How Did Claude's Revenue Beat Its Old Boss in Two Years? The Story of Anthropic and Dario Amodei

Video Information

Original link: https://youtu.be/wzxCvd3rWZk
Note: This article is based on a local transcript. Ads, subscription prompts and unrelated remarks have been filtered out. Only segmentation, headings and sentence cleanup were done. The original video's information and order are kept as much as possible, with no added explanatory paraphrasing.

Opening

OpenAI and Anthropic are the two most influential AI companies in the world.

The CEOs of these two companies were once put on stage together, ready for a handshake photo, and they absolutely refused to shake hands.

These two former colleagues who fell out are now outright rivals.

And on Dario's side, they even ran a brilliantly ironic ad during the Super Bowl.

That move was ruthless. You could say it really rattled OpenAI's Altman.

How did Anthropic manage to knock OpenAI flat in the enterprise AI market in just two years?

The Numbers

Let's look at the numbers first.

In 2023, the enterprise AI market was still OpenAI's turf, with half the market share. Anthropic had just 12% back then, and everyone assumed the outcome was already clear.

Then, by the end of 2025, the story had completely flipped.

Anthropic rose to around 40%, while OpenAI fell to just over 20%.

The revenue numbers are even more dramatic.

Anthropic had only about $100 million in 2023. It grew to $1 billion in 2024, and in 2025 it approached $10 billion. That's nearly a hundredfold increase in three years.

Epoch AI's latest analysis from this February said that at this pace, Anthropic's total revenue would overtake OpenAI's by the middle of this year, making it the most profitable AI company in the world.

How did this company overtake its rival so quickly in two years?

Dario Amodei's Background

We have to start with why Dario Amodei, their boss, left to found Anthropic.

Dario was born in San Francisco in 1983.

His family was nothing like the Silicon Valley elite. His father was an Italian-American leather craftsman, and his mother was a Jewish-American library project manager.

He grew up in the Mission District of San Francisco. Anyone who knows the city knows it was never a high-end neighborhood.

In high school, the dot-com bubble was in full swing. Everyone around him was thinking about how to write code and get rich, but he had no interest in that at all.

In 2000, 17-year-old Dario was selected for the US Physics Olympiad national team. He studied at Caltech first, then transferred to Stanford to study physics, and then went on to Princeton for graduate school, where he earned a doctorate in biophysics.

So he's a physicist, and he also holds a PhD in neuroscience. This double academic background gives him a way of looking at things that is completely different from the AI founders in Silicon Valley.

How His Father's Death Affected Him

In 2010, his father died of chronic myeloid leukemia.

But only a few years after his father passed, second-generation targeted drugs were developed. For that kind of cancer, survival rates jumped from 50% to 95%.

If scientific progress had been just a little faster, his father might still be alive today.

That completely changed Dario.

He began to think: if some technology could dramatically speed up scientific research, a tragedy like his father's would never have to happen again.

In the end, he saw that opportunity in AI.

From Baidu to OpenAI

In 2014, at 31, he joined Baidu's AI lab in Silicon Valley, and later went to Google Brain.

In the summer of 2015, Musk and Altman held their famous dinner at the Rosewood hotel and invited top AI researchers, including Dario and Ilya, to attend. That's how Dario became one of OpenAI's earliest core members.

During his years at OpenAI, he personally built GPT-2 and GPT-3. His team also developed RLHF, reinforcement learning from human feedback. This technique is what made AI obedient, easy to use, and safer.

Why He Left OpenAI

But during his four years at OpenAI, he and Altman had many disagreements about how AI should be developed.

It's like inventing a new drug: the animal tests look promising, but the human trials aren't finished yet.

Altman's approach was to launch first, watch the side effects as the product sells, and use scale to gather data.

Dario felt that you should complete the human clinical trials first, make sure it was one hundred percent safe, and only then release it, even if a rival grabbed the market first.

By the end of 2020, Dario couldn't stay any longer. He resigned together with his sister Daniela and several friends from the Google Brain days, seven people in all. It was the biggest exodus of talent from OpenAI's early days.

How Anthropic Was Founded

In 2021, these seven people founded Anthropic, splitting the equity almost evenly among themselves.

They also deliberately registered the company under a rather special corporate structure.

It means that if the CEO one day refuses to release a new model for safety reasons, shareholders cannot legally oust him.

They really did write "building safe AI" into the company's charter.

Holding Back Claude, Even Though It Could Have Launched Earlier

In fact, back in the summer of 2022, before ChatGPT came out, they had already trained the first version of Claude.

They could have launched first, but they chose to run more safety tests.

A few months later, OpenAI released ChatGPT the moment it was finished, and it exploded worldwide instantly.

Anthropic did miss the chance to be first, but it also proved that they really put safety and reliability first.

Why the Core Team Is So Stable

Now fast-forward to today.

Look at Musk's xAI: half of its twelve co-founders have left. It's not just the co-founders. Even its CFO stayed only a short time before walking out.

Turn back to OpenAI, where only three of its eleven founding members remain. Those who left include Musk, who fell out with Altman; people who departed after the failed boardroom coup; and textbook-level heavyweights like Andrej Karpathy and John Schulman. All of them are gone.

And now look at Anthropic.

Seven co-founders, working under the equal-split equity structure that startups fear most, five years in, and not one of them has left.

