The Victory of Values: How Can Anthropic Catch Up to and Surpass OpenAI?

Writing by: Xiao Bing, Deep Tide TechFlow

This might be the most passionate AI revenge drama of the year.

The once-dominant giant model leader OpenAI is no longer in the spotlight. Its former employees, who left with six others to establish Anthropic, are gradually eroding OpenAI’s lead in revenue, valuation, and enterprise market share.

The temperature difference in the secondary market is the most intuitive. Ken Smythe, founder of Next Round Capital, is faced with a transfer request for $600 million of old OpenAI shares, with six hedge funds and venture capital firms lining up to sell. At this time last year, these stocks would be snapped up within days. Now? He’s scoured hundreds of institutional investors’ pools and can’t find a single taker.

Meanwhile, $2 billion in cash is queued up to buy Anthropic’s shares.

On the on-chain derivatives platform Ventuals, Anthropic’s implied valuation briefly surpassed OpenAI’s, at $863.6 billion versus $846.1 billion.

What’s more telling is Goldman Sachs’ attitude. Selling old OpenAI shares to high-net-worth clients no longer yields profit sharing, effectively a discounted fire sale. Selling Anthropic’s shares? Still charging a 15% to 20% carry, take it or leave it.

Founded only five years ago, how has Anthropic gradually surpassed its former parent OpenAI step by step?

Leaving

The story begins in 2020.

That year, Dario Amodei was still OpenAI’s Vice President of Research, involved in building GPT-2 and GPT-3. There are many versions circulating in Silicon Valley about why he left—some say it was because Microsoft’s investment changed OpenAI’s nature; others say it was fundamental disagreements over safety philosophy.

Dario himself discussed this on Lex Fridman’s podcast, roughly saying: Arguing over others’ visions is extremely inefficient. Instead of trying to change others there, it’s better to bring trusted people and do what you want to do.

In 2021, Dario left with his sister Daniela and five other core OpenAI researchers to establish Anthropic.

Sam Altman probably didn’t pay much attention at the time. OpenAI was riding high, losing a few researchers was no big deal.

But during the “board coup” in November 2023, which was the most intense, OpenAI’s board even approached Dario, asking if he would replace Altman as CEO and merge the two companies.

He refused. What he wanted wasn’t the CEO position at OpenAI, but to build a new system from scratch according to his own logic.

From 2021 to 2024, Anthropic was almost invisible to the outside world.

When ChatGPT exploded globally at the end of 2022, Claude was still in internal testing. The Anthropic team felt it hadn’t met safety standards and wasn’t in a rush to launch. Meanwhile, competitors were grabbing users and headlines, while Dario was still obsessing over a training method called “Constitutional AI,” which constrained the model according to a set of written “constitutional” principles.

At that time, many thought Anthropic was a bit stubborn—market window was so small, if you didn’t seize it, others would.

But looking back now, Anthropic made a crucial choice during this “invisible period”: from day one, it focused on API and enterprise clients, almost neglecting consumer product promotion.

When Claude first launched in 2023, its awareness among consumers was worlds apart from ChatGPT; most users didn’t even know it existed.

Dario’s logic was probably this: consumer attention comes and goes quickly; signing enterprise contracts is where real revenue lies.

This judgment seemed conservative at the time, but proved correct by 2026. Of course, whether Anthropic’s “long-term enterprise focus” was a strategic choice or a forced shift after failing to compete in the consumer market with ChatGPT, both narratives have some truth.

By early 2025, Anthropic’s annualized revenue quietly reached $1 billion. At that time, this number didn’t attract much attention—after all, OpenAI was already in the hundreds of billions, and no one expected what would happen next.

Counterattack

Numbers tell everything.

Anthropic’s annualized revenue (ARR): January 2025 at $1 billion, by the end of the year $9 billion, February 2026 $14 billion, March $19 billion, early April surpassing $30 billion.

OpenAI during the same period: about $13 billion in 2025, around $25 billion by April 2026.

In 15 months, Anthropic grew 30-fold, from trailing OpenAI by a factor of ten to surpassing it by 20%. OpenAI’s own growth was not slow, but compared to Anthropic’s explosive rise, it looked like “steady growth vs exponential explosion.”

