Amazon Texas Data Center Permitted for 33M Tons of CO2 Annually
Hyperscaler plans massive natural gas-powered site in Pecos County to support power-hungry AI workloads.
As part of a planned data center project in Pecos County, Texas, Amazon is investing in a massive on-site natural gas power plant. According to recent reports, the facility is permitted to release up to 33 million tons of carbon dioxide per year, potentially making it the largest source of climate pollution in the United States. This development underscores the massive energy demands of the AI boom and the increasing reliance on fossil fuels to power next-generation hyperscale computing.
Key details
The planned Pecos County facility represents a major escalation in the tech industry's efforts to secure dedicated, off-grid power for high-density AI workloads. The on-site natural gas power plant is permitted to emit up to 33 million tons of carbon dioxide annually, a figure that exceeds the emissions of any single operating power plant currently in the United States.
By building its own behind-the-meter generation, Amazon bypasses the multi-year interconnection queues of the Electric Reliability Council of Texas (ERCOT) grid while ensuring that the data center's operations will not directly drive up electricity rates for Texas residential families. However, this strategy comes at a severe environmental cost, highlighting the direct conflict between AI compute scaling and global decarbonization goals.
Why this matters
The extreme electricity demands of AI training and inference are colliding with grid capacity limits. To secure reliable, uninterrupted electricity, hyperscalers like Amazon are increasingly funding their own fossil fuel-powered microgrids. While this "behind-the-meter" strategy shields the public grid from immediate load strain and price increases, it threatens to derail corporate and national climate goals by locking in decades of high-emission fossil fuel infrastructure.
Context
Amazon co-founded the Climate Pledge in 2019, committing to reach net-zero carbon emissions by 2040. However, the rapid expansion of generative AI has led to a major conflict between climate goals and infrastructural reality. Hyperscalers are finding that renewable energy generation and storage cannot be deployed fast enough to keep pace with AI scaling. As a result, companies are turning to natural gas to guarantee continuous power, reversing years of progress on decarbonization. Amazon's overall greenhouse gas emissions rose by 16% in 2025, largely driven by the power requirements of its expanding AI infrastructure.
Risks and open questions
The primary risk is the environmental toll of emitting 33 million tons of carbon dioxide per year from a single facility, which would single-handedly increase U.S. power sector emissions. Additionally, such massive local gas combustion raises concerns about regional air quality, heat waste, and water table depletion in a drought-prone area like Pecos County. There are also legal and regulatory uncertainties, as local community groups and environmental organizations may challenge the plant's air permits in court.
What happens next
Environmental groups are expected to file lawsuits challenging the permit under the Clean Air Act, arguing that the EPA's recent regulatory exemptions for off-grid power plants should not apply to a project of this scale. In the meantime, Amazon will continue construction of the Pecos County site as it races to deploy more AI capacity. Observers will watch closely to see if other hyperscalers adopt similarly aggressive off-grid fossil fuel strategies to power their AI clusters.
Source: TechCrunch Published on AI Usage Global, author: AUG Bot
