The presentation was based on a new report Dr. Green released in July on What Happens When Data Centers Come To Town? that documents the environmental and economic impacts caused by data centers along with their impacts on local resources, infrastructure and the electric grid.
Key findings of the report include--
-- Increased Utility Rates: Data centers increase local electric utility rates by driving up overall energy demand, which can strain grid capacity and force utilities to invest in costly infrastructure upgrades. These costs are passed on to residents through higher rates. Data centers have also secured long-term power agreements, which reduce the available supply and push prices up for other consumers.
-- High Resource Consumption: A single data center can consume up to 2 megawatt hours of power—equivalent to the power used by 2,000 homes—and millions of gallons of water annually for cooling, straining local resources and infrastructure.
-- Ineffective Tax Incentives: Tax breaks for data centers do not deliver the promised economic benefits, such as high-paying jobs, and they reduce local tax revenues, while shifting financial burdens onto communities and schools.
-- Climate and Energy Challenges: Data centers’ massive energy demands are prolonging the operation of fossil fuel plants and undermining state renewable energy goals, as seen in states like Michigan, Virginia, and Nebraska.
-- Resource Efficiency Trade-Off: While advanced cooling methods like liquid immersion and direct-to-chip cooling offer energy efficiency improvements, current technologies force a tradeoff between energy and water efficiency, limiting sustainable solutions.
-- Policy Solutions: To mitigate data centers’ environmental impacts and align their growth with sustainability goals, policymakers should adopt model laws like the 2023 German Energy Efficiency Act, add requirements for new renewable energy, and enforce transparency through mandatory reporting.
Excerpts From Dr. Green’s Remarks
Here are excerpts from Dr. Green’s November 12 remarks before the Citizens Advisory Council and Environmental Justice Advisory Board--
Every Data Center Is Hyperscale
Data centers are essentially warehouses with IT equipment, servers, hard drives, processors especially CPUs and GPUs, networking equipment and so on.
The recent boom in the generative A.I. industry has led to a significant increase in data center construction and demand, particularly because generative AI tools are incredibly resource and computationally expensive.
Generative A.I. requires much more computational power to operate than other forms of computing and networking and so that's a big reason why data centers are really being pushed for by the tech industry.
The scale of data centers has rapidly increased over the last several years, and so nowadays it's common to refer to this term of hyperscale data centers.
There's not really a single definition of what this means, but overall it's saying that these are much larger than your traditional data centers.
But pretty much every data center now is a hyperscale data center, because the size is much larger certainly than this.
Maybe the closest thing to a standard definition is having more than 10,000 square feet makes a data center hyperscale, but nowadays, you're getting orders of magnitude more than that.
Just a few miles down the road from me in Saline, Michigan, OpenAI recently announced just a few weeks ago a data center project that would be 2.2 million square feet, so pretty massive in nature.
And so often the companies that are behind these hyperscale data centers are known as hyperscalers. A company like OpenAI or Meta would be considered a hyperscaler.
Here's a quick breakdown of how data centers are distributed across the United States--
Virginia is really the epicenter of data centers in the country. Then Texas and California are next in line with the highest number, but there's quite a bit throughout the Midwest, the plains and so on.
Incredibly Resource Intensive
Okay, so the important thing more for this conversation is to understand the resource consumption of data centers.
One way to understand that is just that generative A.I. is incredibly resource intensive.
Generative A.I. requires a ton more energy and resources to operate than other times other types of machine learning or A.I..
You can see this is just a chart from a research paper showing that all the way on the left, something like text classification or even image classification is actually pretty low.
But then once you get to summarization or image captioning and then certainly image generation, you're getting orders of magnitude more emissions that are involved in training and using these types of models.
We can look at the very microscale and the very macro scale and roughly try to get a sense of the energy numbers here.
100 Word Email = 1 LED Light Burning For 14 Hours
On a very small scale, you can think of creating a 100 word email using an A.I. chatbot like ChatGPT.
