The U.S. AI power leader has changed! Anthropic’s annual revenue is 300 billion, crushing OpenAI

Written by: New Intelligence Institute

A historic moment!

Just now, a financial report worthy of being recorded in business history dropped like a deep-water bomb, completely shattering OpenAI’s growth myth.

The historic moment has arrived—Anthropic, the “Avengers” founded by former OpenAI employees who defected, has officially reached an annual revenue of $30 billion, surpassing OpenAI’s $24 billion.

Today, foreign media WSJ exclusively revealed a number that made Silicon Valley gasp: Anthropic’s annualized revenue (ARR) exceeded $30 billion.

At the beginning of 2025, Anthropic’s ARR was only $1 billion. In 15 months, it increased thirtyfold.

In just 15 months, the annualized revenue skyrocketed from $1 billion to $30 billion—a terrifying growth rate never seen before in American business history, even making Google and Meta look like they are crawling.

The fastest-growing company in American business history is not OpenAI, not Nvidia, but a company founded only in 2021 by former OpenAI employees.

The American AI dominance has officially changed hands!

Now, OpenAI has been mocked by foreign media as “Anthropic’s little brother,” with revenue and valuation both surpassed, and internal panic has triggered a “crazy copying” mode.

What’s more brutal is that on the same day, WSJ delivered another blow: OpenAI failed to meet its internal target of 1 billion weekly active users, and the CFO expressed concerns about whether they can afford the $600 billion in computing power bills, leading to a divergence in strategy between the CEO and CFO.

With $122 billion in financing still fresh, the back yard is already on fire.

Now, the dragon-slaying youth has become king.

How did Anthropic’s $30 billion come about?

Moreover, by the end of 2025, Anthropic’s ARR was still about $9 billion; by April this year, it was officially confirmed to have surpassed $30 billion. In just four months, it tripled again.

This acceleration is truly terrifying.

How did Anthropic achieve an annualized revenue of $30 billion?

Why does OpenAI, with 900 million weekly active users, get overtaken in revenue by “latecomers”?

The secret is only three words: enterprise side. Yes, 80% of Anthropic’s revenue comes from the enterprise sector.

It has 300k corporate clients, eight of the top ten Fortune 500 companies use Claude, and over 1,000 large clients spend more than $1 million annually. Moreover, this number doubled in two months.

OpenAI’s embarrassment lies in falling into the most typical “revenue growth without profit” trap of the internet era.

Among its 900 million users, the vast majority are “free riders” using ChatGPT for homework, weekly reports, or just casual chatting. Maintaining this traffic costs sky-high inference expenses.

This is a clear contrast of two completely different business models: one is a C-end traffic funnel with many users but low payment rate; the other is a B-end subscription engine with fewer but high-value users.

Now, the real market results are out—B-end wins.

Even more interesting is Anthropic’s spending pattern. While achieving this success, Anthropic’s investment in model training is only a quarter of OpenAI’s.

In terms of return on technological investment, Anthropic is not just winning, but crushing.

While Ultraman was still begging for “trillion-scale computing power plans,” his former subordinate Dario Amodei had quietly taken Claude and encroached on OpenAI’s backyard.

Dario Amodei’s “defection,” Ultraman’s nightmare

The most exciting script in the business world is “the exile’s comeback.”

In 2021, due to disagreements over OpenAI’s increasing commercialization and safety philosophy, Dario Amodei, along with his sister Daniela and a group of core researchers, left en masse to establish Anthropic.

At that time, industry insiders even joked that they were “idealists suicidal with passion.”

Ultraman probably didn’t take this matter seriously at the time.

Four years later, this “rebellious team” gave him an answer with $30 billion in revenue.

As mentioned above, the key lies in their focus on the B-end enterprise market.

B-end cash ability vs. C-end wool-pulling

By April 2026, over 1,000 enterprise clients pay Anthropic more than $1 million annually. Giants like Amazon, Google, Salesforce, Accenture, and Deloitte are not just investors but deep-paying users.

And Claude Code has become a key weapon. Enterprise developers use it to write code, tune architectures, and deploy—an embedded workflow necessity. Once B-end clients adopt it, switching costs are extremely high, and renewal rates are astonishing.

Where does Anthropic truly win?

It turns “safety” from a marketing label into a business barrier. Financial, medical, legal industry clients fear not that AI isn’t smart enough, but who is responsible if AI goes wrong.

Anthropic’s Constitutional AI framework, interpretability research, and responsible deployment commitments have become hard requirements on CTOs’ procurement lists.

In contrast, OpenAI, WSJ revealed a set of painful data: ChatGPT repeatedly failed to meet the internal target of 1 billion weekly active users, and revenue in several months was below expectations.

Backyard on fire! OpenAI’s power game

If external competition is a sharp pain, internal cracks are a fatal wound.

Just as Anthropic celebrates victory, WSJ and The Information have repeatedly dropped heavy bombs—OpenAI is already in chaos internally!

Disputes among leaders

Ultraman and the company’s new CFO Sarah Friar are at odds.

As a top financial expert from Square and Nextdoor, Friar’s task was to safeguard OpenAI’s IPO, valued at $122 billion or even $1 trillion. But now, she is on the verge of “going berserk.”

In internal meetings, Friar strongly questioned Ultraman’s aggressive expansion plans for computing power. Ultraman wants to buy all the graphics cards on Earth, and Friar is trembling at the bills.

She has privately warned colleagues that if revenue doesn’t accelerate, OpenAI may be unable to pay future compute contracts, and Ultraman has excluded her from key infrastructure meetings.

The compute black hole

Last year, Ultraman signed a $600 billion compute contract.

This means that even with the recent $122 billion in record financing, if current spending continues, the money will only last three years.

