Amazon's new AI data center power plant will become the largest source of CO2 emissions in the US

The first artificial intelligence center in the Amazon, located in Belém, Paraná state, has become the subject of an investigation by the Paraná State Prosecutor's Office. The investigation examines the risk of 'heat islands' formation and the lack of public consultations with coastal communities and residents of areas neighboring this facility.

The project, named BEL1, involves investments of 250 million Brazilian reais and is scheduled to begin operations in 2027. Its initial capacity will be 7.5 megawatts, enough to power over 22,500 households.

A study conducted by the University of Cambridge indicates that data centers are capable of raising the average land surface temperature by 2°C, with peaks reaching 9.1°C. This scenario is the basis of the concern raised by the Paraná State Prosecutor's Office.

Data center structures dissipate energy as heat, which could potentially raise the external temperature in an area that already frequently reaches thermal stress levels of 34°C or higher, thereby creating so-called 'heat islands' in the Amazon region.

Elea Data Centers, the company responsible for implementing the project, states that it will use air conditioning systems to cool the equipment instead of water. Furthermore, the company explains that more in-depth studies on thermal impact will only be conducted for the 100 MW phase, which does not yet have a defined implementation date.

Amazon is investing in the construction of a natural gas power plant to power its massive artificial intelligence data center in Texas. According to The New York Times, this facility could potentially become the largest individual source of climate pollution among power plants in the United States.

The project, located in Pecos County, Texas, illustrates the challenge of AI expansion: this technology requires enormous amounts of energy, while Amazon aims to achieve a goal of actual greenhouse gas emission elimination by 2040.

The future power plant will be equipped with 35 natural gas turbines capable of generating up to 7.65 gigawatts of energy, which will be directed to the planned Amazon data center.

Potential Emissions

Licenses allow this facility to emit about 33 million tons of carbon dioxide annually. This permitted volume is more than double the annual emissions of the James H. Miller Jr. coal plant in Alabama, which emits approximately 16 million tons.

However, this does not mean the plant will necessarily reach this limit, as such facilities typically do not operate at the maximum capacity allowed by licenses.

AI data centers consume vast amounts of electricity to perform the trillions of calculations necessary for creating and running these systems. As AI expansion puts a strain on electrical grids, companies are seeking alternatives to avoid long waits for grid connections from suppliers.

One such solution is building their own power plants directly on the data center premises. Natural gas is being considered as a relatively fast and simple option to implement.

Expert Michael Thomas, founder of Cleanview, told The New York Times about the rise of off-grid gas plant projects in Texas and other regions.

Amazon's emissions have increased in recent years, and the company itself has acknowledged that the growth of AI data centers could complicate its environmental goals. Margaret Callahan, an Amazon representative, noted that the current situation differs from when the company helped create the Climate Pledge, but confirmed its commitment.

The company is also exploring the use of solar energy and batteries in the data center and stated that the project will create thousands of jobs.

Ecological Concerns

Environmental groups are concerned about the potential impact of the project on the climate and air quality. In addition to carbon dioxide, gas power plants release pollutants that contribute to smog formation and may be linked to issues such as asthma and heart disease.

This case demonstrates the scale of the energy challenge posed by the AI race: how to expand the necessary infrastructure for these systems without proportionally increasing the environmental impact.