Equitable Growth in Conversation is a recurring series where we talk with economists and other academics to help us better understand whether and how economic inequality affects economic growth and stability. In this installment, Shayna Strom, president and CEO of Equitable Growth, speaks to Costa Samaras, the director of the Carnegie Mellon University Scott Institute for Energy Innovation, the trustee professor of civil and environmental engineering, and an affiliated faculty member of the university’s Departments of Engineering and Public Policy. From 2021–2024, he served in the White House Office of Science and Technology Policy as the principal assistant director for energy, chief advisor for energy policy, and then chief advisor for the clean energy transition, where he worked in coordinating federal activities on U.S. energy policy and assessing energy technologies for meeting U.S. climate, resilience, equity, and security goals. In their conversation, Strom and Samaras discuss:

• Samaras’ career path and how it shaped his energy infrastructure research agenda
• Infrastructure and physical systems today and how they intertwine with artificial intelligence
• The relationship between energy demands, the growth of AI, and equitable growth
• Regional economies and the benefits and drawbacks of the build-out of data centers
• The coordination of AI deployment amid a clean energy transition
• State and local communities’ resistance to data center development
• Samaras’ proposal for an electricity grid “New Deal”
• What this means for growth: A grid New Deal that benefits entire communities, lowers electricity costs, and spurs clean energy development can deliver shared prosperity over the next 5 to 10 years

 
Shayna Strom: Thank you so much for joining us, Costa.

Costa Samaras: I’m delighted to be here.

Strom: You have spent your career at the intersection of energy systems, climate policy, and engineering. You worked as an engineer on large-scale public infrastructure projects, then advised the White House on energy policy, and now lead the Scott Institute for Energy Innovation at Carnegie Mellon University. How has your career path shaped your research agenda?

Samaras: As an engineer, I always think about infrastructure and physical systems, and what they mean for our society and our national objectives and the environment. Like a power plant or a bridge or a water treatment plant, infrastructure and physical systems provide functions that exist outside of policy. They work no matter who is in charge—or they don’t work no matter who is in charge.

What I found in the early part of my career was that the infrastructure systems are such a key enabling part of what we want to do for policy, yet policymakers really have not made an attempt to be explicit about what they want to get from our infrastructure systems. In the early part of my career and working with city and state government and then working at the White House, I was trying to be a boundary agent between the world of infrastructure and the world of policymaking, trying to make sure that we are getting what we want from both of those systems.

Strom: That is extremely important work. Fast forwarding to 2026 and your more recent work, how do you think artificial intelligence and energy are intertwined? The advancement of AI and the energy transition toward sources of clean and renewable energy are not parallel trends, obviously, but are deeply connected, drawing on the same supply chains, infrastructure, and institutional capacities. Can you explain why it matters that policymakers and researchers should consider these issues jointly? Where is new research needed most to inform policymaking perspectives?

Samaras: I think it is safe to say that the AI boom that we are in is no longer just a software issue. It is also a hardware issue. So, it is different from other computing advances, such as in modern telecommunications or broader use of digitization over the past several decades. Electricity was always the underpinning part of those technological advancements, but there was a societal focus mostly on the computing advantages.

The AI boom is different in that electricity is the foundational layer for what is going to enable the success, the failure, the benefits, the impacts, the cost, and the benefits of AI. And the challenge is, how do we as a society make big public choices around what this means for us? How much pollution does this generate? What does it mean for climate and water? What does it mean for the cost of electricity? What does it mean for jobs and communities?

I think up until this moment, there was a feeling in the broader AI software industry that electricity would always be there at the cost that we needed it to be, and that environmental systems would be treated separately from the innovation system needed for AI. And that is very much not what is happening. Energy is the foundational element to making all of this work. So, that means we as society have a deep role in ensuring that we get what we want out of this moment.

Strom: I’m wondering, where does the research stand on the relationship between energy demand spikes, AI expansion, and equitable growth? What do we know with confidence? What is uncertain? What should researchers or policymakers be prioritizing right now? Especially given the current pace of AI development, do we have the right indicators to measure these trends?