The reason it has stayed so steady is actually another key figure that most people have overlooked: Dario's sister, Daniela Amodei.

Daniela Amodei's Role

Most AI companies are run by technical people. The founders are brilliant scientists, but they don't know how to manage people, money, or an organization.

Anthropic's leadership is a very rare combination: siblings running the company together.

Daniela studied English literature in college. In 2013, she became Stripe's 45th employee.

During her five years at Stripe, she watched with her own eyes how a startup grew from dozens of people to thousands without falling into chaos.

She knows very well that when a talented engineer is hard to work with, it is devastating to team collaboration.

So she built a process specifically to screen out brilliant jerks.

That means at Anthropic, no matter how skilled you are, if your values don't match, you won't be hired.

They would rather go without top talent than let people with mismatched values in.

Why They Turned the Enterprise Market Around

Now that we understand Anthropic's stable core team, let's look at how they turned the enterprise market around.

First, from day one, Anthropic's strategy was set in stone: no mass market, no flashy products, only the needs of enterprise customers.

Many people think AI safety is just a hollow marketing slogan. But for enterprise customers, safety is a firewall, and a firewall is exactly what big companies care about most.

Think about it. Would a bank dare let an AI leak customer data? Would a drugmaker dare let an AI give wrong medical advice? Insurers require every AI interaction to be traceable and auditable, and every regulated industry needs its AI to comply with European and national regulations.

If you can't clear these basic hurdles, you don't even get a seat at the briefing.

Anthropic has built these requirements into its underlying architecture from the very first day.

The result is that transnational giants with extremely strict security requirements, such as IBM, Pfizer, Novo Nordisk, and Norway's sovereign wealth fund, have chosen Anthropic one after another.

Why Claude Code Was the Turning Point

If safety was the foundation that got enterprises to nod, then Claude Code, launched in May 2025, was the key that let Anthropic completely turn the tables.

Claude Code is no longer just a tool. It's a senior engineer that can help you finish a project directly.

First, it can close the loop on its own.

If you ask it to refactor an API, it will run the tests in the terminal by itself. When it sees errors, it fixes them, and it keeps going until every test turns green.

Second, it has a strong grasp of very large codebases.

Most AI handles a single file very well, but it gets lost in an enterprise project with tens of thousands of lines. Claude Code can move between different files and precisely track variable definitions and all sorts of dependencies.

Third, its hallucination rate is far lower than its rivals'.

When it's unsure how to use a certain library function, it checks the documentation itself instead of inventing a syntax that doesn't exist.

A principal engineer on the Google Gemini API team publicly said that Claude Code took only an hour to write a pricing design that Google's internal team had spent a whole year on.

Even Andrej Karpathy, a founding member of OpenAI, often says that when it comes to coding, Claude is definitely his first choice.

Even the developers of Claude Code admit that much of Claude's own code was written by Claude itself.

Why Enterprise Customers Are Willing to Pay

As a result, Anthropic's enterprise customers grew from fewer than a thousand to over 300,000 in two years.

Large customers spending more than $100,000 a year grew more than sevenfold within a single year.

Now let's look at the essence of the battle between Anthropic and OpenAI.

A large part of OpenAI's money comes from consumer subscriptions, where the average person pays maybe twenty-something dollars.

A large part of Anthropic's money comes from enterprise contracts and API calls. The amount each paying user contributes on average differs enormously.

People treat ChatGPT as a super search tool: looking up information, translating, asking everyday questions. In people's minds, it's priced at Google's level. Free is best, and paying $20 already feels expensive.

Anthropic is completely different in the enterprise. It solves real business problems.

Traditional AI at best gives you suggestions, but Claude Code takes over the entire development environment, runs the tests, and catches bugs on its own. When enterprises do this math, the benchmark is a senior engineer on a seven-figure salary.

Compared with that, an API bill of a few hundred dollars is ridiculously cheap.

OpenAI and Anthropic Are Fighting Different Battles

As a result, Anthropic captured the most valuable part of enterprise AI spending with only a fraction of ChatGPT's user base.

And the enterprise market has a trait that the consumer market doesn't.

Today you use ChatGPT, tomorrow you try Gemini, the day after you switch to Claude. Switching costs are almost zero.

But if a bank has spent six months integrating Claude into its lending, review, and internal systems, telling it to switch to GPT tomorrow is simply impossible.

That's why Anthropic's revenue structure is so healthy.

Enterprise contracts often run for several years, so the numbers are predictable and gross margins are high.

While OpenAI was charging into the consumer market, trying to win over more free users, Anthropic was quietly signing up big customers and focusing all its energy on the enterprise market.

On the surface, the two companies run on the same track, but they are fighting entirely different battles.

By the time OpenAI turned around and wanted to seriously compete for the enterprise market, Anthropic had already locked in the most valuable of those customers.

Ending

So now you know: Anthropic's comeback was practically a foregone conclusion.

Compared with Altman, Dario has always been very pragmatic. He puts everything into technology and R&D, and puts building relationships and hype far behind.

As Eric Schmidt, former CEO of Google, said when he invested in Anthropic: at the very early stage, you have no data, no revenue, no market, and no product. What you're betting on is Dario, the person.

In this most important AI showdown in human history, the market bet on speed, the investors bet on the person, and Dario bet on a safe, responsible AI that, had it arrived a little sooner, might have saved his father.