The Carbon Cost of the Internet?�Understanding data center trends, estimates, and opportunities for action

Eric Masanet, Ph.D.

emasanet@ucsb.edu

​

Mellichamp Chair in Sustainability Science for Emerging Technologies

Head, Industrial Sustainability Analysis Laboratory

University of California, Santa Barbara

​

​

Massachusetts Climate Action Network, April 13^th, 2026

http://industrial-sustainability.org/

Office/Department/Division Name

Data centers: the backbone of the internet

https://www.institut.veolia.org/sites/g/files/dvc2551/files/document/2026/02/AI%20for%20Energy%2C%20Water%2C%20and%20Waste%20Management.pdf

Office/Department/Division Name

Data centers: the backbone of the internet

https://www.institut.veolia.org/sites/g/files/dvc2551/files/document/2026/02/AI%20for%20Energy%2C%20Water%2C%20and%20Waste%20Management.pdf

Office/Department/Division Name

Data centers: the backbone of the internet

https://www.institut.veolia.org/sites/g/files/dvc2551/files/document/2026/02/AI%20for%20Energy%2C%20Water%2C%20and%20Waste%20Management.pdf

https://www.nytimes.com/2024/07/11/climate/artificial-intelligence-energy-usage.html

https://www.nytimes.com/2026/01/27/technology/microsoft-water-ai-data-centers.html

https://www.politico.com/news/2026/03/11/data-centers-ai-electricity-virginia-00815219

https://news.vcu.edu/article/northern-virginia-data-center-air-pollution-rivals-power-plant-emissions

Office/Department/Division Name

Data center energy analysis: a day in the life

• How much energy do data centers and AI currently use?

​

Key questions for the energy systems analyst:

Office/Department/Division Name

Data center energy analysis: a day in the life

• How much energy do data centers and AI currently use?

​

• What is the trajectory of future energy demand and what are the implications for emissions and the grid?

​

Key questions for the energy systems analyst:

Office/Department/Division Name

Data center energy analysis: a day in the life

• How much energy do data centers and AI currently use?

​

• What is the trajectory of future energy demand and what are the implications for emissions and the grid?

​

• How might technology, policy, and behavior “bend the curve?”

Key questions for the energy systems analyst:

Office/Department/Division Name

Data center energy analysis: a day in the life

• How much energy do data centers and AI currently use?

​

• What is the trajectory of future energy demand and what are the implications for emissions and the grid?

​

• How might technology, policy, and behavior “bend the curve?”

Key questions for the energy systems analyst:

Data requirements, model

complexity, and uncertainty

Office/Department/Division Name

Data center energy analysis: a day in the life

• How much energy do data centers and AI currently use?

​

• What is the trajectory of future energy demand and what are the implications for emissions and the grid?

​

• How might technology, policy, and behavior “bend the curve?”

Key questions for the energy systems analyst:

Data requirements, model

complexity, and uncertainty

Office/Department/Division Name

The state of data center reporting

Office/Department/Division Name

Challenges with estimation: Part 1

https://bsky.app/profile/mims.bsky.social/post/3m5qx4nbxac2z

Office/Department/Division Name

Challenges with estimation: Part 2

Facebook leased data center (Ashburn, VA)

70+ MW (~50,000 homes)

https://sabeydatacenters.com/blog/hyperscale-data-centers/

A large data center before the “AI Boom”

Office/Department/Division Name

Challenges with estimation: Part 2

Facebook leased data center (Ashburn, VA)

70+ MW (~50,000 homes)

https://sabeydatacenters.com/blog/hyperscale-data-centers/

A large data center before the “AI Boom”

A large data center after the AI Boom (1 GW = 1000 MW)

Office/Department/Division Name

Challenges with estimation: Part 3

Office/Department/Division Name

Challenges with estimation: Part 3

Office/Department/Division Name

Challenges with estimation: Part 4

Office/Department/Division Name

Challenges with estimation: Part 5

Office/Department/Division Name

Challenges with estimation: Part 6

https://developer.nvidia.com/blog/scaling-token-factory-revenue-and-ai-efficiency-by-maximizing-performance-per-watt/

Office/Department/Division Name

Where is U.S. data center energy use headed?

Office/Department/Division Name

What about carbon emissions?

https://iea.blob.core.windows.net/assets/de9dea13-b07d-42c5-a398-d1b3ae17d866/EnergyandAI.pdf

Office/Department/Division Name

U.S. data center water consumption scenarios

How much is 300 billion liters? Equivalent to ~ 1 day of U.S. crop irrigation.

​

However, location matters!

Office/Department/Division Name

In fact, location matters a LOT!

Office/Department/Division Name

What is my personal footprint?

Source:

Office/Department/Division Name

What is my personal footprint?

Source:

What about the carbon, energy, and/or water footprint of:

​

• A streaming video?

​

• An email?

​

• A search query?

​

• An AI model query?

Office/Department/Division Name

Inherent variance in “per activity” estimates

https://carbonliteracy.com/the-carbon-cost-of-an-email/

https://www.thetimes.com/uk/technology-uk/article/thirsty-chatgpt-uses-four-times-more-water-than-previously-thought-bc0pqswdr

Office/Department/Division Name

Understanding the variance in estimates

Greenhouse gas (GHG) emissions per passenger kilometer traveled by model

Source: Chester and Horvath (2009). https://iopscience.iop.org/article/10.1088/1748-9326/4/2/024008/pdf

Office/Department/Division Name

Understanding the variance in estimates

Greenhouse gas (GHG) emissions per passenger kilometer traveled by model

Source: Chester and Horvath (2009). https://iopscience.iop.org/article/10.1088/1748-9326/4/2/024008/pdf

“Per activity” metrics:

Total impact (kWh electricity, kg CO2 emissions, etc.)

Total activity (queries, passenger-km, etc.)

Office/Department/Division Name

Understanding the variance in estimates

Greenhouse gas (GHG) emissions per passenger kilometer traveled by model

Source: Chester and Horvath (2009). https://iopscience.iop.org/article/10.1088/1748-9326/4/2/024008/pdf

Office/Department/Division Name

Understanding the variance in estimates

Greenhouse gas (GHG) emissions per passenger kilometer traveled by model

Source: Chester and Horvath (2009). https://iopscience.iop.org/article/10.1088/1748-9326/4/2/024008/pdf

Key variables:

• What is the bus capacity?
• How many passengers?
• How fuel efficient is the bus?
• What is the fuel?
• Which life-cycle stages are included?

Office/Department/Division Name

https://huggingface.co/spaces/AIEnergyScore/Leaderboard

Office/Department/Division Name

Bending the curve: Key conditions

1. Smaller models for specific problems
2. Data center siting 🡪 maximize renewables, minimize water stress
3. Stringent efficiency standards (PUE, WUE, and more)
4. Maximize waste heat recovery
5. Utility resilience 🡪 onsite storage and demand flexibility
6. Maximum operator reporting transparency
7. Measurement and verification of AI application impacts
8. Consumers choosing wisely!

Office/Department/Division Name

Community and policy action

Office/Department/Division Name