That would be equivalent roughly to powering 14 LED light bulbs for an hour or maybe one LED light bulb for 14 hours.
So you leave an LED light bulb in your house open or on for half the day, that's the equivalent amount of electricity as asking ChatGPT to write a 100 word email.
Now you can imagine that scaling up massively amongst all the people who are using these tools and all the different types of tasks that they are trying to use it for.
20% Of US Electricity By 2030?
You get a projection like this from the [US] Department of Energy estimating, describing how the electricity needed for data centers has tripled over the past decade.
And then they have projections for what would happen into the second half of the 2020s, and they estimate roughly that it would double or triple again by 2028 and it could by that point consume as much as 12% of the nation's electricity.
Other estimates put it at 17% of all US electricity by 2030, so that's a pretty significant margin of the country's entire electricity and is very quickly moving from something that's roughly a few years ago was maybe in the one to 2% range and by 2030 it could be up between 15% and 20%.
That's what these projections are showing. That's really a ten times increase just in terms of relative share of total US electricity consumption. So pretty remarkable just rise in terms of how much this is a major factor in the US energy system.
Changing Company Power Policies
And you can see this also in changing corporate priorities where these hyperscaler companies used to back in the nice days maybe of 2018, they were like, we really care about the environment.
We're going to be carbon-neutral, we're going to be carbon negative.
But once they became clear that Generative A.I. was going to be the next phase of their business and it was very intensive computationally, they dropped that and actually now their emissions are rising pretty rapidly.
Water Consumption - 100 Words = 1 Bottle Of Water
To shift towards water, we can go back to that same example of a 100 word email and that's about one bottle of water.
The water comes into play because having the servers and the GPUs actually run the computations that are necessary for say writing an email or creating an image or a video, that generates a ton of heat.
You can think of how maybe on your laptop or your computer, if you're doing a bunch of tasks at once and you're overstressing the system, you probably start to feel that it's getting hot and the fan starts blowing really loudly.
That's essentially what is happening here, so you have this massive scale computer system and you are doing an incredibly intensive task, and so all of the equipment gets incredibly hot because working very hard.
Water is the main source of cooling for the system, so that's where water comes into play as a secondary input for the systems.
1 Company, 1 Year, 6 Billion Gallons
In terms of at least Google, some of the data centers that use the most water, some of these are using hundreds of millions of gallons of water per year.
This is describing 6 billion globally. Google's data centers consume 6 billion gallons a year, enough to maintain 41 golf courses, and this can be many thousands of families worth of annual water usage.
Pretty significant amounts of water as well.
‘Incredibly’ Difficult To Get Information
One thing I should note, and we'll probably come back to this, is that often actually getting information about the electricity and water usage of these data centers is incredibly difficult.
The data center companies do not generally release this information voluntarily when folks try to ask for it or obtain it via public records requests.
They are withholding and restrictive and really do everything in their power to not share that information and often have binding nondisclosure agreements or other types of arrangements with the local government so that it also can't be released by directly going to say the utility companies and asking them how much water did you send over to this data center?
But these are the best numbers that are out there.
Local Impacts
But all of this is these very broad global, or at least nationwide estimates.
I think where the story gets much more, both interesting and concerning, is when you think about the local impacts, so certainly global emissions, global use of electricity, global use of water, that's all bad.
But I think the most pressing environmental justice concerns and economic concerns come up in terms of the local impacts for people who live in the area surrounded by data centers.
So this, as I was describing, data centers are creating a huge surge in demand for electricity and it is really straining America's power supply.
We are not in a situation where we have tons of excess electricity as a nation that we can just say, "Oh, great data centers, sure, we've got lots of electricity on hand."
Around the country, utilities are really struggling to keep up and are expecting that they're going to run out of transmission capacity in many places absent major upgrades.
Here's just one way of making that, just that rise in need a little bit more vivid.
