AI is short for Artificial Intelligence. AI uses algorithms, or a set of rules/instructions, to analyze and process data with the end goal of identifying patterns and being able to make decisions. AI requires the “feeding” of lots of data in order to replicate the processes of human intelligence.
Data centers are buildings which house rows of computers and equipment, which store, process, and transmit massive amounts of data. Certain data centers, called AI data centers are built specifically for the purpose of training, deploying, and delivering AI services.
Since AI and the need for lots of data is driving much of the growth we’re seeing in the number, complexity, and size of data centers, the relationship between the two is becoming increasingly intertwined. Data centers provide a feedsource for AI to be trained and operated and in return, AI optimizes data center storage, processing, and even security.
Goldman Sachs Research estimates that data center power demand will grow 165% by 2030, driven fiercely by the “AI revolution.”
AI/Data centers x energy consumption
Like most things, data centers need energy to run. And as data centers get larger and serve the needs of AI, their energy demands increase. Data centers currently only make up 4% of US electricity demand, but that’s forecasted to triple by 2028. This means data centers will need to rely heavily on the U.S. grid, composed of mostly (60%) fossil fuels, in order for them to operate.
It also means data centers are increasingly interested in accessing more energy to ensure there’s enough and reliable energy for their operations. In recent years there has been a notable uptick in tech companies and fossil fuel corporations announcing plans for opening energy production facilities to serve the energy demand needed for data centers. Some recent developments include:
- Chevron is partnering with Engine No. 1 and GE Vernova to create multi gigawatt-scale co-located power plant and data center
- In January U.S. nuclear power provider Constellation Energy bought Calpine Corp in one of the biggest power industry acquisitions, driven in part by AI data centers
- Big tech companies have plans for building or recommissioning nuclear power plants (like Three Mile Island) to supply data centers with power
The International Energy Agency, an intergovernmental agency specializing in the global energy sector, notes that one large data center uses the same amount of energy as 400,000 electric vehicles.
Researchers at UMass Amherst, who looked into the AI training process for teaching machines human language (think ChatGPT), found that training an AI model can emit nearly five times the lifetime emissions of the average American car. Asking ChatGPT for an answer can use up to 10x as much electricity as a regular google search.
AI/Data centers x ratepayer costs
As mega tech companies and fossil fuel corporations build or connect to energy production facilities like nuclear and natural gas power plants to power data centers, ratepayers have both pointed out and voiced concerns over projected increases in utility costs. Utilities bake the cost of construction for new power plants and other infrastructure into utility bills under the premise that society as a whole benefits from that increase in power. But in a society where data centers are increasingly requiring existing and new power solely for their operations while households still face barriers to accessing the internet and experience energy insecurity (often both at the same time), just who is benefiting and at what cost? In a recent report, Harvard Electricity Institute accurately points out that “serving data centers is a lucrative opportunity that is incentivizing utilities to offer attractive rates to Big Tech companies.”
AI/Data centers x water consumption
Have you ever had one too many things running on your computer (like a seemingly infinite number of tabs, or maybe Sims with one too many mods…) and it starts to sound like an airplane taking off? That’s because computers require a lot of power to run and with that power comes heat. Heat requires the equipment to be cooled down so it can continue operating. For your laptop or computer at home, you have a fan (that’s what makes the noise by the way).
Data centers are no different. As computers in data centers complete super-fast processing, things get hot. Data centers cool the massive computers down in order to keep things running smoothly and processing times low.
There are a lot of ways data centers maintain temperatures inside buildings, from air conditioning to liquid coolant, but many cooling systems require water in some way. The use of water for data center cooling often requires pulling water from nearby sources- depleting local water sources. NPR reports that a mid-sized data center uses as many gallons of water as 1,000 households in a single day. To add on to the sheer amount of water data centers use, the majority of data centers are located in drought-prone areas like Dallas and Los Angeles. Researchers at Lawrence Berkeley National Laboratory found that 20% of data centers in the United States already rely on watersheds that are under moderate to high stress from drought and other factors, putting at risk local communities, local waterways and wildlife.
Ironically, publicly available data for data center water usage is limited and there’s often lack of monitoring on site. In some places, like The Dalles where local news organizations sued over access to Google’s water consumption information, they found that Google’s data centers used 355 million gallons of the region’s water in 2022, 29% of the city’s total water consumption.
AI/Data centers x perpetuating other systems of harm (land grabs, militarization)
AI can be used to perpetuate systems of harm like displacement. Data centers require land, which is increasingly limited because of competition with other uses. In places like North Carolina where large AI/data centers are located, they drive up the cost of housing/land and threaten to encroach on important and meaningful places like state parks.
There’s also a deeply intertwined relationship between technology, the use of AI and militarization through things like surveillance and impacting or limiting peoples’ rights to information.
AI/Data centers x Black Communities x Climate/Energy/Economic Justice
Black communities are hit three-fold by AI and data centers.
- Black communities face the highest impact from AI and data centers because they’re disproportionately impacted by climate change, which is driven by CO2 emitted in part by the energy production involved in running data centers.
- In areas where data centers and energy production risk raising utility prices, Black communities are already more likely to be energy insecure and have difficulty paying for their utility bills.
- In thinking about AI and its applicability in the world, roles to be replaced by automation and AI roles – such as food service workers – are held disproportionately by Black people. This jeopardizes Black community’s economic stability and threatens to perpetuate the already existing gap seen in Black household wealth.
What can you do?
- Know about the data centers by you: https://www.datacentermap.com/
- Limit your AI usage where you can, and use it responsibly. Turn off Apple Intelligence and other unnecessary forms of AI. Use web browsers that do not use AI (DuckDuckGo)
- Support and engage in call-to-actions by organizations organizing against data centers that threaten to harm already overburdened communities, like Memphis Community Against Pollution’s petition to deny air permits to a facility that pollutes the community and perpetuates environmental racism.
Demand AI development regulation by federal and state governments and tap into resources from organizations and coalitions demanding accountability and responsibility for AI and its impacts. Check out Green Screen Coalition’s joint statement on AI, signed by over 130 organizations.