Data Centers Are Booming
Will they be energy hogs or energy angels?
Generative AI models like ChatGPT are delivering artificial intelligence to the masses. Anyone with an internet connection can now use sophisticated deep learning tools to complete complex tasks in seconds – from writing term papers to creating weird blog post art.
These models are trained on petabytes (a word I just learned) of data to generate new content based on patterns and information they have learned to recognize and mimic. All of this learning and inference requires a lot more computing power than quaint Google searches of yesteryear. Alphabet’s John Hennessy estimates that my AI-supported art project required 10 times the computing costs of a standard internet image search.
What will happen to global energy use/GHG emissions as AI computing demand growth is compounded across 5.3 billion internet users? The range of opinions on this topic is pretty astounding. Some see climate destruction in the rise of AI. Others see a smoother clean energy transition. I can’t pretend to know where all this is going. But I can see complications ahead.
Data centers as energy hogs?
Nearly all the world’s internet traffic travels through data centers. These are large, increasingly “hyper-scale” buildings that house computing machines and related equipment. If we want to understand how increased computing demand will impact electricity demand, data centers seem like an important place to start.
The IEA estimates that data centers consumed 220-330 Terawatt hours in 2021. To put this into perspective, the entire state of California uses around 278 TWhs per year.
I’ve been curious to understand what rapidly expanding AI applications might imply for data center electricity demand. If you ask the internet, the estimates are all over the place: AI is on track to consume our entire energy supply; US data center demand will more than double by 2030; Data centers will draw up to 21% of the world’s electricity supply by 2030.
These projections are kind of alarming given all the other challenges we have on our decarbonization plate. However, it is important to keep in mind that analysts have, in the past, overestimated data center electricity use. In the 2000s, there were similar demand increases forecast. Big improvements in “power usage effectiveness” (PUE) meant that global data center energy use increased by only 6% between 2006-2018 while computing output and storage capacity increased by a factor of 6 and 25, respectively.
The chart above tracks the remarkable improvements in PUE prior to 2019 using survey data from the Uptime Institute. But it also shows that these efficiency gains have flatlined in recent years. If efficiency gains stagnate while computing demand explodes, we’ll have an energy hog problem on our hands.
Data centers = climate angels?
There is a more optimistic counter-narrative out there arguing that data centers are not an energy problem, but rather a critical component of green and innovative climate solutions.
Barry Fischer, Google’s “Data Storyteller”, has several podcasts devoted to this topic (see here for example). And if you ask ChatGPT about the climate impacts of data centers and generative AI, it will quickly generate a very long list of innovations that will “enhance the development, deployment, and optimization of clean energy technologies”.
As I understand it, the angelic vision is that data centers will procure their own additional 24×7 renewable energy so as to meet their electricity needs. Relatively flexible data center loads will provide value to the grid. And the innovative AI applications that data centers support will help optimize/accelerate the clean energy transition. I see the potential here. But I think we’ll need more than potential and private sector promises to keep data center energy consumption on the right track.
It’s getting complicated
Across the country, we’re already starting to see tensions between our decarbonization goals and a pressing need to accommodate growth in data center electricity demand. Some recent examples include:
- New data centers in Virginia are driving significant and unexpected electricity demand increases. To keep up, Virginia utilities are asking to extend the lifespan of gas- and coal-burning plants.
- Wisconsin is also looking into delaying coal plant retirements to accommodate a new Microsoft data center.
- In Maryland, a new data center applied to have 504 MW of backup diesel generators exempted from environmental regulations on the grounds of “public convenience and necessity.”
A few anecdotes do not a crisis make. But delayed coal plant retirements and angry ratepayers do underscore some challenges ahead. These challenges are greatly exacerbated by a dearth of energy data from data centers. A lack of reliable information makes it really hard to anticipate and plan for data center energy demand, let alone monitor and manage their environmental impacts.
Directing responsible data center growth requires data
Mis-information about data center energy use is feeding exaggerated claims about where data center energy use is trending and what we should do about it. This is a problem we can solve. Why aren’t we systematically collecting good data on data center energy use?
Back in 2018, the EIA did “assess the feasibility of collecting data and publishing estimates for data center buildings”. Participation rates in a pilot survey were very low. A frustrated EIA concluded that collecting reliable data on data center energy use would require “cooperation from the industry” and that effective data collection is “likely not feasible with current methods”.
Recently, Europe managed to establish detailed reporting requirements for the “energy performance and sustainability of data centers” which will take effect next year. Here in the US, Senator Whitehouse is developing draft legislation that would similarly require data centers in the U.S. to report operational data and environmental performance information. Following!
The widespread use of generative AI is raising all sorts of promise-or-peril-type questions around political polarization, intellectual property, higher education, the clean energy transition. Policymakers are scrambling to understand how we can use regulation and policy tools to steer the responsible growth of AI. When it comes to managing data center energy use and GHG emissions, it seems we already have the policy tools in hand. What we need now are better data and the political will to use them effectively.
Keep up with Energy Institute blogs, research, and events on X @energyathaas
Suggested citation: Fowlie, Meredith, “Data Centers Are Booming”, Energy Institute Blog, UC Berkeley, October 9, 2023, https://energyathaas.wordpress.com/2023/10/09/data-centers-are-booming/
Categories
Energy Institute Blog 
As Meredith noted, “When it comes to managing data center energy use and GHG emissions, it seems we already have the policy tools in hand. What we need now are better data and the political will to use them effectively.”