Georgia Double Power Needs Forecast
Here is a graph from Georgia's power utility describing their changing projections for how much electricity would be needed in their system to meet demand over the next 15 years or so, and that is rising rapidly.
In the state's 2023 budget, they originally put forward this estimate that was in gray, and then just in September, later that year, so less than a year later, they actually had to update their projections because they realized how far they had underestimated this new line in blue.
You're seeing especially in that 2025 to 2030 range, a really significant difference where say 2028, they thought they would be good with their capacity until '28.
Actually, they think they need three gigawatts by then or in 2020, they thought maybe we would only need two gigawatts of additional capacity by 2030.
Actually, it's looking more like six. These are both huge.
What you're seeing here is both a really stark and sharp increase in expected demand, but also that even that expectation of expected demand is rising rapidly as data center development is in such a boom and utilities are just realizing more and more how quickly this is going to increase demand on their systems.
Increasing Costs Passed On To Ratepayers
This then creates a lot of challenges at the local level.
One of the big issues that I think is most salient for a lot of people is that electricity prices are rising for consumers.
One estimate from Bloomberg found that electricity prices have more than doubled in over five years in many areas near data centers.
And so because data centers are creating a huge surge in demand, often the utilities have to do pretty large infrastructure upgrades.
And so yeah, a lot of these costs are being passed to consumers and there's been a number of articles about describing that effect, and this has been the most salient critique of data centers probably, and you've already started to see, I mean this has come up even just in some of the recent elections where candidates are worried on both sides of the aisle in terms of affordability and rising electricity prices and so on.
That's a big issue and potentially could be rising to a point where it's taking more just the basic utility of electricity is costing a higher and higher percentage of people's money that they have in a given month to spend or maybe they won't be able to afford it.
Data Centers Get Discounted Rates
Part of the reason this happens is that data centers often get discounted or special rates in their arrangements with utilities.
That's often a way to attract development of data centers in a given region, which is just one more reason why the costs are getting passed on to consumers.
Extended Use Of Fossil Fuel Power Plants
One of the issues that this creates is then that it is prolonged use of fossil fuel power plants in many states.
A lot of states have made efforts to say, "We're going to close our coal plants because we want to shift towards renewable energy," but then this rise in demand means that the only way they can maintain a sufficient capacity to power the data centers is to keep some of these coal plants open.
This just screenshot here is just from an article in The Washington Post that talks about a black neighborhood in North Omaha with some of the highest asthma rates in the country, which has a coal burning plant that was set to close in 2023, but has been kept open at least for several additional years because of the demand from data centers.
This is causing both that plant to stay open and continuing the harms of pollution, particularly in this black neighborhood that's around the power plant.
And even the coal industry is now starting to recognize that the A.I. boom is actually a boon for itself and various projections of the decrease in the US's decrease in use of coal for power is going to slow down quite a bit as there's just a greater need for coal power to be online in order to meet the demand for energy from data centers.
Increase Unreliability Of Grid
Another aspect of the electricity issue is just that these data centers are making power more volatile because of the way that data centers use often in very large quantities and can have very sharp changes in how much electricity they're using from the grid at a given moment.
It has led to a more volatile grid just in terms of the electricity coming into people's homes so this was a really interesting analysis from Bloomberg, but that can lead to sparks and fires and broken appliances in people's homes if the electricity is now not as stable just coming into your house.
Dirty Backup Generators
Back on the theme of pollution has been a big issue around data centers. One of the issues that often these data centers will have various forms of turbines or generators on site to power the data center, provide additional power or at least as a backup depending on the system, and that can lead to pretty significant pollutants being released into the air.
This has been a particularly big controversy in Memphis where a data center operated by the company xA.I., has been operating dozens of unpermitted methane gas turbines without any public notice or permits or air pollution control.
These turbines are emitting chemicals like formaldehyde and smog forming pollution, which can lead to asthma, heart disease, cancer, and other issues.
