of 2026. More than $7.3B in cumulative funding. Most-recent reported valuation around $61.5B. Multi-billion-dollar revenue run rate. Hundreds of millions of users across consumer, enterprise, and developer surfaces.
Is Anthropic profitable?
Profitability for frontier AI labs in 2026 depends heavily on how you account for training compute. Operational margins on inference (running the models for users) are positive. Training new frontier models is enormously expensive, and Anthropic continues to invest aggressively in research. The company has not publicly disclosed detailed financials.
What is Constitutional AI?
Anthropic’s signature safety-training technique. Instead of relying primarily on human-labeled examples of good and bad model behavior, Constitutional AI trains the model against a written set of principles. The model is then taught to critique and revise its own outputs according to the constitution. The technique reduces labor cost and makes the model’s values more inspectable.
The bottom line
Anthropic is the AI company most committed to research depth and safety positioning. It is also a successful commercial operation by any reasonable measure — multi-billion-dollar revenue, hundreds of millions of users, and the kind of enterprise partnerships (Apple, Microsoft, Amazon, Google) that suggest the depth-first strategy is working.
For most users in 2026, Anthropic is “the company that makes Claude.” That’s a fair short version. The longer version is that Anthropic is shaping how the AI industry talks about safety, plugin ecosystems, and enterprise deployment — and the next 12–24 months will determine whether the careful approach scales as well as the breadth-first approach has at OpenAI.
For daily reads on what Anthropic and its competitors are shipping, subscribe to the free Beginners in AI newsletter. For Anthropic-specific products: Claude basics, Claude Code, Claude Cowork, Claude for Legal, Claude Mythos.
Sources
- Anthropic, official company pages, research papers, and product blog.
- Anthropic press releases for funding rounds and enterprise partnerships (2022–2026).
- Reuters, Bloomberg, and Wall Street Journal coverage of Anthropic valuation, partnerships, and product releases.
- Dario Amodei public interviews and Senate testimony.
- Constitutional AI: A research paper published by Anthropic (2022) describing the technique.
Learn Our Proven AI Frameworks
Beginners in AI created 6 branded frameworks to help you master AI: STACK for prompting, BUILD for business, ADAPT for learning, THINK for decisions, CRAFT for content, and CRON for automation.
Get Smarter About AI Every Morning
Free daily newsletter — one story, one tool, one tip. Plain English, no jargon.
Free forever. Unsubscribe anytime.