The biggest structural difference lies in revenue sources: over 80% of OpenAI’s income comes from ChatGPT consumer subscriptions. With 900 million weekly active users, that’s impressive, but only about 5% pay, the remaining 95% free-ride on compute.

Anthropic, on the other hand, is the opposite: 80% of its revenue comes from enterprise clients and API calls.

Enterprise revenue and consumer revenue are entirely different species.

Enterprise contracts, once signed, are hard to switch; deep integration creates switching costs, renewal rates are high, and revenue grows year by year.

Consumer subscriptions can be canceled at any time; a new product launch can cause churn.

In trading terms, one is a long-duration asset, the other a short-duration asset.

Looking at specific data: by April 2026, Anthropic had over 1,000 enterprise clients paying more than $1 million annually, doubling in two months. Eight of the top 10 Fortune companies use Claude. In the core code generation track, Claude captured 42% to 54% of the global market share, while OpenAI only held 21%. Ramp’s corporate expenditure data shows Anthropic’s share of enterprise AI spending soared from 10% at the start of 2025 to over 65% by February 2026.

Do these numbers mean OpenAI is “done”? Not necessarily. But they do show one thing: the so-called unbeatable first-mover advantages—brand, user base, ecosystem—have almost no effect in the enterprise market, where procurement decisions follow a different logic.

Claude Code

The trigger for Anthropic’s explosive revenue was a product called Claude Code.

Launched in May 2025, by November ARR exceeded $1B, and by February 2026 over $2.5B. From zero to $2.5B in nine months.

Looking at SaaS industry records, there’s no faster case. Cursor took over a year to reach $500 million, GitHub Copilot took longer.

What exactly makes Claude Code different from previous AI coding tools?

Simply put, GitHub Copilot helps you autocomplete the next line while you code—you’re still the worker. Claude Code, on the other hand, is told “I want a user login module,” and it writes code, creates files, runs tests, commits changes—all automatically, while you watch.

This difference sounds like a matter of degree, but it’s actually a paradigm shift: one is a “better tool,” the other a “colleague that replaces your work.”

Anthropic’s internal data better illustrates this.

Boris Cherny, head of Claude Code, says he now writes 100% of his daily code with Claude Code, and the entire engineering team generates 70% to 90% of their code with it. 90% of Claude Code’s own codebase is also self-written.

Pragmatic Engineer’s developer survey in February 2026, with 15,000 respondents, ranked Claude Code as the most popular AI coding tool. By early 2026, 4% of public commits on GitHub came from Claude Code, expected to exceed 20% by year’s end.

Claude Code’s success reveals a harsh reality many in the AI industry are reluctant to face: the commercial ceiling of chatbots may be very low. The real value for enterprises is embedding AI into workflows and replacing specific roles.

ChatGPT opened the door to AI, but whether you turn left or right after entering determines who can convert users into revenue. Anthropic took the right turn, into the enterprise production chain.

In January 2026, Anthropic released Cowork, extending the same approach from developers to all white-collar jobs. A team of four engineers built it in 10 days, most of the code generated by Claude Code itself.

Since its launch, the global SaaS sector has evaporated about $2 trillion in market value.

People

Product and strategy are obvious differences, but the real key is: people.

Looking at OpenAI, from 2024 to 2025, the company experienced a systematic wave of executive departures.

Co-founder and Chief Scientist Ilya Sutskever left to start Safe Superintelligence. CTO Mira Murati left to found Thinking Machines Lab. Co-founder John Schulman and the head of the Superalignment team Jan Leike went to Anthropic.

Chief Research Officer Bob McGrew left, VP of Research Barret Zoph left, Co-founder and President Greg Brockman took a long sabbatical. By summer 2025, at least seven researchers had been recruited away by Meta’s superintelligence lab.

Of the original 11 co-founders, only Sam Altman and researcher Wojciech Zaremba remained full-time by the end of 2025. A former employee told Fortune: “Without Ilya, OpenAI is a different company; without Greg, it’s a very different company.”