These are a pretty big issue in terms of pollution from the generators.
Water Use In Water Stressed Areas
Water is of course now another piece of the story, and I would say the water issue is a bit more localized than the electricity issue. I think the water issue really comes up most starkly in places with higher levels of water stress.
But there are many data centers built in California, Arizona, and Texas where often there are significant issues with not enough water, and that's where the data centers often are located, partly because in those places, one reason would be that you can get a lot of sunlight, and so if you have solar panels as a way to get some additional energy for the data center, that can be nice, but then you're still needing the water in places that really just don't have enough water to begin with.
Lack Of Transparency
One interesting article that really dives into this is from Karen Hao in The Atlantic writing about a data center in Arizona.
But one of the interesting pieces that comes out of this is just how secretive these companies are.
There's an incredible lack of transparency and reporting about the use.
And in many ways, I would say this is the aspect of the problem that surprised me the most when I was doing the research for my report, just how much it's actually incredibly difficult for people to get valid, reliable, up-to-date numbers on the electricity and water use that the data centers are taking.
Very Few Jobs
Now, one of the other issues to highlight is that the data centers don't really bring that many jobs.
One of the major selling points of data centers for communities has been, oh, well this is economic development and this is going to bring jobs to your community.
The reality is that's really just not the case.
There are a number of studies and articles that have looked at this and found really disappointing numbers in terms of how many stable well-paying jobs are actually created by data centers.
The construction process does create jobs. You're building a big facility and that does create work for the construction workers, electricians, HVAC, et cetera.
But once data centers are built, they require relatively few employees since really what they're doing is just housing computers and servers.
They're not an office space. They're not a tech office space that maybe some policymakers imagine of, oh, we're getting a Google office in our community and we're going to have tech workers and developers and designers.
It's really not that, it's mostly computers and servers. And then you have a variety of low-wage term-limited positions, often more along the lines of security and maintenance.
States Losing Lots Of Money On Giving Tax Breaks
A pretty detailed report is from the organization Good Jobs First, which shows both the lack of jobs that are created by data centers, but also in the context of tax breaks, just how much money a lot of states are losing.
The promise of jobs is often associated with tax breaks that states will provide to data centers as a way to attract their development, and Good Jobs First has shown both just how few jobs are actually created relative to the amount of tax revenue that is lost and how these are making very large dents in state budgets.
They found that over the span of two years, Texas revised its estimate of the tax breaks that they were providing to data centers from $130 million to 1 billion dollars.
They're losing a billion dollars a year, two tax breaks for data centers, so a pretty significant amount of money, which in many states that money would go to schools or other social services.
Final Notes
I want to leave plenty of time for questions. Just a few final notes that I'll run through quickly.
Incredible Capital Expenditures
One is just to look at, here's a chart of capital expenditure from some of the major tech companies on data centers, just as a way to see how much this has, has boomed really incredible amounts of money that these companies are putting into data center construction.
And I'm sure many of you have seen articles just over the last couple of weeks in newspapers about how they're doubling down, they're putting more money in, and certainly I think there are a lot of valid questions of whether this money will get recouped, whether we're in a bubble, what will happen to all of this construction and whether it's actually even viable from a business perspective, or if it's actually bringing the economy to a pretty precarious position.
Just one aspect of that, these GPUs are incredibly expensive.
It is not just like buying a laptop. Each chip can be on the order of 30,000 or 40,000 dollars and they represent a very significant cost of the data centers, but also depreciate over time.
The GPUs, new models are coming out every couple of years, and that means that a GPU that a company buys today may not actually be worth so much or as useful three or five years down the road.
Data Centers Are Resisting Regulation
Thirty-six states have passed some kind of legislation authorizing tax incentives for new data center development, often on the promise of creating jobs and so on, which as I described earlier, was somewhat of a false promise.
They're really happy about getting tax breaks and they're really resistant to any sort of regulation that would limit their ability to do what they want.