Friar worries that if revenue growth cannot remain exponential (in fact, ChatGPT’s weekly active users and revenue have repeatedly underperformed), OpenAI will crash into a financial iceberg.

Forecasts suggest OpenAI might burn over $300k before achieving stable cash flow—an unprecedented rate of spending!

IPO deadlock

Ultraman is eager to ring the bell by the end of this year, completing a historic listing. But Friar privately told colleagues that OpenAI’s internal controls are terrible and not ready for the rigorous audits of the secondary market.

Now, their differences are public. Although the official statement claims they are “completely aligned,” this pale denial is powerless against Nasdaq’s 1% dip.

The plunge in SoftBank, Nvidia, and other partners’ stock prices already signals market concerns about OpenAI’s stability.

A quarter of costs, three years’ delay

Revenue overtaking is only superficial. The real difficulty for OpenAI lies in cost structure gaps.

According to internal financial documents from WSJ and third-party analysis by SaaStr, a leading SaaS industry authority, OpenAI’s projected annual compute expenditure by 2028 could reach about $121 billion, while Anthropic’s peak training cost is around $30 billion—less than a quarter of the former.

What does this number mean?

Using the same dollar, Anthropic can train models that cost OpenAI four dollars to achieve.

When your costs are a quarter of your competitor’s, your pricing, profit, and survival space are all several times larger.

Who wins the profit marathon first?

Even more deadly is the difference in profit timelines: Anthropic expects to turn positive cash flow by 2027, while OpenAI’s target is pushed back to 2030.

Three years. In the AI industry, three years is an epoch. It’s enough for a company to build an insurmountable ecosystem moat or for a once-unicorn to collapse due to cash flow issues.

Now, the situation is very clear.

Anthropic: expects to achieve positive cash flow by 2027. With a healthy B-end paid model and high R&D efficiency, it is moving lightly.

OpenAI: profit goal set for 2030. Burdened with heavy C-end baggage and astronomical compute debt, it struggles forward.

Anthropic just announced a partnership with Google and Broadcom for compute resources, further securing its infrastructure advantage. Meanwhile, OpenAI’s board is questioning whether Ultraman’s expansion strategy is sustainable.

Internal panic is spreading: OpenAI’s Chief Revenue Officer directly named Anthropic, warning employees to stay alert, “The market competition is fiercer than I’ve ever seen.”

More symbolically, OpenAI’s response has been to cut projects like Sora, shift fully to B2B, and aggressively follow up with products like Claude Code, step by step.

According to The Atlantic, a related report shows Claude-generated infographics

When you start mimicking your competitor’s strategy, it means you admit one thing: their route is right, yours is wrong.

The $122 billion just arrived, and the truth is out

What’s even more intriguing is the timing of this overtaking.

Just a few weeks ago, OpenAI completed a round of shocking funding: $122 billion in total, with a valuation soaring to $800 billion.

Before the money even dried, WSJ revealed a layer of the curtain.

OpenAI’s current annual revenue is about $24 billion, with operating costs expected to exceed $20 billion. Profit margins are razor-thin, and profitability is far off.

What truly hangs over OpenAI is a bigger bill: the “Stargate” project’s total investment is up to $600 billion.

$122 billion against $600 billion looks more like a painkiller.

This forms a concerning financial cycle: chip manufacturers invest in OpenAI, OpenAI uses that money to buy chips, the chips’ computing power sustains free users who don’t generate profit, and to keep growing, OpenAI needs more funding.

OpenAI’s revenue executives even internally accused Anthropic’s revenue calculation method of “inflated,” but no one denies the fact: Anthropic has indeed overtaken.

This is also the root of internal conflicts at OpenAI.

Accountants’ rationality tells Sarah Friar that the company is heading toward a bottomless capital black hole; while dreamers’ enthusiasm keeps Ultraman going—stop expanding, and the $300 billion valuation could instantly loosen.

The new players in the trillion-dollar club

The capital market’s reaction is more honest than any analysis.

Anthropic’s secondary market valuation is approaching $1 trillion.

Some early investors in OpenAI are already wavering—investors are reassessing the reasonableness of OpenAI’s valuation.

Concept stocks deeply tied to OpenAI like Oracle and CoreWeave have pulled back. The market is voting with real money.

But what’s most interesting about this reversal isn’t who wins or loses, but the industry pattern it reveals.

OpenAI chose the “first attract users, then find a business model” internet approach. Anthropic is adopting the enterprise software approach: “first create product value, then wait for scale effects.”

This rivalry between two routes is actually the industry’s answer to a fundamental question: Will the commercialization of large models ultimately lead to consumer internet or enterprise infrastructure?

Anthropic bets on the latter with $30 billion in revenue. At least for now, the market is on its side.

Someday, OpenAI or Anthropic may conquer cancer or reshape the world, but for now, they still need to pay the bills.

The rebel’s masterstroke

In 2021, when Dario Amodei left OpenAI, almost no one believed in him.

Four years later, Anthropic not only survived but also overtook.

This raises a question: if Ultraman could go back to 2021, would he stop that “defection”?

The answer might be: even if he did, it wouldn’t help. Anthropic’s success isn’t just because a few people were poached; it’s because it found a path that OpenAI overlooked.

The battle for AI dominance is far from over. OpenAI still has GPT-5 technology, Microsoft’s ecosystem support, and the world’s largest AI user base, but the balance of power has shifted.

Ultraman’s anxiety is now off the charts. He has to deal with Musk’s lawsuits, soothe internal rebels, and answer investor questions.

This epic AI war has only just entered the second half. Who will laugh last?

But one thing is certain: the era of OpenAI’s monopoly has officially ended.

References: