Loading summary
Transcript25 lines
- [00:01]
A
You're listening to TechTank, a BI weekly podcast from the Brookings Institution exploring the most consequential technology issues of our time. From racial bias and algorithms to the future of work, TechTank takes big ideas and makes them accessible.
- [00:27]
B
Welcome to our Tech Tank podcast. I'm Darrell West, a senior fellow in the center for Technology Innovation at the Brookings Institution. Data centers are generating a lot of attention due to their role in AI and the massive amount of money being invested in them. They are powering the AI revolution and the movement towards a digital economy, yet they also are requiring enormous energy and needing both electricity and water. Today, I am pleased to be joined by Joe Kane. He is a fellow in the Brookings Metro program and he recently authored a paper entitled AI Data Centers and Water. And it looks at the water needs of data centers and how states and localities should respond to those requirements. Joe, welcome to our Brookings Tech Tank podcast.
- [01:17]
C
Thanks for having me, Darrell.
- [01:19]
B
So your paper does a terrific job discussing the resource limitations facing data centers and the need for new approaches. And there's so much attention being paid about the need for electricity. One of the things I appreciated about your paper is you focus on water, which is also a very important need for data centers. Now, before we get into your specific recommendations, let's just talk about the water needs. Why do data centers need so much water?
- [01:46]
C
Yeah, so a great, great context setting, Darrell. So to offer some additional context, data centers are obviously growing importance given the increased demand to store and manage digital information. You know, doing that requires a ton of servers and other equipment, which not only takes up space in terms of land, but also needs other natural resources, including energy to power these systems and then water to cool them down. You know, the specific tech for cooling can vary widely, which I think we'll, we'll probably get into a little bit more later on, depending on the size, location and function of a given data center. But. But in general, air conditioning and evaporative cooling are commonplace, which require sustainable and predictable water resources. So think of chillers, cooling towers, and other infrastructure that need water. And water can come from surface water sources like rivers, streams, reservoirs, can also come from groundwater sources or even be reused in some cases. So, you know, the devil is in the details of how and how much water is needed. But it's pretty obvious that the appetite of tech firms, economic development leaders, and others in this space is for more water to support such developments.
- [03:10]
B
No, I think that's a great point. And, you know, for these hyperscaler data centers, which are the Large scale ones, they often have 5,000 to even 10,000 file servers located in them. So you know, you mentioned the need for chillers, cooling towers and so on. Like when you have thousands of file servers operating simultaneously, like the need for water to cool them does become a very important factor. So what scale of water are we talking about here? Like what, what is going to be required as data centers and AI starts to move forward?
- [03:44]
C
Yeah, I mean, it's going to depend. I mean, let's be clear, not only is the specific amount of water needed going to vary from facility to facility, as you were describing, but you also have to think of the various and evolving needs for water. I've already briefly mentioned its purpose for cooling, specifically for on site water usage. I mean, for example, a typical data center uses about 300,000 gallons of water each day. That's, that's equivalent to demands of about a thousand households. But, but large data centers can use an estimated 5 million gallons of water each day, which is equivalent to the needs of a town of up to 50,000 residents. You know. Moreover, projections show water used for cooling may increase 870%. That is not a typo in the coming years as more facilities come online. But there is also considerable off site usage, including water for power plants to generate electricity and all the water consumed during the manufacturing process to create chips and other equipment. And we don't always have reliable, consistent data to track this water usage in real time or even over time. So just as data centers and AI are part of a larger technological and economic development ecosystem, they're also part of a larger natural and built environment. You know, water is an essential interconnected resource that regional leaders, not just individual firms and developers, need to weigh carefully in ongoing projects.
- [05:20]
B
Now, the good news is that there are some new technologies coming on board that may help with this water issue. And in your paper you discuss some of the ways that renewable energy can help, as well as some new approaches to cooling, closed loop cooling systems and so on. So I'm just curious, what are some of the technologies or new approaches that could help us manage this resource need?
- [05:45]
C
Right, so there are ultimately bigger planning and policy dimensions at play here that, that we'll talk about in a bit. But, but yes, new technologies are helping alleviate some of these water demands. And obviously for both tech firms that need this resource and water utilities that provide this service, there is a mutual interest in seeking and implementing more sustainable and cost effective solutions. I mean, right now we have a situation of high water consumption since so much water for cooling is lost to evaporation. And then you also have wastewater that has contaminants and other issues associated with it. But this is where more circular solutions and technologies can come into play, especially around water reuse and recycling. You know, for instance, as you had mentioned, closed loop cooling systems may reuse wastewater or even harvest rainwater, which can reduce freshwater use by up to 70%. Other design improvements, such as immersion cooling, where servers can be literally dipped in a special solution, can remove heat and help limit water use too. And taking advantage of renewable energy sources, including on site solar, can coincide with these upgrades to boost overall sustainability. Leaders do need to think about our water resources alongside other energy and natural resources more holistically. And, and we already see this in other parts of the infrastructure space around the energy water nexus, for example, so. So a lot of promise around these, these different technologies.
- [07:18]
B
I mean, I think one of the nice things is as part of developing data centers, companies are trying to become more efficient because they need to. And they're going to be payoffs in a lot of other areas because water is a limiting factor, not just on data centers, but in many economic development activities and population is growing. So hopefully there'll be some payoffs in some other areas. But in your paper, you also point out the technology alone probably is not going to solve the water problem. You argue that we need better regional coordination of water resources. So I'm just curious, what should states and localities be doing to address this water issue in regard to data centers?
- [07:58]
C
Great question, and there's honestly a lot bundled into that. Probably too much for just this, this one podcast, but. But I want to shine more light, especially on the individual local water systems responsible for providing service. And increasingly more systems in more places are grappling with data centers, water impacts, and those of an AI driven world more generally, including significant and ongoing water use, expanded infrastructure costs, such as new distribution lines to reach exurban facilities, and other regulatory and financial pressures. These concerns are generally out of sight and out of mind when it comes to regional economic development planning. Aging pipes, treatment plants and other systems are weighing heavily on local water systems. But many businesses and households for that matter, do not think about these physical challenges outside of their rising bills or a sudden service disruption. For example, you can add in climate impacts and pullbacks in federal and state infrastructure funding, and what you see is, is a need for greater capacity to proactively stay ahead of all these needs, technically, financially, managerially, and so on.
- [09:08]
B
So one of the things I like about Brookings is we focus not just on problems, but we also make recommendations, we try and think about what are the possible remedies that could help with whatever issue that we have identified. And I think one of the strengths of your paper is you actually do talk about ways to deal with these water issues. And I know that not every region faces the same type of water challenges. So I'm just curious, which localities face the biggest kinds of problems in regard to water and what are they doing to meet these challenges?
- [09:43]
C
Right. And as I was just describing, I mean, there's an immense scale and variety to all these infrastructure needs. I mean, more than 50,000 local water systems actually operate across the country, which is mind boggling for a lot of people. So there's huge fragmentation and localization to these challenges, not just associated with data centers and AI in isolation. It's those places, I'll say, with declining populations and shrinking customer bases and other economic struggles that often have the biggest challenges to planning and paying for their water infrastructure needs. So you think of places like Flint, Michigan, Jackson, Mississippi, cities that have really struggled for decades, not just in a year or two, but decades, to stay ahead of their needs. And we're seeing with the rise of data centers and AI, huge challenges emerging, even in economically growing areas with significant water constraints. So you think of parts of the west, the southwest, areas that have had pretty significant drought and water scarcity challenges amid housing growth, amid other economic growth. I mean, this is just adding a new variable into the mix. So it is this balance, Darrell, of, of kind of slower growing, economically struggling places with, with sort of paying for and staying ahead of their infrastructure challenges. But then also many places too, that, that are seeing new pressures, new triggers that are requiring them to fundamentally rethink how they plan and manage their water resources more sustainably.
- [11:21]
B
And I know one particular idea that you focus on in your paper and many of your colleagues in the Metro program also focus on this is just building new infrastructure. And I think you've mentioned, you know, the aging pipes, the old treatment plants, pipes that are leaking. What is it this city should be doing to kind of build new infrastructure? How can they handle stormwater runoff? Questions. How do these things help provide possible remedies for the water limitations that some communities are facing.
- [11:55]
C
Yeah, I mean, there's a physical sort of capital upgrades, of course, and incorporating new designs and technologies, more distributed and flexible water systems. When we think of, you know, green infrastructure, for example, which can absorb heat and better manage runoff, but. But ultimately it takes a new approach as well. And so in particular, what I've Stressed and have heard from many communities across the country, is the need for more regional thinking and planning. Doing so can help build capacity and ultimately seize economic opportunity. So given the fragmentation we see of all these small individual systems, it's quite literally impossible for them by themselves, many of them anyway, to handle all of this. And I think, you know, they're grappling with that reality increasingly over time out of necessity. I mean, some rural communities with particularly acute capacity constraints, you know, with, with really, I mean, not even able to find staff to manage these systems, they're considering regional approaches. But, but leaders in larger urban regions too are recognizing the benefits of formal and informal regional collaborations. So we see, for example, joint water investment and economic planning have evolved for decades in places like Las Vegas. We've also seen emerging tech focused efforts taking root in places like Chicago and the Great Lakes. Even climate and workforce driven regional plans and partnerships have evolved in markets ranging from New Orleans to San Francisco. More directly in terms of AI as part of capital plans and budgets, we see data center hubs such as Loudoun county in Virginia, right outside Washington D.C. you know, leaders have started to better measure and forecast tech impacted infrastructure needs. So I want to emphasize there are big sprawling water challenges all over the place, but also a lot of promising collaborations and efforts unfolding too.
- [14:03]
B
Yeah, it does seem like regional thinking and comprehensive planning would be a big part of the remedies because you mentioned just how fragmented the water management systems are. We know governance in a lot of places is pretty fragmented as well. Suburbs often don't like to cooperate with cities and vice versa. So any governance improvements that would address those issues certainly would be helpful. In your paper, you also kind of mentioned the idea of spreading the financial risk. Like when we talk about upgrading the infrastructure, it probably challenging for any particular locality to deal with this on its own. How can we spread the financial risk in a way that would benefit the entire ecosystem?
- [14:45]
C
Yeah, I mean, and this is not a new challenge either, right? I mean, it certainly, it was heightened, you know, has been heightened, you know, in pullback of federal resources, you know, but, but utilities have had to grapple with sort of the proactive and reactive nature of staying ahead of their infrastructure needs, including a changing customer base, having to increase water rates, which are the traditional lever, if you will, that they have available to them to pay for infrastructure upgrades. And then many places are taking on debt, right? Municipal bonds, for example, are really common in this place. But, but places can't just indefinitely do that. And individual systems recognize that if they keep going it alone, you know, operationally, managerially, you know, it's a real struggle. And as we're seeing, you know, not again, not just in the context of data centers and AI, but, but that's a tipping point, you know, for a lot of places of maybe, maybe for example, there have been proposals for, you know, increased water usage in data centers. Should there be a recalculation or rethinking of the water rates for those users? Right. We've, we've seen places also rethinking, you know, agricultural water use, which is a huge, huge challenge in parts of California and Arizona, for example. And so I think there is a willingness and hunger for experimentation and breaking free from business as usual mindsets. But, but again, the traditional nature of this sector and again, that sort of the fragmentation of it and the regulatory nature of it, you know, a lot of water utilities, you know, they have to comply with the Clean Water act and the Safe Drinking Water act. So, you know, it's not usually to their advantage to try a bunch of new things, you know, rocking the boat. So, but, but certainly seeking alternative funding and financing is, is important and certainly part of the equation we're talking about here.
- [16:45]
B
And I think one thing that companies could do is just do a better job at measurement and transparency of what their needs actually are. A few weeks ago, Nicole Turner Lee and I wrote a paper on the future of data centers, and we argued that it would help communities a lot in terms of their planning process and their management, both on electricity as well as water, just to be more transparent about what their needs are so that people understand what the issues are. It then becomes, I think, easier to think about possible remedies. Are there particular states or cities that you think are doing a good job managing water with data centers, it's really
- [17:24]
C
across the board, honestly. I mean, the, I think again, it's because of necessity and reaction in real time. In most cases where I mentioned parts of Virginia, parts of the west, the southwest areas, where they're already seeing data centers proliferating, if not projected to increase over time. And so I think those places, you know, have had to adapt and have had to, you know, basically figure out how to address these issues. And so, I mean, those are the examples I have seen across the country. But with that said, I mean, as, as data centers of all different shapes and sizes and varieties emerge, you know, this is going to be an issue in regions that have, you know, not as much water scarcity. They may have a lot of water, but, but there could be other uses and use categories they need to consider. There could be other environmental quality and pollution concerns. And, and all of this needs to be balanced as well with broader sort of land use and resource conversations that, you know, water isn't, isn't just in its own silo, ideally, you know, and that's, that's been certainly a topic for, and a priority for many water water utility leaders is, well, how do they coordinate more effectively as part of the economic development process? They're not just sort of an offshoot or a tangent to it, but they're actually integrated as part of that. And, you know, I don't want to sound Pollyannish. Part of that is setting the table, you know, literally and figuratively, where these leaders can get together and not just, you know, do it out of necessity, but, but ideally do it in advance of, of the next project or proposal that could come their way. And so those are the sorts of, of efforts, and I mean, not so much a specific place or two, but, but the, the types of efforts I'm monitoring all across the country.
