On May 29, 2026, a news flash quickly spread within the AI ​​investment community: Anthropic completed a $65 billion Series H funding round, bringing its post-money valuation to $965 billion, and its annual recurring revenue "has reached $47 billion, surpassing OpenAI's $25 billion." Anthropic has overtaken OpenAI in ARR.

This overtaking didn't happen suddenly in May 2026. If we go back to the starting point in 2021, Anthropic's valuation was $550 million, while OpenAI's ARR had already exceeded $1 billion during the same period. Within five years, a small team that left OpenAI completed a full curve from following to keeping pace and then surpassing it in metrics.

It started with 550 million, but the spotlight wasn't on it.

In 2021, Dario Amodei led a group of former OpenAI researchers to found Anthropic. According to public reports, the company raised approximately $124 million in seed and Series A funding rounds, with a post-money valuation of approximately $550 million.

While its initial valuation was low, the founding team's credentials and its differentiated positioning as "safe AI" quickly attracted strategic capital. Between 2022 and 2023, Google and Spark Capital invested, driving the valuation up gradually. By early 2024, Anthropic's valuation had risen to approximately $18 billion, with Amazon's investment being a significant driver of this valuation.

Valuation logic at this stage is almost unrelated to commercial revenue. Anthropic released its first product, Claude 1, in March 2023, followed by Claude 2 in July of the same year. Both models have done a lot of work on safety alignment, but they lag significantly behind OpenAI's GPT-4 in terms of inference capabilities, context windows, and developer experience.

The bigger problem was timing. In November 2022, OpenAI released ChatGPT, instantly capturing all public attention in the AI ​​industry. By the time GPT-4 was released in March 2023, OpenAI's annual recurring revenue had already surpassed $2 billion. Anthropic's Claude 1 was released around the same time, but it was completely overshadowed by the brilliance of GPT-4.

From a financial perspective, Anthropic generated almost no large-scale commercial revenue between 2021 and 2023. Its valuation growth relied on the appeal of the "safe AI" narrative to strategic capital. Google and Amazon's investments in Anthropic were less a bet on a revenue engine and more a defense against potential threats from OpenAI in cloud services and search.

The gap between Anthropic and OpenAI during this period was not a lag in the same field, but rather a fundamental difference in their commercialization paths. OpenAI had consumer traffic, a developer ecosystem, and ARR that doubled year after year; Anthropic only had a security narrative and two generations of models that had not yet penetrated the market.

A single model update boosted ARR from zero to $1 billion.

The turning point came in 2024.

In March 2024, Anthropic released the Claude 3 series, including Haiku, Sonnet, and Opus versions. This marked the first time Claude had matched GPT-4 in some benchmarks, but it did not trigger a large-scale migration in the developer community.

What truly changed the situation was Claude 3.5 Sonnet, released in June 2024.

In coding and inference tasks, Claude 3.5 Sonnet achieved performance superior to GPT-4o at the same stage with lower API call costs, and its cost-effectiveness quickly sparked discussion in the developer community. On the ClaudeAI channel on Reddit and Hacker News, many developers began sharing their experiences migrating from GPT-4o to Claude 3.5 Sonnet, with coding capability being a frequently mentioned keyword.

More importantly, there's the distribution channel on the enterprise side. During the same period, Anthropic's models were deeply integrated with the enterprise customer systems of two major cloud platforms: AWS Bedrock and Google Cloud Vertex AI. For enterprises already running workloads on AWS or Google Cloud, accessing the Claude API is much more natural than accessing OpenAI's API separately. Call logs, billing management, and security compliance can all be handled within the same cloud account.

ARR's response came quickly. According to data reviewed by Anthropic in its subsequent Series F funding announcement, the company's annual recurring revenue exceeded $1 billion by the end of 2024.

Valuation reflects this breakthrough. According to Forbes, Anthropic's valuation rose from approximately $18 billion at the beginning of the year to the $60 billion range in 2024. This time, the core logic behind the valuation is no longer the "security narrative," but is supported by actual API call volume and enterprise contracts. The growth engine has shifted from defensive investment-driven to product-driven.

However, Anthropic's $1 billion ARR is still an order of magnitude smaller than OpenAI's approximately $6 billion ARR during the same period. Claude 3.5 Sonnet merely opened a gap in the enterprise market, moving Anthropic from being "incomparable" to "lagging behind by half a step in the same track." True catching up requires a more explosive growth engine.

Claude Code becomes a money printing machine

The engine will be ignited in 2025.

In May 2025, Anthropic released the Claude 4 series, including Opus 4 and Sonnet 4. The focus of the two models is not on increasing general capabilities, but on in-depth optimization for specific scenarios: long context processing and agent encoding.

Shortly after the release of Claude 4, Anthropic launched Claude Code, a code generation and automation agent tool for developers and enterprises. It's not just a simple code completion plugin, but an execution agent capable of autonomously planning multi-step tasks: reading the entire code repository, understanding the project structure, generating complete functional modules, executing commands in the terminal, and iteratively modifying them.

This product form addresses high-frequency needs in enterprise development scenarios. According to industry tracking data cited by China Venture Capital Network, by early 2026, the annual recurring revenue of Claude Code alone exceeded $2.5 billion. Among mainstream coding tools such as GitHub Copilot and Cursor, Claude 4 and subsequent versions have significantly better user reviews than competing products of the same period.

