Welcome! You're about to become an investigator, uncovering the hidden environmental costs behind the AI you use every day. This worksheet will guide you through your own research to understand the energy and water needed to power artificial intelligence.
Your Task: Use the internet and your critical thinking skills to answer the questions in each section.
Step 1.1: What is a Data Center?
In your own words, what is a data center and what is its role in powering AI? Hint
Step 1.2: Inside the Machine
What are the three main functions of a data center? (Hint: Think compute, storage, and networking). Briefly describe each and explain why cooling is so essential.
Step 2.1: Find the Data
According to the IEA, what percentage of global electricity did data centers and data transmission networks account for in 2022?
Step 2.2: The Carbon Connection
The environmental impact of electricity depends on how it's made. In your own words, explain the difference in carbon footprint between a data center powered by fossil fuels and one powered by renewable energy.
Step 2.3: Critical Thinking
Based on the IEA report and what you know about technology today, why do you think the energy demand for AI is projected to grow so much in the coming years?
Step 3.1: Why Water?
Explain the process of evaporative cooling. Why is this method, which is common in data centers, so water-intensive?
Step 3.2: Dig into a Report
Visit Google's 2023 Environmental Report: Google 2023 Environmental Report.
How many billions of gallons of water did Google consume in 2022? (Hint: Check the tables under "Environmental data tables.")
Step 3.3: Local Impact
Imagine a large data center that uses billions of gallons of water is built in an area that often experiences droughts, like California or Arizona. What potential problems or conflicts could arise for the local community and environment?
Step 4.1: Research Greener AI
Using your own research skills, find and describe three strategies that tech companies are using to reduce the environmental impact of their data centers and how AI can be part of the solution.
Use the calculator below to estimate the energy, carbon footprint, and water of using a generative AI model. The total energy includes the energy for your queries (inference) plus a tiny fraction of the energy used to train the model in the first place (amortized training).
The one-time training cost is "paid off" over its life. This estimates how many billions of queries the model will serve.
Find live data from Electricity Maps.
This varies based on the data center's cooling method. Evaporative cooling uses more water.
Total Electricity Used:
0.043 kWh
Equivalent to running a 10W LED bulb for 4.3 hours.
Total Carbon Emitted:
20.35 g CO2
Equivalent to driving an average car 0.05 miles.
Total Water Used:
0.03 Liters
The impact of a few queries might seem small. But when you consider that millions of people use AI daily, how does this change your perspective on its overall environmental footprint? Does this exercise make you think differently about how you use AI?