On the other side, Anthropic presents a different picture.

Seven co-founders—Dario Amodei, Daniela Amodei, Jared Kaplan, Jack Clark, Sam McCandlish, Ben Mann, Tom Brown—all still there, with no public executive departures in five years.

This stark contrast prompts a question: what has Anthropic done to keep people?

Forbes estimated early 2026 that each of the seven co-founders held about 1.8% of the shares, a very close split. At a $380 billion valuation, each share is worth roughly $6.8 billion. This nearly equal ownership structure is very different from Silicon Valley norms, where typically the CEO holds the largest share, and others hold decreasing amounts. Equal shares help eliminate common fissures—no one feels they’re getting shortchanged.

But equity is only the surface. More important is Dario Amodei’s time commitment to management.

He said on Dwarkesh Podcast that he spends about one-third to 40% of his time “ensuring Anthropic’s culture is good.” For an AI company CEO, this is an unusually high proportion. As the company grows to 2,500 people, he can no longer participate in every technical and product decision, so he chooses to focus on more “leverageable” things: aligning everyone’s direction.

How does he do that?

He holds a company-wide meeting every two weeks, called “DVQ”—Dario Vision Quest. The name was coined by employees; Dario himself once wanted to change it because it sounded like a hallucinogenic experience. Each session, he prepares three or four pages of documents and speaks for an hour in front of the entire company, covering product strategy, geopolitical trends, and industry outlooks. Most employees attend in person or remotely.

On a more daily level, Anthropic has a “notebook channel” Slack culture. Every employee, including Dario himself, maintains an open Slack channel to share ideas, work progress, and even doubts.

Growth lead Amol Avasare compared it on Lenny’s Podcast to “internal Twitter feeds,” where you can jump into research or other team channels to see what they’re thinking. Dario encourages employees to “debate with him directly.”

In an interview with Fortune, he said: “My goal is to build a reputation for telling the truth about the company, pointing out problems directly, avoiding ‘corporate speak’—that defensive, politically correct language. If you hire people you trust, you can communicate completely unfiltered.”

This “anti-public relations” internal communication style sharply contrasts with OpenAI. During the 2023 board crisis, internal information was so fractured that even the CTO wasn’t sure what was happening.

Anthropic’s culture filtering begins at recruitment. Every candidate, regardless of position, must go through a standardized “culture interview.” Only after 30 days of onboarding and multi-stage cultural training can someone become a culture interviewer. The logic: culture transmission is too important to be left to someone who hasn’t fully grasped what the company culture is.

Reportedly, one question in the culture interview is: “If Anthropic decides not to release a model because of safety concerns, and your equity is wiped out, would you accept that?”

This isn’t rhetorical. No matter how strong your technical skills, those who fail this question won’t be hired.

Another detail: all technical roles at Anthropic—from new hires to senior executives—use the same title: “Member of Technical Staff.” There are no distinctions like “Senior,” “Chief,” or “Distinguished.” Internally, employees call each other “ants,” from the abbreviation of Anthropic.

The company even hired a full-time philosopher, Amanda Askell, whose job is to shape Claude’s moral judgment framework. She told Time: “Sometimes it feels like you have a 6-year-old child, teaching it what’s good. But when it grows to 15, it will be smarter than you in every way.”

Daniela Amodei’s role is often underestimated within this system.

Dario is the technical visionary and external spokesperson; Daniela handles execution, culture, talent, and operational infrastructure. Reportedly, the research, product, sales, and operations executives all report directly to her. She has a clear hiring preference: find people with strong communication skills, high EQ, kindness, curiosity, and a willingness to help others. In an industry dominated by technical founders, this emphasis on “soft skills” is uncommon.

All seven Anthropic founders have signed a pledge to donate 80% of their wealth. Nearly 30 employees signed up for the 2026 EA (Effective Altruism) conference in San Francisco, more than double the combined attendance of OpenAI, Google DeepMind, xAI, and Meta’s superintelligence labs.

The core asset of an AI company is the human brain. Code can be copied, compute can be bought, but researchers’ intuition and judgment cannot be taken away.

