Source: The Verge
Anthropic's regulatory disclosures reveal that training and running large language models requires significant natural gas consumption—a climate cost largely invisible to users and investors betting on AI's clean future. The company's public filing suggests the industry can no longer hide the energy-to-capability tradeoff. Regulators and capital markets now face pressure to price sustainability risk into AI valuations. Methane's 28-year global warming potential is roughly 84 times higher than CO2, making AI's infrastructure carbon footprint materially different from the software sector's traditional story.