Anthropic's ARR growth curve will begin to rise sharply from mid-2025. According to Reuters, Anthropic's annual recurring revenue will reach $3 billion by mid-2025; according to industry tracking data, this figure will soar to approximately $9 billion by the end of 2025.

The valuation response was almost simultaneous with ARR. In September 2025, Anthropic completed a $13 billion Series F funding round, led by ICONIQ, valuing the company at $183 billion post-money. Five months later, in February 2026, the company completed a $30 billion Series G funding round, led by GIC and Coatue, doubling its valuation to $380 billion.

From 61.5 billion to 183 billion and then to 380 billion. Each of these three valuation leaps in 2025 was directly supported by triple-digit ARR growth. The growth engine is no longer just "product-driven," but has found a model with even greater commercial leverage than C-end subscriptions: enterprise-level code generation and automated agent workflows.

Claude Code's value proposition is fundamentally different from ChatGPT's subscription revenue. A company can pay tens of thousands of dollars per month for Claude Code because it replaces not the "cost of acquiring information," but the "cost of hiring developers." This represents a much higher base of willingness to pay, and is the core reason why ARR surged from $1 billion to $9 billion in 18 months.

Meanwhile, OpenAI's growth did not stagnate. According to official OpenAI announcements and Sacra's tracking data, OpenAI's annual recurring revenue exceeded $20 billion in 2025, still more than double that of Anthropic. However, the growth rate had slowed, from three times in 2024 to approximately 3.3 times. Anthropic, on the other hand, experienced exponential growth, with the slopes of its two ARR curves undergoing a substantial change by the end of 2025.

47 billion versus 25 billion: A belated numerical reversal

The reversal will materialize in the first five months of 2026.

According to Yahoo Finance, Anthropic's ARR reached $19 billion in March 2026. Industry tracking data from The AI ​​Corner and arr.club shows that ARR surpassed $30 billion in April 2026, officially exceeding OpenAI's approximately $25 billion during the same period. By May 2026, industry analyst tracking data from Sacra and LinkedIn indicates that Anthropic's ARR had climbed to between approximately $44 billion and $47 billion.

OpenAI has not been inactive. According to an official OpenAI announcement, in March 2026, the company completed a massive $122 billion funding round, led by institutions such as SoftBank, valuing the company at $852 billion post-money. During the same period, its ARR was approximately $25 billion, corresponding to a price-to-sales ratio of approximately 34. This valuation logic includes the expected premium for OpenAI's consumer brand, advertising model, and platform ecosystem.

However, the fact that ARR was overtaken changed the relative positions of the two companies in the capital market. Two months later, in May 2026, Anthropic completed a $65 billion Series H funding round, with a post-money valuation of $965 billion, surpassing OpenAI in valuation metrics. Based on an ARR of approximately $47 billion, this corresponds to a PS ratio of about 20. In the SaaS and high-growth AI sectors, this ratio is at the high end of a historically reasonable range, but not extreme.

The release of Opus 4.8 in May, at this juncture, seems more like a signal than a growth engine. According to Anthropic's official announcement, the pricing of Opus 4.8's regular API remains the same as its predecessor. The main improvements focus on the reliability of Agent task execution and "code honesty." Models are more inclined to acknowledge rather than generate flawed code when they find themselves unable to complete a task. Anthropic's continued emphasis on reliability rather than general-purpose capabilities after reaching the top of the valuation charts indicates that it will not deviate from its proven high-value B2B workflow approach in the short term.

Two forked roads meet five years later.

The double overtaking in ARR and valuation is a fact from a financial perspective, but to understand the significance of this result, we need to go back to the fork in the road five years ago for the two companies.

OpenAI has followed a path built on a foundation of consumer-facing (C-end) traffic. Since its explosive growth at the end of 2022, ChatGPT has rapidly accumulated hundreds of millions of users, establishing the strongest public brand recognition in the AI ​​industry. Based on this, OpenAI has expanded into a multi-layered commercialization structure, including enterprise APIs, Team subscriptions, advertising models, and a platform ecosystem. ARR (Average Revenue Per Transaction) has grown linearly but healthily, from $2 billion to $6 billion to $25 billion.

Anthropic took an almost opposite path. Between 2021 and 2023, it almost completely missed the consumer-facing (C-end) boom sparked by ChatGPT, lacking its own ToC applications and brand recognition. However, after 2024, through deep integration with AWS and Google Cloud, it directly embedded its model into existing enterprise IT infrastructure. When Claude 3.5 Sonnet established a localized advantage in coding tasks, and Claude Code proved the commercial value of agent workflows in 2025, the migration cost for enterprise customers was far lower than switching chatbots—because they were switching API providers on the same cloud platform using the same security and compliance framework.

ARR's overtaking is due to the time difference between these two paths during the 2025-2026 window. Anthropic's choice cost it two years of silence in the early stages, but its leverage effect only became apparent after its value to enterprises was realized. OpenAI's choice has brought it public awareness and brand barriers that continue to this day, but the ceiling for C-end payment and the uncertainty of the advertising market are structural risks that have not yet been fully tested.

ARR's overtaking does not mean the debate over the right path has been finalized. Anthropic's revenue concentration, where a few large enterprise customers or AWS/Google Cloud channels contribute the vast majority of revenue, will be a variable in its growth sustainability. Meanwhile, OpenAI's $122 billion funding means it still has ample ammunition to launch a new round of competition on either dimension. The two paths meet in May 2026, but neither has yet proven to be the only correct answer.