Anthropic quarterly profit exceeded 1 billion USD, large language models can truly generate profits now

Large language model companies may truly be evolving into highly profitable businesses. 

On July 8, semiconductor and AI research firm SemiAnalysis released its latest report, projecting that Anthropic's GAAP EBIT profit in the third quarter of this year will exceed $1 billion.

Meanwhile, the company's annual recurring revenue (ARR) has surged from approximately $9 billion at the end of 2025 to over $60 billion.

It is important to emphasize that these figures do not represent Anthropic's official financial statements.

As a privately held company that has not yet gone public, Anthropic does not fully disclose its quarterly performance. SemiAnalysis used its Tokenomics model to estimate the company's revenue and costs from the bottom up, segmented by product, service tier, and customer type.

Even so, these numbers remain remarkably striking.

Over the past few years, the biggest question the public has had about large language model companies is whether they can actually retain profits after revenue growth.

But Anthropic is now presenting a different outcome:

A cutting-edge AI company can not only turn a profit, but also rapidly grow into a highly profitable enterprise with quarterly earnings exceeding $1 billion.

ARR Surges from $9 Billion to $60 Billion

Connecting the financial trajectory of Anthropic over the past several quarters makes the transformation even more intuitive.

At the end of 2025, the company's ARR stood at around $9 billion.

By the first quarter of 2026, Anthropic's three-month revenue had already reached $4.8 billion.

In the second quarter, the company's projected revenue further rose to $10.9 billion, more than doubling quarter-over-quarter.

More critically, Anthropic is expected to record an adjusted operating profit of $559 million in the second quarter.

This marks the first time the company has achieved quarterly profitability under this accounting metric. This profit figure incorporates new model training costs but excludes stock-based compensation.

By the third quarter, SemiAnalysis is no longer discussing when Anthropic will become profitable, but instead projecting that its ARR will exceed $60 billion and its GAAP EBIT will cross the $1 billion threshold.

In just over half a year, ARR has surged by more than 6 times.

As a point of reference, for most SaaS companies labeled as high-growth, doubling their full-year revenue already places them in the top tier.

Salesforce took nearly 20 years to grow from zero to $30 billion in annual revenue; Anthropic reached a comparable scale in less than 3 years.

Of course, these figures cannot be simply regarded as a set of audited consecutive financial statements, but the underlying trend is consistent: Anthropic's revenue is exploding, and its profit margins are rapidly turning positive.

This is precisely why, after Anthropic confidentially filed for its IPO on June 1, the capital market has paid extremely close attention to its financial status.

Once successfully listed, it will not only likely become one of the largest AI lab IPOs in history, but also place the actual performance of a pure cutting-edge model company fully on display in the public market for the first time.

How Robust is Anthropic's Performance?

To understand how exceptional Anthropic's results are, OpenAI serves as the most intuitive benchmark.

OpenAI posted 2025 revenue of $13.07 billion, an operating loss of $20.92 billion, and a net loss attributable to the company reaching $38.53 billion.

In the first quarter of this year, according to The Information, OpenAI's revenue was approximately $5.7 billion, with an adjusted operating margin of -122%.

This equates to roughly $1.22 in operating losses for every dollar of revenue earned.

The revenue structures of the two companies are also markedly different.

Per SemiAnalysis estimates, OpenAI still derived over 65% of its revenue from the subscription model in the first quarter of this year, with consumer ARR accounting for around 40%;

For Anthropic, 75% to 85% of its ARR comes from usage-based API business.

This means OpenAI's growth relies more on converting paying users, expanding seat counts, and upgrading subscription plans, while Anthropic can more directly capture revenue from the continuously increasing token consumption of the same group of enterprise customers.

While these are not strict performance comparisons, even setting aside accounting standards, one trend has become increasingly clear:

Competing in the same high-cost cutting-edge model race, Anthropic is demonstrating that revenue and profit growth do not necessarily have to come at the expense of burning cash.

Looking at the domestic market, no company has yet replicated a similar success story.

Zhipu reported 2025 revenue of 724.3 million RMB, representing a 131.9% year-over-year increase, but its adjusted net loss reached 3.182 billion RMB.

MiniMax generated full-year revenue of $79 million, with an adjusted net loss of $250.9 million.

Both companies maintain rapid growth, but they are still far from the profitability inflection point where model revenue covers R&D and operational investments.

Cutting-Edge Models Possess Pricing Power

Why is it Anthropic that has managed to turn a profit first? The most straightforward answer is Claude Code.

SemiAnalysis estimates that Claude Code-related contributions now account for over 7% of all code submissions on GitHub.

In the first quarter of this year, fueled by the explosive growth of Claude Code, Anthropic's net monthly new ARR surged from approximately $3 billion in January to around $11 billion in March.

Unlike ordinary chatbots, coding and Agent scenarios are typically high-value, high-frequency use cases where enterprises are purchasing tangible productivity gains.

This has also transformed the revenue logic for model companies: the more frequently Agents are invoked, the greater the token consumption, and the higher the revenue the model company generates.

But what deserves more attention behind this is not just increased token sales, but the fact that cutting-edge models like those from Anthropic may wield stronger pricing power.

SemiAnalysis also regards pricing capability, business model, gross margin, and profitability as Anthropic's most critical current advantages.

If SemiAnalysis's estimation direction holds true, cutting-edge model companies could form a genuine closed business loop:

Improved model capabilities unlock access to more high-value enterprise scenarios; increased enterprise usage drives high-quality token revenue; enhanced inference efficiency lifts gross margins; profits are reinvested in next-generation model training to further expand the capability lead.

It is fair to say that Anthropic has once again raised the bar for model competition.

Moving forward, all model companies that still rely on burning cash to drive growth will have to confront a harsh question: why can't they achieve the same results?

This article originates from the WeChat public account "World Model Workshop", authored by World Model Workshop, and published with authorization from 36Kr.