Anthropic revenue surpasses OpenAI: A historic turnaround in the AI arms race

The leading companies in the AI industry are set to experience a historic reversal in revenue by 2026. The Information reports that Anthropic’s annualized revenue has surpassed OpenAI, making it the highest-earning AI company worldwide.
(Background: Anthropic’s enterprise adoption rate has overtaken OpenAI)
(Additional background: Anthropic’s valuation has surged to $800 billion)

The revenue rankings of the two AI giants were completed with a dramatic turnaround in the first half of 2026. On May 27, The Information reported that Anthropic’s current annualized revenue is close to $450 billion, while OpenAI’s latest annualized revenue is about $330 billion—meaning Anthropic’s revenue scale has surpassed OpenAI by at least 35%.

Table of Contents

Toggle

• Enterprise AI becomes a key variable
• OpenAI’s hidden cost burden
• What it means for the AI industry
• AI adoption status in Taiwan

The pace of this reversal is faster than the market expected. At the end of 2025, Anthropic’s annualized revenue was only about $9 billion—less than half of OpenAI’s. In less than half a year, Anthropic’s revenue surged by roughly 5 times, while OpenAI grew by only about 50%. The revenue gap between the two instantly shifted from “the chaser” to “the leader.”

Enterprise AI becomes a key variable

The Information’s report points to a key difference: Anthropic’s explosive growth is mainly driven by enterprise-level AI, code generation, and white-collar work scenarios. This stands in sharp contrast to OpenAI’s strategy of heavily betting on consumer business (ChatGPT Plus subscriptions).

Corporate customers aren’t just buying “faster language models”; they’re automating entire workflows. Anthropic’s Claude series has accumulated three major advantages among enterprise customers:

• Code generation: Tools such as Claude Code and Claude Projects are seeing rapidly rising penetration among enterprise developers
• White-collar workflow: From file analysis to meeting notes, companies are replacing some “middle-layer” manpower with Claude
• Enterprise pricing: Enterprise customers are willing to pay a premium for “predictable output,” rather than being as easily swayed by price wars as consumers are

The revenue characteristics in these areas can be described as “high stickiness” and “high ARPU (average revenue per user),” enabling Anthropic to generate larger annual revenue with a smaller number of customers.

OpenAI’s hidden cost burden

By contrast, OpenAI faces a heavier cost structure. The Information notes that OpenAI’s compute spending and consumer business costs have continued to climb, mainly due to:

• Slowing growth in ChatGPT subscriptions: Customer acquisition costs in the consumer market are rising, but ARPU is being squeezed by price wars
• The compute arms race: OpenAI’s top-tier compute investments (especially its reliance on Nvidia) have not yet fully been translated into revenue
• Cost allocation across diversified products: From GPT-4o to Sora, and then to the Agent ecosystem, OpenAI’s product line is longer than Anthropic’s, but the revenue contribution of each product may not be proportional

This leaves OpenAI facing a classic startup dilemma: high revenue, but profit margins are diluted by multiple product lines and compute spending.

What it means for the AI industry

Anthropic’s revenue turnaround is not just competition between two companies—it also reflects a broader shift in the AI industry’s center of gravity, moving from “consumer applications” to “enterprise productivity tools.”

When AI evolves from “chatbots” into competitors like “programmers, analysts, and copywriters,” the premium that enterprise customers are willing to pay is far higher than what consumers will pay. Anthropic’s early layout in the enterprise market (especially Claude’s reputation in the developer community) is entering a harvest phase in 2026.

AI adoption status in Taiwan

This surge in enterprise AI also has profound implications for the Taiwan market. Taiwan’s manufacturing, semiconductor, and 3C industries are at a critical stage of “optimizing capacity with AI,” and the penetration rate of enterprise AI tools (such as Claude Code) among Taiwan-based developers and engineers is rising.

However, enterprise AI adoption in Taiwan still faces challenges: small and medium-sized enterprises are more sensitive to the costs of AI tools, while large enterprises face trade-offs between “data privacy” and “cloud deployment.” Compared with the United States, Taiwan’s enterprise AI market is more oriented toward “specific industry applications” rather than “cross-industry platforms.” This means that the opportunities for Anthropic and OpenAI in Taiwan may be more concentrated than in the U.S.

The Taiwan government’s policies supporting AI startups are also accelerating enterprise adoption. If Taiwan can successfully move AI applications in manufacturing from “pilot projects” to “scaling up to production,” the benefits from this wave of enterprise AI could be longer-lasting than those from consumer AI.

The first-round results of the AI arms race are already clear: Anthropic, focused on the enterprise market, has overtaken OpenAI, which has heavily bet on consumer services. But this race is far from over—the next moves by these two companies will determine the future landscape of the entire AI industry.