Anthropic S-1 filing reveals Claude Code as primary revenue driver

Crossing the profitability line

Anthropic's recent S-1 filing reveals a significant shift in the company's unit economics. For the first time, the company reported that revenue surpassed its compute and infrastructure expenses. This transition occurred in the first quarter of 2026, with revenue reaching $11.6B against $5.6B in compute costs by the second quarter.

The growth rates highlight a major leap in efficiency for the AI provider. While compute expenses grew by 300% year over year, the company's revenue surged by 1,368% during the same period. This suggests that the cost of running the models is becoming a smaller portion of the overall business as the technology matures.

The company also reported that it is getting cheaper for customers to perform individual tasks. Specifically, Anthropic noted that the Opus 5.5 model reduced the cost per task for customers by approximately 40% compared to the previous Opus 5 version. This trend indicates that the company is successfully optimizing its internal infrastructure to provide more value at lower costs.

The rise of Claude Code

The filing makes it clear that Claude Code is the primary engine for the company's current growth. Anthropic describes the coding agent as the product that most concretely defines what autonomous AI work looks like today. This tool has been a major driver of high-value engagement among enterprise clients who need to automate complex software development workflows.

The spending data for this specific tool is striking. While average Enterprise users without Claude Code spent less than $75 per month, the top 5% of users who adopted the coding agent spent about 20 times that amount. This disparity shows that the most sophisticated users are willing to pay a significant premium for the capabilities provided by the autonomous agent.

Anthropic is also using its own tools to build its products. In September 2026, Claude was responsible for writing over 90% of the contributions to the company's own codebase. This internal adoption serves as a proof of concept for the operating efficiency the company aims to provide to its external customers.

Governance and founder control

The S-1 filing also outlines a complex and restrictive ownership structure that limits the influence of public shareholders. The company utilizes five different share classes, some of which are designed to protect the founders' vision. A group of five financially disinterested trustees elects a majority of the board, while the co-founders maintain a dominant voting position.

Specifically, an LLC held by the seven co-founders owns Class F shares, which carry 50.1% of the total voting power. This structure effectively eliminates the ability of Class A common stockholders to influence major decisions, including a change in control. The company notes that this setup is intended to reduce the pressure to prioritize short-term stock prices over long-term development.

This governance model means that the direction of the company is tied closely to the founders' priorities rather than market demands. While this can provide stability for long-term research, it also means that users and investors have very little say in how the models are deployed or how the company evolves. The filing warns that the company may change rate limits, model capabilities, or even shut down specific models based on their own internal schedules.

Regulatory and legal hurdles

Anthropic has faced significant friction with the U.S. government regarding its safety stances. In February 2026, the company publicly stated that its technology should not be used for mass domestic surveillance or fully autonomous weapons. In response, the Secretary of War designated the company as a supply-chain risk to national security, which effectively barred many military contractors from doing business with them.

The company has fought back against these designations in the court system. While they won a district court ruling that characterized the government's actions as retaliatory acts violating their First Amendment rights, they lost a separate challenge in the D.C. Circuit Court of Appeals. That court ruled that the supply-chain risk designation was proper and did not violate the company's constitutional rights.

These legal battles create a complex environment for any customers who do business with the federal government. The company's refusal to cooperate with certain government requests has led to direct restrictions on its commercial activity with military partners. As the company continues to navigate these disputes, the availability of its tools for government-adjacent work remains uncertain and subject to ongoing litigation.

Future risks and competition

The filing highlights several risks that could impact the company's trajectory in the coming years. One major concern is customer concentration, as the top five customers accounted for approximately 25% of the company's revenue over a nine-month period. Additionally, the company faces constant pressure from competitors who are attempting to replicate, distill, or extract the functionality of their models through various means.

The company also noted that the legal landscape regarding copyright and open-source licenses for AI training remains unresolved. Because courts have not yet established authoritative doctrines on these issues, Anthropic faces ongoing uncertainty regarding the legality of its training data. These factors represent significant risks for investors who are looking for a clear regulatory path forward for the AI industry.

Despite these challenges, the company appears to be in a healthy financial position with $131.7B in cash and short-term investments as of June 30, 2026. They are successfully scaling their revenue while managing the high costs of compute. However, the filing suggests that users should remain aware that model availability and capabilities can change rapidly based on safety, alignment, and infrastructure needs.