Samaras: I think the important research questions start with how much electricity AI will use, and that is something where more transparency from the industry can help. But policymakers and researchers alike cannot just be thinking about that. The intersection of energy and AI causes rapid changes in the ways that communities, the power system, and industry interact. And so, the policy research that is needed is multifaceted. Under what rules and under what conditions and under what assumptions does this future unfold? Who pays? Who benefits? What are the impacts? And what are the robust pathways forward that will enable good outcomes for both communities and the country under uncertainty?

There also are uncertain questions about what is going to power this boom and what pollution will look like, and where geographically, and disproportionately, the impacts and benefits will be felt.

Strom: Just to pick up on what you are saying about impacts and benefits, the historical record tells us that large infrastructure projects such as these can, without proper design, unintentionally create winners and losers. What do we know about whether this moment will be any different, and is there anything in particular required to make it different?

Samaras: Well, think about the canal system in the country, the expansion of the railroads, the growth of the early electricity grid, and the growth of the federal highway system in the middle of the previous century. All of these infrastructure systems fundamentally changed the U.S. economy, its trajectory, and the economic outcomes, with whole sets of people who benefited and whole sets of people who were either left behind or disproportionately impacted.

Consider what is happening right now as a result of our transportation infrastructure. Looking back at the growth of the highway system in the middle of the 20th century, it was an amazing feat of infrastructure that was completed quickly, yet had a real impact on many parts of communities that were divided, where communities were relocated, or communities were sited next to infrastructure that then caused a bunch of challenges that reverberate to this day.

I look at this moment in AI as a similar point in history, where we as a society should be considering what our aspirations are in terms of what we want the country and this infrastructure to look like in 10 years, 20 years, 30 years—even 50 years. If we do so now, then we might not be saying in the second half of the 21st century that we wish we had done it differently and now are going to spend a bunch of money to fix stuff.

There were important resources in the bipartisan infrastructure law enacted by Congress during the Biden administration to help reconnect communities that were divided by highways down the middle of them. With AI, I would rather we try to get this more correct now than spend the rest of our careers fixing it.

Strom: It is worth noting that the data center build-out has been quite regionally concentrated, with places such as northern Virginia hosting the largest concentration of data centers in the world. What are some of the near- and long-term consequences for those regional economies that are building out energy systems to support this infrastructure?

Samaras: There is a similar theme. Who pays, who benefits, and where are the coalitions that pull together for better outcomes for everybody? Because of the growth of the first part of the internet and the agglomeration economies that happened in northern Virginia—and the proximity to the federal government and federal clients—the Commonwealth of Virginia had long enacted policies that encouraged the development of data centers. That is why Virginia is the epicenter of data center development—not just in the United States, but in the world, too.

But there are a whole bunch of other states that are catching up quickly: Texas, Oregon, Ohio, Arizona, Georgia, and others. And there are still many, many policies in states that are trying to encourage data center development because there are tax benefits, there are local economic benefits, and there is a perceived opportunity for economic growth with the very, very large investment that many tech companies are making right now.

Under these conditions, the benefits of such substantial reindustrialization often do not flow to the communities in or near where these developments are happening. And that is where the opportunity is to align meaningful partnership between communities, state and local governments, and companies to really get the outcomes that we want ahead of time. In the United States, electricity for data centers—which includes not only all of artificial intelligence, but also everything that we ever do on the internet, such as Netflix, email, video, basically everything digital—is about 4 percent to 5 percent of total electricity demand, with projections that that will rise to between 9 percent to 17 percent by the end of the decade, according to the most recent analysis by the Electric Power Research Institute.

There is lots of uncertainty around that, but the growth of data centers is happening faster than the planning and capacity mechanisms that generally exist in the electricity system to deliver these types of growth at the scale that is currently envisioned. We have done this type of growth in the past. We have met a lot of that in the past with fossil fuel energy. We will need a new set of capacities and vision to meet all of this with clean energy.

