Anthropic reveals $17 billion Q3 revenue and $895.4 billion in infrastructure commitments

Scaling the AI business

Anthropic recently updated its financial disclosures in a revised S-1 document submitted to the SEC. This paperwork provides a more detailed look at the company's finances as it moves toward a potential initial public offering (IPO). The filing covers the nine months ending September 30, showing a period of rapid expansion.

The company reported total revenue of $33.24 billion for that nine-month stretch. This represents a roughly 1,200% increase compared to the same period from the previous year. These figures suggest that the demand for Anthropic's AI services is accelerating significantly as the company matures.

During the third quarter specifically, Anthropic generated approximately $17 billion in revenue. While the company is growing quickly, it also reported a net loss of $23.29 billion for that same three-month period. The filing serves as a roadmap for how the company intends to manage this high-growth, high-cost trajectory.

The cost of compute

A major portion of Anthropic's financial picture is dominated by the need for massive computing power. The company disclosed that it has entered into purchase commitments and lease agreements totaling about $895.4 billion. According to the filing, substantially all of this spending is tied to compute and infrastructure.

The costs associated with this infrastructure are rising at a blistering pace. Anthropic noted that its compute and infrastructure expenses increased by $14.4 billion over the nine months ending September 30. This represents a 329% increase compared to the same period in the previous year.

The company explained the reason for these rising costs in the filing. Anthropic stated, "As our compute and infrastructure capacity grows, the associated compute and infrastructure expenses have increased." This confirms that as the company scales its capabilities, the physical requirements for hardware and data centers become exponentially more expensive.

Market jitters and supply

Despite the high revenue, Anthropic faces a complex environment regarding how it acquires the hardware it needs. The company relies on major partners like Amazon, Google, and SpaceX for much of its capacity. It also utilizes neoclouds, which are smaller cloud service providers, to help meet its expansion goals.

Investors are currently expressing concern about the broader AI buildout. There are worries that high interest rates might make it difficult for data center developers and neoclouds to afford the debt needed for construction. Recent quarterly reports from tech giants like Microsoft, Google, and Amazon have added to these concerns regarding cash flow.

The core issue for Anthropic is whether compute capacity can be built fast enough to meet its demands. If the infrastructure cannot come online quickly enough, it could create a bottleneck for the company's growth. This tension between the need for rapid scaling and the physical limits of hardware production is a primary focus for Wall Street.

Safety and risk factors

Anthropic has positioned itself in the market as a safety-conscious alternative to its primary rival, OpenAI. To reflect this, the company included more than 90 pages of risk factors in its S-1 filing. These risks include potential political headwinds and the difficulty of balancing commercial goals with safety protocols.

The company also highlighted the complexities of its infrastructure needs as a primary risk. Because the AI industry is so capital-intensive, any disruption in the supply chain or a spike in costs could impact the business. The filing makes it clear that the path to profitability involves navigating these significant operational hurdles.

While OpenAI has also filed a prospectus, it is aiming for a 2027 debut. Anthropic's move to update its filing now puts it on a different timeline. The company must now prove to investors that it can manage its massive infrastructure commitments while maintaining its focus on responsible AI development.