Newsom just signed a bill requiring companies operating in California with revenues of at least $1 billion to disclose their carbon footprints, and that will apply to about 5,400 companies according to the LA Times. I don’t see why legislation can’t be introduced next year to compel data centers to provide energy consumption information. But that requires strong political will because I’m quite sure the IT firms that run these centers will argue, with a phalanx of lobbyists, that this data would reveal trade secrets and compromise intellectual property — blah blah blah. Well, as Meredith points out, “Recently, Europe managed to establish detailed reporting requirements for the “energy performance and sustainability of data centers” which will take effect next year.” The California legislature loves to cite and often emulate what the EU is doing on energy and sustainability. So, if Europe can do this, why can’t we? Best of luck to Sen. Whitehouse, but I don’t think the prospects for success of his legislation are particularly positive.
Thanks for all the references! This part of your post-
“The widespread use of generative AI is raising all sorts of promise-or-peril-type questions around political polarization, intellectual property, higher education, the clean energy transition. Policymakers are scrambling to understand how we can use regulation and policy tools to steer the responsible growth of AI.” got me thinking a bit about recent discussions around the role of nuclear power.
It will be interesting to see how AI frames the CEC’s recent findings as noted by PG&E-
“PG&E finds the CEC’s conclusions to be reasonable. That is, California does not have sufficient resources adequate to substitute for DCPP and will not be able to maintain the same level of reliability or progress towards its decarbonization goals without extended operations at DCPP. PG&E appreciates the opportunity to comment on the Staff Report.”
Mark Miller
the lowest hanging fruit is ‘usage’. missing from this essay is the energy used per search > 0.0003 kwh EACH. but add to that what comes after the search — a streaming, a video up or download. if all 5b users did ONE search a day it would be 1.5 MILLION kWh.
No mention of nuclear power and the movement to power data centers and factories with small nuclear reactors, as Dow and Microsoft plan. Do you really believe in a “green transition”?
Data centers will also have considerable land use requirements that will conflict with restoration of habitat and sustainable land and water management practices. This will further conflict with solar farms and community solar, which needs to be secondary to restoration of ecosystem services.
Much of the policy Haas has influenced over the past years has omitted benefits of rooftop solar, such as land use, need to re-wild on land competing with solar farms, localization of the economy, transmission cost reduction, and cost effectiveness to the electrical system as a whole. Needs for our land will continue to come up as long as people remain on this planet.
I live in Santa Clara, CA, the “Data Center Capital of Silicon Valley”. Our utility, Silicon Valley Power, is a city-owned nonprofit that charges about 40% less than PG&E for commercial power. (Residential is 58% less.) The city’s annual power consumption in 2022 was 4414 GWh; the 2031 projection is ~7100 GWh, mainly from expected data center growth. The city is trying to build out its renewal energy portfolio but is hampered by transmission bottlenecks. (According to an SVP manager, “What used to take 2-3 years to complete a project can now take 3-7 years to get interconnection approval and then you still need to build.”) If you look at north Santa Clara on Google Earth you can see vast expanses of flat industrial roof space (including data centers) and parking lots that could be used for distributed solar (e.g., I would guess the Great America/Levi Stadium parking lots could supply ~10 GWh/yr). Solar generation within the city’s 18 square miles probably wouldn’t go very far toward filling the >2000 GWh shortfall by 2031, although the city has not evaluated and does not know the resource potential. The San Jose industrial zone around the airport has much greater potential, which might be exploited if the city and/or PG&E could develop a more cost-competitive utility business model.
Whether relating to economics or energy, efficiency gains can be expected to stagnate over time as they approach a theoretical limit. Though breakthroughs in technology occasionally bring about corresponding increases in efficiency, those leaps become progressively more scarce.
Similarly, improvements in emissions efficiency (per unit of electrical energy) have stagnated over time in California, the U.S., Germany, and Australia, due to reliance on intermittent energy from the sun and wind. The inherent limit is a product of dependence on natural gas to ensure grid reliability.
Electricity consumption will increase dramatically in coming years, with increased reliance on data centers and electrified transportation. Thus, linking carbon emissions to it creates an impenetrable barrier to effectively addressing climate change. It isn’t consumption of energy that’s causing climate change; it’s reliance on fossil fuel combustion. Minus any progress in adoption of carbon-free, dispatchable sources of energy, we can only expect progress on emissions reductions to stagnate over time.
Yes, let’s put this in perspective. California’s population is about 0.5% of the global population. California’s energy consumption is 278TWh and global consumption is about 27,100TWh or 0.8% Which means that data centers, globally are using about the the same.
I agree that data centers should take steps to reduce their energy usage. We all should. But there is lower hanging fruit. One such plum is rooftop solar. All ownership costs are borne by the roof owner, not the ratepayer. The energy is delivered to the nearest consumer, not over long transmission lines.
Conservation is the most cost effective way to approach energy solutions. After that, local generation. Therefore, data centers need more efficient computers and memory devices. Then put solar panels on their roofs.
This article is utility prestidigitation of another sort.