Strom: The energy transition and AI deployment both require enormous coordination across utilities, regulators, grid operators, and public finance institutions. From your time at the White House Office of Science and Technology Policy and from your research since, I am curious where you see the biggest institutional gaps. What does closing them require?

Samaras: This comes back to the software versus hardware framing, the public versus private challenges. The data center companies want to connect in months or in a small number of years, where the utilities and broader state capacity generally plan in years to decades. So, there is just a timing mismatch between what private actors are signaling they want and what the institutions that we have established are designed to deliver. We need investment in expertise and state capacity at the state and local level and at the federal level to make sure that the technology and the timelines are not ahead of the policymaking and regulatory infrastructure, and that, physically, there are enough people to deliver beneficial outcomes for the folks in communities along the timelines on which the industry is moving.

We need investment in state capacity. We need investments in data infrastructure. We need investments in shared tools and resources. We need investments in the type of innovation for growth that we have not thought about in the electricity system for 20 years because the amount of electricity demand was growing at a near-zero pace for the past two decades but now, all of a sudden, is tasked with delivering many, many multi-billion-dollar investments in growth for the AI build-out.

So, there is a role for public finance. There is a role for philanthropy. There is a role for the federal government. There is a role for state and local governments. Those institutions have taken a giant backseat to the energy and AI industries right now, and there is an opportunity to scale those partnerships up.

Strom: As you know, the second Trump administration is accelerating federal permitting processes for data center build-outs, and there have been federal attempts to block states from regulating AI companies, preempting state policies while federal regulation remains absent. Yet Maine recently tried to pass a moratorium on future data center build-outs, and other state and local policymakers are considering similar moves. How can policymakers navigate these dynamics while advancing the energy transition and climate goals?

Samaras: Currently, the vast majority of states are still encouraging data center companies to come build in their states with incentives and regulatory opportunities. But there is a bottom-up groundswell of communities across the political spectrum that are saying, “hold on.” Maine was advancing these efforts and ultimately did not succeed, yet is emblematic of communities voicing their concerns about the speed of this build-up. Indeed, there are lots of local level moratoria that have been advanced across different communities. 

That also raises a challenge about which local communities have the capacity to organize to do that, thinking once again about who pays and who benefits in the long run. Ultimately, I think it is not about fast versus slow but rather unaccountable versus accountable. I published a policy brief in 2025 on smart AI fast lanes, arguing that speed is possible for infrastructure in this country, and with speed comes responsibility. Speed in infrastructure development needs to be earned, and there is a way that that can be done.

Strom: There has been a wave of pledges from large AI companies and utilities that aim to protect households from bearing the brunt of costs of data center-driven grid upgrades. What should researchers and policymakers be taking away from those pledges? Are there policy mechanisms in policymakers’ toolkits that would make rate-payer protection enforceable?

Samaras: As a policy person, I love pledges. It is a great opportunity to get people on the record to understand what folks’ values are. As an infrastructure person, a pledge will not run your microwave. A pledge will not pay your electricity bill. A pledge will not turn your heat on.

I think that these types of policy pledges are important first steps to show what the set of shared values are. But in the end, pledges have to come back to what the actions are, and those actions are either investments, regulatory moves, consumer protection laws, or some combination that advances faster connection in exchange for a community-responsive set of infrastructure.

Strom: We have heard you and others mention a grid “New Deal.” What would that take, and what does it look like?

Samaras: I have been talking about a grid New Deal. Because, really, I am trying to shift the lens of how we are thinking about the electricity grid as the platform for modern economic growth. All of the things that we want to do for competitiveness, innovation, security, and the environment will depend on the electricity grid. And we have a grid that was largely constructed physically in the middle of the past century, although it includes many components that are much older.

So, we need to be thinking about the grid as a national asset differently. It starts with generation, and all the type of clean power and clean firm power that we need to be constructing. It continues with transmission, the big lines that we see on big transmission towers. And it means both reinvesting in those technologies but also finding more ways to connect clean power across the country into more places. And then, it goes into the power poles that we might see in our neighborhoods, in what is called the distribution grid.