- [19:30]
B
And you mentioned Northern Virginia, and of course, that area is kind of the Silicon Valley of data centers, just because there's an enormous concentration of data centers in this part of the world, in part because of the proximity to federal agencies and the energy sources that were available here. And it's interesting, there's a new governor who is just elected in Virginia, and data centers and rising electricity rates actually were a part of her campaign, and she talked a lot about affordability. So I think Virginia will be an interesting place to watch, just both on the electricity side as well as on the water side, in terms of she's thinking about this issue. She knows Virginia is kind of a hotbed of activism, community activism in this area. It's very much on her agenda. And so I think it'll be instructive for people to watch how she handles some of these issues. The water crisis is occurring at a time when there also is a reduction in federal aid in a number of different areas, as well as staffing reductions that could compromise the ability to handle some of the issues that we have been talking about. What should the federal government be doing to help states and cities with data centers?
- [20:46]
C
Yeah, absolutely, Darrell. And I mean, I think your Virginia example might actually be very fitting here too. But this is top of mind for so many state and local leaders at the moment involved in economic development, infrastructure investment, and just so many other functions. The reality is that most spending, most planning, and most of the ownership and operation of all this infrastructure Is especially is at a local level. And water systems get that. I mean, that's the reality they have existed in for some time. According to Congressional Budget Office data, the federal government is only responsible for about 10% of all public spending on water infrastructure each year. And, and most of that support actually comes in the form of loans, not, not even grants. So, you know, the cavalry really hasn't been coming for some time in the water sector. You know, with that said, of course, there's a continued need and desire for greater federal resourcing and leadership on our water infrastructure issues, you know, including setting clearer policy and strategic priorities that individual systems can more easily follow. You know, this has come up, for instance, around workforce development needs in this space, affordability needs and other broader technical assistance efforts to again, you know, build greater knowledge and capacity at a local level. We saw a lot of this unfold during the rollout of the Infrastructure Investment and Jobs act that, that did present historic and once in a generation funding for our water infrastructure. But, but we're still, you know, significant challenges in actually implementing that across the country, which is probably its own conversation. So, you know, to me, the federal government has a role to play as a, as a convener, a thought leader, a knowledge builder around these data center and AI issues, especially given the prevailing local fragmentation and capacity constraints we're seeing across the country.
- [22:45]
B
So this water problem also needs to be addressed in the context of the ongoing challenges related to climate change and natural disasters that also have their own effects on water. How should cities coordinate their data center plans with these broader issues? Climate change, natural disasters, and economic development in general?
- [23:08]
C
Yeah, I mean, very timely. I like to say water policy is climate policy and vice versa. I like to stress that it isn't just mitigating future climate impacts, but also adapting to current and evolving climate impacts too. And certainly technological shifts, including AI and data centers are part of that equation. You know, there is quite literally a zero sum situation around water resources in different regions and places cannot indefinitely and indiscriminately keep building more data centers, more housing, and striving for more economic growth overall without bearing in mind these fundamental resource concerns. You know, you add in fire risk, drought risk, freeze risk and more, and you have a complicated situation that many leaders are having to balance that at the moment. But, but again, I think there's, there's a lot of good happening across the country too, and including even on local and state level climate planning and water planning and water investment. And, and it helps to highlight those efforts and look towards I think greater scaling, replication of best practices where we
- [24:22]
B
can well, Joe, I want to thank you very much for sharing your thoughts on data centers. You have a terrific paper. It's entitled AI Data Centers and Water, and our listeners can read it at brookings. Edu and our scholars write regularly on AI in data centers, and you can find that work on our Brookings Tech Tank blog. Thank you very much for tuning in.
- [24:53]
A
Thank you for listening to Tech Tank, a series of roundtable discussions and interviews with technology experts and policymakers. For more conversations like this, subscribe to the podcast and sign up to receive the Tech Tank newsletter for more research and analysis from the center for Technology Innovation at Brookings.