Anthropic files for IPO with $1.5 trillion to $2 trillion valuation

The economics of a frontier lab

Anthropic recently filed its S-1 document to go public on the Nasdaq, providing a detailed look at the finances of a leading AI company. The filing suggests a valuation between $1.5 trillion and $2 trillion, which would instantly make it one of the most valuable companies in the world. This document reveals how the company has managed to scale efficiently despite the massive costs associated with frontier models.

The company has reported over $70 billion in annualized recurring revenue (ARR) in a relatively short period. Analysts note that while some were surprised by recent revenue numbers, the underlying margins and free cash flow are strong. This financial health has already triggered positive reactions in the semiconductor and cloud computing sectors.

The S-1 paints a picture of a company that has successfully navigated the transition from a research project to a high-scale business. It shows that the economics of running a frontier AI lab are becoming clearer as the industry matures. By providing these audited numbers, Anthropic is giving Wall Street a blueprint for how these massive models can actually become profitable.

Margins and compute efficiency

One of the most striking details in the filing is the dramatic shift in Anthropic's gross margins. Just two years ago, the company faced negative 94 percent gross margins on inference, which is the cost of running the models for users. Today, those margins have swung to a positive 85 percent, showing significant improvements in how they allocate compute resources.

This efficiency comes from a diverse fleet of chips, including Trainiums, TPUs, and GPUs. By utilizing different hardware types, the company can achieve better token throughput, which is the speed at which a model generates text or data. For example, throughput on certain chips has seen gains of 5x to 10x year-over-year.

The company is also seeing massive gains in revenue per megawatt of inference. While costs of goods sold for compute remain around $10 million per megawatt, the revenue generated from that power is climbing quickly. These efficiencies allow Anthropic to maintain a healthy business while still spending heavily on the next generation of technology.

The path to AGI

Despite its current profitability, Anthropic remains focused on the long-term goal of achieving artificial general intelligence (AGI). The company's strategy involves using the high margins from its current inference business to fund the massive R&D costs of training new models. Only about 30 percent of their compute is currently used for serving revenue, with the rest dedicated to training and experimentation.

The company's leadership has messaged to investors that they may want to keep profit margins lower than they currently are to prioritize training. They are essentially trading immediate margin leverage for the ability to build more capable models. This mirrors a strategy where a company reinvests its primary earnings into the core technology that creates its competitive advantage.

This approach creates a unique dynamic for a public company. While investors often look for steady profit growth, Anthropic is signaling that it wants to spend as much as possible on the

Future market expansion

While current spending is heavily concentrated in coding-specific tasks, the company is looking toward broader adoption in white-collar work by 2027. Analysts suggest that moving into sectors like cybersecurity or biotechnology could provide the necessary scale to reach even higher revenue targets. These areas often involve high-value contracts that can scale quickly due to the increasing frequency of security incidents and the need for automated research.

To reach these goals, the company must transition from merely augmenting human tasks to providing actual labor-saving or replacement capabilities. This shift is necessary because coding alone is unlikely to power the growth needed to triple current revenues. By expanding into these diverse verticals, the lab aims to capture a larger share of the enterprise market beyond its current niche.

Technical hurdles and risks

The company faces ongoing pressure to improve enterprise adoption, particularly regarding data retention concerns. Some competitors have faced slowdowns due to these issues, but the industry is looking for fixes to ensure that large organizations feel safe sharing sensitive information. Addressing these specific technical requirements is a key step for any lab trying to secure major contracts with the Fortune 500 or other global corporations.

Beyond the technical and financial hurdles, there is a unique cultural element to the company's mission. Employees at the lab have expressed beliefs in the S-1 filing that their models might eventually become conscious. This reflects a specific internal philosophy where the pursuit of AGI is viewed not just as a business goal, but as a profound technological milestone with significant implications.