Rebuilding that last key part of our infrastructure, that last mile of infrastructure, is key for the needs of not just of data centers but also the needs of all the electrification that we are going to need in the next two decades—our buildings, our businesses, our factories, and our vehicles.

A grid New Deal is about a stronger planning function, national and regional visibility, financing that is shared between public and private institutions—and sometimes even philanthropies—and permitting functions that are legitimate, equitable, transparent, and rapid. And then, deep early community involvement that keeps bills affordable, trains workers for the types of jobs that are going to be created, and ensures that we can meet this commitment while also meeting our climate, water, and other environmental objectives.

Right now, we are wish-casting that the grid that we want will just show up by itself. And it will not. Without intentional policy and public-private partnerships, it will take longer. It will be less equitable. It will be less clean. It will be higher cost. What a grid New Deal framework starts to think through is how do we accelerate innovation, competitiveness, and deep action on climate and environment, while also building the foundational infrastructure for the 21st century.

Strom: Given all of these challenges, I am curious what you are working on right now that you are most excited about.

Samaras: A lot of the work that we do here at Carnegie Mellon’s Scott Institute for Energy Innovation cuts across a lot of these themes around what should the future look like? What do we want the future to look like, and how we can design the technologies, strategies, policies, and other systems to deliver that future?

So, our work spans everything from new materials in the lab to new approaches to making materials, such as iron and steel and aluminum, and from how we power the AI boom in ways that do not burden communities and meet our climate objectives to how we ensure that the broader transition to clean energy is good for everybody.

In my own research group, we are doing a lot of work on how the impacts of AI on the energy system can be managed and quantifying the opportunities and challenges that AI brings to the electricity sector. And that might mean more demand, more powerplant use, trying to figure out how to make those impacts have the least amount of burden on communities, and what are innovations that AI could bring to the electric sector to integrate things such as wind and solar quicker? All these developments need rapid but rigorous analysis so that policymakers can rely on science to make very important and consequential decisions around these issues in the near term.

Strom: Well, it all sounds like incredibly important work.

Strom: I want to close us off with a final question. You have talked a lot about who wins and who loses as part of this conversation. And right now, we are obviously in a moment that could either deliver broadly shared growth or exacerbate inequality. What would you point to in 5 years or 10 years as evidence that we got it right or got it wrong?

Samaras: Well, a lot of times, in policy analysis, in order to evaluate whether we did a good job or not, we might need to do a long set of studies with econometric controls and other statistical methods. And we love doing those studies, and I hope there will be a lot of new scholarship about AI and the environment. But I do think this is going to be easier to know if we got it right. We will be able to see this with our own eyes. And in 5 years or 10 years, if we get this right, electricity bills will be more affordable, especially for low- and moderate-income households that carry a lot of energy burden right now.

I also think we would see clean electricity deployment accelerate rather than stall. We would see data centers become flexible grid participants and not just connected demand centers that run all the time. We would see benefits in host communities—more tax revenue, cleaner infrastructure, better public services. And, for the climate, we would see emissions keep falling, as the energy system continues to shift from fossil-based to clean sources of energy.

But if we get it wrong, the signs will definitely be obvious. We will see pollution. We will see people getting sick from the pollution that might be generated from diesel generators burning on site at a data center. We will see electricity bills rising. We will see infrastructure costs rising because the benefits will accrue to companies and not to communities. We might see new fossil fuel generation that gets reinvested in and locked in for many more decades.

And I think the biggest thing that we will feel is a lack of trust. I think communities will see that this boom is not for them. And I think that that is a broader challenge and risk to both our climate, our competitiveness, and our national security objectives. So, growth alone is not what I would expect as what we would see as success. I would say success is seeing growth that is clean, reliable, affordable, and broadly shared. And that is the type of test that we can apply to see if growth is equitable.

Strom: That is music to our ears here at Equitable Growth. Well, thank you so much, Costa. This was a great conversation. I really appreciated the opportunity to talk with you about this.

Samaras: It was so great to be with you.