This, my lazy Brave AI generated info, summarizes well what I've garnered about the subject:
Data centers in the US are increasingly building their own "behind-the-meter" power plants to bypass grid interconnection delays that can stretch 7 to 10 years. A Cleanview report identified 46 data centers with a combined capacity of 56 GW planning on-site generation, representing roughly 30% of all planned US data center capacity. 90% of these projects were announced in 2025 alone, driven by the urgent need to power AI infrastructure where traditional utility connections are too slow.
Natural gas is the dominant fuel source for these private plants, with approximately 75% of identified equipment being gas-powered. Developers are prioritizing speed over efficiency, utilizing mobile generators, aeroderivative turbines, and reciprocating engines rather than waiting for long-lead-time heavy-duty gas turbines. This shift has triggered a boom in private natural gas proposals, which tripled in 2025 according to the Global Energy Monitor, with some projects emitting millions of tons of greenhouse gases annually.
Political and corporate responses are accelerating this trend. President Donald Trump issued a mandate during his State of the Union address for major tech companies to build ttheir own power supply to protect ratepayers from higher bills. Companies like Google, OpenAI, Oracle, and xAI are deploying on-site gas plants or fuel cells, with Google’s Texas campus permitting a facility that would emit over 4.5 million tons of greenhouse gases per year. This "shadow power grid" represents a significant deviation from previous clean energy commitments, as speed to market becomes the primary driver for AI infrastructure investment.
This, my lazy Brave AI generated info, summarizes well what I've garnered about the subject: Data centers in the US are increasingly building their own "behind-the-meter" power plants to bypass grid interconnection delays that can stretch 7 to 10 years. A Cleanview report identified 46 data centers with a combined capacity of 56 GW planning on-site generation, representing roughly 30% of all planned US data center capacity. 90% of these projects were announced in 2025 alone, driven by the urgent need to power AI infrastructure where traditional utility connections are too slow.
Natural gas is the dominant fuel source for these private plants, with approximately 75% of identified equipment being gas-powered. Developers are prioritizing speed over efficiency, utilizing mobile generators, aeroderivative turbines, and reciprocating engines rather than waiting for long-lead-time heavy-duty gas turbines. This shift has triggered a boom in private natural gas proposals, which tripled in 2025 according to the Global Energy Monitor, with some projects emitting millions of tons of greenhouse gases annually.
Political and corporate responses are accelerating this trend. President Donald Trump issued a mandate during his State of the Union address for major tech companies to build ttheir own power supply to protect ratepayers from higher bills. Companies like Google, OpenAI, Oracle, and xAI are deploying on-site gas plants or fuel cells, with Google’s Texas campus permitting a facility that would emit over 4.5 million tons of greenhouse gases per year. This "shadow power grid" represents a significant deviation from previous clean energy commitments, as speed to market becomes the primary driver for AI infrastructure investment.