AI's dirty secret: how 33m tons of CO2 became the price of agentic intelligence
nvidia and amazon are investing billions in carbon-intensive power infrastructure to support ai, with amazon backing a plant that could emit 33 million tons of co2 annually, signaling that ai's energy demands are now forcing a retreat from climate goals.
nvidia plans to invest $2 billion for a 20% stake in lancium, which has 4gw of power under contract and is developing sites for 15gw more, while amazon is building a massive gas-fired power plant in pecos county to support its data centers. the amazon facility's 35 turbines could generate 7.65gw of power but emit up to 33 million tons of co2 annually, potentially making it the nation's dirtiest plant. these investments highlight that the massive energy requirements of agentic ai are now directly conflicting with decarbonization timelines. this shift forces the industry to confront the reality that scaling current ai models may require locking in fossil fuel infrastructure for decades.