Current Events: The Environmental Costs of AI

Generative AI is a monumental innovation with far reaching effects.

One of the most troubling impacts of Generative AI technology is the environmental demand for water and electricity at the data centers where all the information is stored. A data center houses rows of servers, all storing and distributing data.

Oregon has among the largest data center markets in the world, home to Amazon, Apple, Google, Microsoft and Meta facilities. Why would a state want data centers? Construction jobs and long term tax revenue. Oregon’s legislature passed SB 611 in 2015 exempting data centers from certain taxes to attract these businesses to the state. Purported as economic drivers, some counties have not seen the anticipated yields. In Crook County, Oregon, the schools missed out on $10 million in local revenue to property tax exemptions to corporations, with well over $20 million of reported lost revenue.

AI data centers also need billions of gallons of water to cool computer hardware. Instead of building data centers where there is ample water, a Bloomberg News analysis reports “two-thirds of new data centers built or in development since 2022 are in places already gripped by high levels of water stress.” Much of that is in the western states of the US, already experiencing drought.

Once facilities are built, they don’t usually support many permanent jobs. But the environmental impacts continue. In Umatilla, existing nitrate pollution of water has become even more contaminated as the water moves through data centers, partially evaporates, then returns to wastewater systems. The contamination was sometimes 8 times over the Oregon limit for water safety, causing Gov. Brown to issue an emergency declaration and requiring households to have water trucked in.

Advances in cooling can improve the situation. For example, a closed-loop system would use less water and waterless cooling systems show even more promise in that sense. But to save water, these systems use more electricity. Already, data centers have increased our state’s electricity consumption up 20% between 2013 and 2023. For comparison, a 30-megawatt data center will use as much electricity as the city of Ashland while a 250 megawatts data center would be the equivalent to the city of Eugene’s needs. The Amazon data centers in eastern Oregon increased the electrical load by 554% at the Umatilla Electric Cooperative over those years.

Meanwhile, households in Oregon watched their electricity bills rise at an alarming rate, in large part due to the demands of these data centers. To combat this effect, this year’s Legislative Session approved the POWER Act, which creates a customer category for data centers so utilities bill these mega users appropriately for their energy usage and separate rates for residential customers.

Throughout the country, these energy demands may be met through electrical generation from coal or natural gas – creating more carbon dioxide pollution and driving climate change. In other regions, transitioning to renewable energy can alleviate some of the environmental problems of AI’s energy consumption.

This week’s current event explores the environmental costs of the data centers Generative AI relies on and possible options for mitigating these effects.

Essential Questions:

Essential Question #1 – What are the economic benefits of data centers to a local community?

Essential Question #2 – How do data centers affect the local environment?

Essential Question #3 – In what ways does artificial intelligence technology affect data centers?

Essential Question #4 – What factors could reduce the environmental impacts of data centers?

Vocabulary:

data center: building that houses computer hardware to store and process digital information

Artificial Intelligence: computer systems that learn to carry out tasks that traditionally required human intelligence, like reasoning and decision-making

tax abatement: a local agreement between a taxpayer and a local taxing unit that exempts all or part of the increase in the value of property from taxation.

Videos: