Before Its IPO, Anthropic Exposes AI Industry’s Energy Vulnerabilities in Its Prospectus

Image credit: Cath Virginia / The Verge

Anthropic, which is advancing towards an initial public offering (IPO), submitted its S-1 prospectus to the U.S. Securities and Exchange Commission, bringing the energy dependence of the generative AI industry into the open.

An S-1 is a statutory registration document required for U.S. companies before going public, where all material risks that may affect operations must be disclosed truthfully and the statements are legally binding. The document was released on Monday local time, and Anthropic explicitly noted in it that "The location of data centers with sufficient power resources and power capacity is currently highly competitive."

The document also confirms that AI data centers deployed by Anthropic have an increasing dependence on natural gas. "Natural gas" is a marketing name used in the fossil energy industry; the core component of this fuel is methane, as traced in Vox's public energy coverage.

The methane-based energy structure directly brings two layers of risks to Anthropic's computing power expansion. First, rising natural gas prices will directly increase the operating costs of data centers. Second, the capacity constraints of the existing energy supply chain may delay the progress of new data centers.

Anthropic also mentioned in the document that the company faces a dilemma similar to "Catch-22", a risk directly related to climate change, and the complete statement of this part has not been disclosed in currently published report excerpts.

Previously, when various AI companies communicated their energy-related plans externally, they mostly focused on purchasing green electricity and carbon neutrality goals for publicity. This legally liable listing document, however, provides information closer to the real operational status of the industry.

For the full report, please see the original article on The Verge.