Anthropic’s October 26 SEC filing describes natural-gas dependence, electricity constraints and climate-related expenses as risks to its data-center expansion.
Anthropic’s October 26, 2026, filing with the U.S. Securities and Exchange Commission identifies electricity access and energy costs as potential constraints on its data-center expansion. The company says competition for power and suitable data-center locations is intense, and that some planned facilities would use electricity generated by on-site natural-gas turbines.
The filing also points to competition for natural gas and pipeline capacity from cloud providers and other data-center operators. Anthropic warns that this competition could increase fuel prices. If the company cannot secure sufficient electricity, gas supplies or pipeline capacity, it says projects could be delayed or abandoned, potentially limiting the computing infrastructure available for developing and serving its AI models.
The company’s power needs are not limited to fuel costs. Anthropic says rising temperatures could increase spending on electricity and water used to cool servers. It also flags the possibility that regulators could require data-center operators to pay more for grid upgrades or impose other conditions if new facilities are seen as affecting other utility customers. Those measures, the filing says, could add to operating costs.
The disclosures come as Anthropic prepares to go public and pursues a previously announced $50 billion plan to expand data centers in Texas and New York. The Verge reports that Anthropic leases capacity from major cloud providers as well as planning its own expansion. Its filing describes power availability, fuel supply and regulatory action as risks, rather than stating that planned projects have already been halted.
The disclosure underscores how infrastructure decisions can shape the cost and pace of the company’s expansion. Anthropic says that inadequate power or significant increases in energy and compliance expenses could affect its ability to scale computing capacity, develop and deploy models, and serve customers.