I live in Northern Virginia (NOVA), which is known as “data center alley.” It started with UUNET, which was the so-called backbone of the internet. UUNET was founded in 1987—in a much simpler time. Since then, it has been part of MCI, and is now Verizon Business.
A quick search comes up with 15 of the purported hundreds of data centers in NOVA. They range in name originality from “Data Center” to “Iron Mountain Data Centers,” and “Evocative Virginia Data Center.” Now that I have done my search, I will likely be bombarded with data center information generated I’d imagine from the very same data centers.
In this day of attempting to reign in energy demand and related carbon dioxide (CO2) emissions, the growth in data centers presents a very real challenge to local politicians a well as our national climate goals. Perhaps no local government is more affected by the growth in data centers than Prince William County, Virginia.
Prince William County is on the periphery of NOVA. It has proximity to the technical expertise of the region, but also has the available land area lacking in the closer-in suburbs. It has a rich history—being home to, for example, Manassas Battlefield Park, an important Civil War site. It’s not the county’s past that is at stake here, but its future. The battle now is between the forces of economic development versus the need to develop the county in a sustainable way, not only in keeping with local considerations of congestion and pollution, but also global considerations in the form of greenhouse gas emissions.
The Prince William County Sustainability Commission is tasked with striking the balance between economic development and environmental integrity. The County’s previous goal was to reduce a projected 4.6 million metric tons of carbon dioxide equivalent (MMTCO2e) to 2.1 MMTCO2e in 2030. This, on its own was a daunting task given that this would be a reduction of over 50%. However, the actual task is much more monumental.
The original projection was based on an average commercial building CO2 estimate of kilowatthours per year per square foot (kWh/yr-sqft). Because of the much higher energy intensity of data centers this was a flawed assumption. This earlier estimate led to the notion that these data centers would contribute to 5% of the 4.6 MMTCO2e value, or .230 MMTCO2e in 2030. However, a revamped analysis using realistic values for both energy intensity and the growth in square footage of the data centers yields a mid-point value of 84% of 22MMTCO2e or 18.48 MMTCO2e. This is an increase of 80 times the original estimate. (1)
So, what can be done? First, we must recognize the reality of the situation. The updated estimate appears to be a very large number compared to the original estimate, however in the context of total U.S. emissions it is not that great considering the vital role that these data centers and the advanced systems such as artificial intelligent (AI) provide. The benefits to the economy and more efficient use of energy and related environmental benefits could be large. However, they are dispersed throughout the U.S. economy and indeed globally.
Some of the dominant companies such as Google and Microsoft have committed to purchasing zero-carbon electricity. The very size of the energy requirements makes it hard to produce that energy on site. Therefore, these centers will depend on bringing clean energy in from outside the area. However, this may be challenging given the disparity between the 1.5 to 2 years that it takes a data center to get approved versus 10 years for new transmission lines to be approved and built. This makes it essential that the permitting reform process continues. While the Federal Energy Regulatory Commission’s (FERC) recent Order 1920 (2) is intended to improve the planning and development process, there remain many issues that hold it up. For example, a permit can be sued for 6 years—just the permit itself and not the actual work being done. This introduces an uncertainty that makes attracting the needed capital very difficult.
The Electric Power Institute (EPRI) estimates that by 2030, as much as 9% of U.S. electricity could be consumed by data centers. (3) This is roughly a doubling from today’s percentage, but also a proportion of a higher number. Locally, in NOVA we may need to reconfigure our climate goals, but if AI is going to live up to its potential there will be beneficial results from these systems for the economy and environment as a whole.
Here are some possible outcomes in the energy and emissions space:
Smart traffic control systems that reduce transportation energy consumption
Efficient industrial systems that optimize all phases of production
Virtual power plants integrating residential and commercial buildings with the grid
An AI-produced roadmap for upgrading and expanding the existing electricity transmission grid
Planning, design, and implementation of ground source heat pumps at neighborhood scale
Renewable energy planning models for wind and solar that include all land-use requirements such as agriculture and forestry
Development of alternative battery materials and longer-term energy storage technologies
After all, what would today’s traffic patterns be like if online shopping and Zoom business meetings didn’t exist. Over time the benefits of these data centers will be better understood. I would propose that an important measure of AI is that it is “energy neutral” or better. In other words, the additional energy required by AI should at least be offset by its improvements to our energy economy. I include clean and renewable energy in this equation. Outside of data centers, there is a movement to electrify everything. If this is to have any chance of success there must be an efficiency component that helps to alleviate the dramatically increased electricity demand.
While many potential solutions are proposed by the Prince William County Sustainability Commission in terms of meeting its own ambitious climate goals, AI itself needs to prove that it can help tackle the pressing issues of our times of which climate change is front and center. The onus should not fall only on the physical location of the data centers and related energy demand, but also on the larger economy that reaps the rewards from AI’s potential benefits.
Special thanks to Randy Freed who provided the links to the Prince William Sustainability Commission material. All comments are my own and should not be attributed to the Commission.
Sources
1 Prince William County Sustainability Commission Meeting of April 25, 2024.
An excellent article, Perry. While many have written about the impact of more electricity needed for AI, you've pointed out well that we also need to consider the other energy uses it might displace. For me, Zoom and on-line shopping have dramatically cut down my auto use, where fewer VMT means fewer gallons of gas and fewer CO2 emissions.