When your chief scientist, CTO, and chief research officer leave within two years, what you lose can’t be measured by funding. Anthropic’s stability at the talent level may be its most difficult advantage to replicate.

All victories are victories of values.

What happened to OpenAI?

A fair word for OpenAI here.

Anthropic’s revenue surpassing OpenAI’s, and the secondary market sentiment shifting—yet OpenAI isn’t collapsing. It just completed a $122 billion funding round, with participants including Amazon, Nvidia, SoftBank, and Microsoft. ChatGPT still has 900 million weekly active users.

In consumers’ minds, “AI” and “ChatGPT” are almost synonymous, but OpenAI does face some structural issues, which exploded in 2026.

Financial pressure is the most direct.

OpenAI expects a loss of $14 billion in 2026. From 2023 to 2028, cumulative losses could reach $44 billion. HSBC analysts believe profitability won’t arrive before 2030. The Wall Street Journal estimates that by 2030, OpenAI’s annual training costs will hit $125 billion, while Anthropic’s will be around $30 billion—four times less. Both are working on cutting-edge models, but the cost gap needs explanation. Part of it is OpenAI’s more aggressive infrastructure investment; another part may be efficiency issues. Capital markets are clearly concerned about this gap. Anthropic expects to turn positive cash flow by 2027, while OpenAI pushes breakeven to 2030.

Product-wise, there are also issues.

Sora shut down in March 2026. This video generation tool reportedly cost $15 million daily to operate, with only $2.1 million in revenue. The shutdown also sabotaged a $1 billion-level partnership with Disney. OpenAI’s new AGI deployment head Fidji Simo told employees plainly that the company “cannot afford to be distracted by side projects.”

Then there’s advertising. In February 2026, OpenAI added ads to ChatGPT’s free and pro versions. Not a big deal in itself—many products have ads. But for OpenAI, it’s stark because Sam Altman explicitly said in 2024 that ads are a “last resort,” and he felt “uncomfortable” with the combination of AI and advertising. Just 15 months later, they launched it. With only 5% of 900 million users paying, this pushed him to make that choice.

Corporate governance is even more complicated. The nonprofit-to-profit restructuring took nearly a year. Elon Musk’s lawsuit, former employees’ joint opposition letter, Nobel laureates’ open letter, and investigations by California and Delaware attorneys general—all culminated in the October 2025 restructuring. The nonprofit foundation retained 26% of shares and control. Critics say this arrangement is largely symbolic.

Individually, none of these issues are fatal. Together, they paint a less-than-rosy picture: a once industry-leading company now dominated by headlines about governance disputes, product shutdowns, and ads.

The War Isn’t Over

Anthropic’s momentum is indeed fierce. Revenue overtaking, market enthusiasm, and the Pentagon’s PR blitz. But one thing to remember: if you asked any industry analyst at the end of 2023 whether OpenAI would be surpassed, 99% would say impossible. The rapid shift in consensus itself should make us cautious about current beliefs.

A few high-confidence facts: Anthropic’s enterprise focus is correct; 80% of its revenue structure is healthier than ChatGPT’s consumer model, supported by solid financial data. Claude Code is a genuine breakthrough, reaching $2.5 billion ARR in just nine months.

But many uncertainties remain. OpenAI still commands 900 million weekly active users and the world’s strongest AI brand recognition. If it finds an effective way to monetize consumers—say, increasing the paid rate from 5% to 10%—the story could be rewritten. The AI industry’s characteristic is that a major breakthrough can reshape everything overnight.

While secondary market capital flows point in one direction, they also chased WeWork.

The more cautious conclusion: in the first wave of AI commercialization, Anthropic’s path has been validated; OpenAI’s approach is being questioned. But “winner or loser” isn’t decided yet—this battle is only halfway through.

When Dario Amodei left OpenAI with six others in 2021, no one could have foreseen today’s situation. A safety researcher, in an industry obsessed with speed, using less funding and more self-discipline, has pushed his former employer into a position where it must write memos to investors explaining its competitiveness.

The most interesting part of this story is—it’s still unfinished.