亚马逊在德州收购了一座大规模离网发电厂，用于为其资料中心供电。该电厂获准每年排放高达三千三百万吨温室气体，成为美国最大的温室气体排放源之一。该电厂完全依赖简单循环涡轮机，这种涡轮机的效率远低于一般电网电厂所使用的复合循环机组，因为后者能回收废热来产生更多电力。即使实际排放量仅为许可上限的一半，其温室气体排放仍将超过七十八座平均规模天然气电厂的总和。（关键数字：33, 78）
使用简单循环涡轮机正在成为资料中心电力开发商日益普遍的做法。Meta在俄亥俄州、Crusoe为微软和Google在德州建设的电厂，以及xAI在孟菲斯的Colossus资料中心，都采用了大量简单循环涡轮机。业界之所以倾向于这种技术，是因为它安装速度更快、成本更低，且能快速启停以应对人工智慧训练所需的巨大能量波动，同时也不需要用水，避免了社区对水资源的争议。（关键数字： a years）
这些离网电厂的快速建设引发了对「平行化石燃料电网」形成的担忧，这些电厂设计寿命长达数十年，将持续产生数百万吨温室气体排放。虽然亚马逊等公司承诺未来会将电厂接入电网，并声称仍致力于二零四零年实现净零排放，但业界专家指出，若人工智慧泡沫破裂或科技公司的能源需求下降，这些庞大而低效的天然气电厂可能成为「搁浅资产」。
Amazon has acquired a massive off-grid power plant under construction in Texas for one of its data centers, permitted to emit up to 33 million tons of greenhouse gases annually. The facility relies entirely on simple-cycle gas turbines, which are significantly less efficient than the combined-cycle units used by conventional on-grid power plants because they waste heat rather than recapturing it for additional energy generation. Even at half the permitted emission levels, the plant would still produce more greenhouse gas pollution than 78 average-sized natural gas plants combined, according to the EPA.
The use of simple-cycle turbines is becoming widespread among data center power developers. Companies including Meta, Microsoft, Google, and xAI are all building or planning facilities powered by these less efficient turbines across Texas, Ohio, and Tennessee. The trend is driven by practical considerations: simple-cycle turbines are faster and cheaper to install amid a years-long backlog for turbine orders, they can rapidly start up and shut down to meet the intense and fluctuating energy demands of AI model training, and they do not require water cooling—sidestepping a growing concern for communities near data centers.
The rapid proliferation of islanded power plants using inefficient turbines raises serious environmental concerns, as many are designed to operate for decades and could collectively create a parallel fossil fuel grid generating millions of tons of emissions. While some companies, including Amazon and Williams, have indicated plans to eventually connect their plants to the public grid, the timeline remains uncertain. Industry experts warn that if the AI boom subsides or energy demands decline, these massive gas plants could become stranded assets with no viable purpose, representing both an environmental and financial liability.