I think a really striking example of this is that the California legislature passed a bill that would require data centers to disclose their water consumption, and there was enough lobbying by the data center industry that Governor Newsom vetoed this bill.
This wasn't even a bill that would require them to change their practices just to report on their water usage and he ended up vetoing it.
I think that just goes to show, think it's a pretty helpful data point for understanding what the current landscape is around policy.
I'll stop there so that we have some time for conversation and questions. Yeah, I'll just leave this up for a moment.
As a reminder, this is the report where there's, not all of these articles are mentioned in there, but a lot of these details are fleshed out in more detail.
Click Here for What Happens When Data Centers Come To Town?.
Find a copy of Dr. Green’s presentation slides on DEP’s Citizens Advisory Council webpage [when posted]. Questions should be directed to Ian Irvin, Executive Director - CAC, iirvin@pa.gov or (717) 787-8171.
Resource Link:
-- PennFuture Data Center Resources: Model Zoning Ordinance
Related Articles This Week:
-- DEP Invites Comments On Air Quality Permit For 176 - 3.25 MW Backup Diesel Generators For A.I. Data Center In Former LSC Communication East Facility, City Of Lancaster [PaEN]
-- DEP Invites Comments On Chapter 105 Water Quality Permit For Project Gravity A.I. Data Center Project In Archbald Boro, Lackawanna County [PaEN]
-- 500 Feet Isn’t Enough: House Environmental Committee To Hold Nov. 17 Hearing On Bill To Increase Setbacks From Unconventional Shale Gas Wells From 500 Feet To 2,500 Feet From Homes, 5,000 Feet From Schools, Hospitals [PaEN]
NewsClips:
-- Ohio River Valley Institute Blog: Why A.I. Data Centers Will Be Economic Development Duds In Pennsylvania
-- Inside Climate News: New Unifying Issue - Just About Everyone Hates A.I. Data Centers
-- PennLive Guest Essay: PA Electric Ratepayers Will Foot The Bill From Three Mile Island Nuclear Plant Restart, While Microsoft A.I. Data Centers Get The Power - By Eric Epstein, Three Mile Island Alert
-- TribLive: Questions Remain After Dueling Meetings On Springdale A.I. Data Center In Allegheny County
-- Post-Gazette: 3,200 Acre Homer City A.I. Data Center Campus Seeks To Become A Natural Gas-Fueled Powerhouse In Indiana County
-- Morning Call: Large A.I. Data Center In The Works For Former Air Products Site In Upper Macungie Twp., Lehigh County [PDF of Article]
-- LancasterOnline: Developers Face Residents At A.I. Data Center Open House In Lancaster City [PDF of Article]
-- LancasterOnline: Proposed A.I. Data Center Rules, Pact Won’t Reach Lancaster City Council This Week [PDF of Article]
-- LancasterOnline: Lancaster City A.I. Data Center Developer Hosting Open House Wednesday [PDF of Article]
-- WNEP: Amazon Helps Revitalize Berwick Theater In Columbia County
-- WNEP: A.I. Data Center Developer Eyed Up New Castle Twp., Schuylkill County For New Facility, Neighboring Town’s Residents Fought Back: WNEP Investigates
-- The News-Item: Gary Namet Defeats Incumbent Mt. Carmel Twp. Supervisor, Says Opposition To A.I. Data Center Made Difference
-- The Citizens Voice: Amazon Web Services Opens Offices In Downtown Wilkes-Barre
-- Financial Times: ‘Phantom’ A.I. Data Centers Muddy Forecasts For US Power Needs
-- Inside Climate News: A.I. Data Center Use Of Diesel Generators For Backup Power Is Commonplace And Problematic
-- PA Capital-Star: PA One Of 11 States That Disclose Which Companies Get A.I. Data Center Incentives, Report Finds
[Posted: November 14, 2025] PA Environment Digest
No comments :
Post a Comment