# US Data Centers Could Become Global Natural Gas Giants by 2035
The explosion in artificial intelligence infrastructure spending threatens to reshape American energy consumption in ways policymakers barely anticipated. US data centers could consume more natural gas by 2035 than Germany and Japan combined, according to recent energy forecasts.
This projection stems directly from the infrastructure buildout required to support generative AI models and large language models. Tech giants including OpenAI, Google, Meta, and Amazon are locked in an arms race to secure computing capacity. Each new generation of AI systems demands exponentially more power than the last. Data centers running these models operate continuously, creating baseline electricity demand that traditional renewable sources struggle to meet alone.
Natural gas has emerged as the bridge fuel for this transition. Unlike coal, it burns cleaner. Unlike purely renewable sources, it provides constant baseline power without the intermittency problems that plague solar and wind. Major cloud providers and AI infrastructure operators have begun signing long-term contracts with natural gas suppliers to guarantee supply.
The scale is staggering. Germany consumed roughly 80 billion cubic meters of natural gas in 2022. Japan consumed about 110 billion cubic meters annually during the same period. US data centers are currently responsible for about 2 percent of total electricity consumption. That figure is expected to double within a decade as AI workloads intensify.
Tech companies argue this energy demand is necessary to deliver AI capabilities the market demands. OpenAI, Anthropic, and other frontier AI labs contend that training models requires computational resources at unprecedented scale. Inference costs, the electricity consumed when AI models actually run in production, dwarf training costs. A single ChatGPT query uses measurable energy resources.
The natural gas trajectory creates several problems simultaneously. Environmental advocates warn that locking in natural gas infrastructure for decades to come entrenches fossil fuel dependence precisely when the grid needs to decarbonize. Power plant operators have less incentive to invest in renewable capacity if tech companies sign guaranteed offtake agreements for natural gas power.
Regional electricity markets face mounting pressure. Virginia, where Amazon and other hyperscalers concentrate data center investments, has already experienced grid strain. Texas, home to massive server farms, negotiates constantly with grid operators over peak demand periods.
Some data center operators explore alternative pathways. Microsoft has invested in small modular nuclear reactors. Google pushes harder on renewable procurement, though actual renewable percentages lag pledges. Most companies hedge by combining natural gas, renewables, and other sources.
The 2035 projection assumes current growth trajectories hold. If AI adoption slows or efficiency improvements accelerate, actual consumption could fall short. If AI deployment accelerates beyond current forecasts, consumption could exceed projections. Either way, the infrastructure decisions made in the next two years will lock in energy patterns for decades.
Policymakers have barely begun addressing the energy implications of AI at scale. Federal regulations fail to incorporate AI-driven electricity demand into long-term grid planning. State utility commissions approve data center projects with energy impact assessments that underestimate actual AI workload requirements.
The collision between AI infrastructure growth and climate commitments arrives sooner than most stakeholders prepared for.