Powering Progress Episode 12 – Rethinking the Grid for AI

Aubrey Bishai [00:00:09]

Welcome back to Powering Progress, a podcast series exploring the infrastructure, energy, and market dynamics shaping the next generation of data centers and AI development.

I’m your host, Aubrey Bishai, Chief Innovation Officer at Vinson & Elkins.

We’ve talked throughout this series about the extraordinary amount of power required to support the growth of AI and data centers.

But having demand for power and actually getting that power to a data center are two very different things.

The regulatory framework governing the electric grid was built for a very different era.

Today, developers and utilities are confronting enormous new loads, compressed development timelines, constrained transmission systems, and evolving rules about who can connect to the grid, where the power will come from, and even who can sell it.

Joining me today is Jeff Jakubiak, a partner in our energy regulatory practice, to help us make sense of this changing landscape.

We’ll talk about FERC’s evolving role, the choices between buying and building generation, different approaches to connecting data centers to the grid, and why regulatory policy may ultimately help determine which states and regions capture the next wave of AI-driven investment.

Jeff, great to have you on the show today.

Jeff Jakubiak [00:01:48]

Thanks for having me. Happy to be here.

Aubrey Bishai [00:01:50]

Let’s start with the big picture. Data centers are creating loads at a scale and speed the electric system hasn’t traditionally had to accommodate. What makes this moment so different from previous periods of load growth?

Jeff Jakubiak [00:02:06]

I’d say there are three things. Scale and speed, as you note in your question, are two of it. And also the third is the state of the grid.

From the scale standpoint, the loads are just simply larger than we’ve ever seen before. People are talking about data centers of 1,000 megawatts, just one GW, which put in context is about the load of a million-person city, and frankly double the load of anything we’ve seen in the United States so far. The largest loads in the United States to date are historically aluminum mills or maybe some mines, and those are the four to 500 megawatt range. So a load at a single location of 1,000 megawatts is really just unprecedented, to be blunt.

And as to the second issue, the speed — the data centers and the data center developers want these up and running within 18 months, which is at a speed we’ve never seen before. To build cities, of course, takes years and decades, and to build even the copper mines and whatnot, it takes many years. So the people developing these projects and the grid operators who have to oversee how they integrate with the rest of the system have to move at a truly lightning pace.

And layered on all of this is that it’s coming at a time in which the electric grid is frankly aging out. There was a massive buildout of the electric grid after World War II until the early ’70s, with the growth in industrialization in the United States, the growth of consumer goods, washing machines, et cetera, that drove the growth of the grid and the need for power. And all that infrastructure is aging out. So the grid is very, very stressed and is in need of replacement. And this growth in load from data centers is coming at a time that frankly couldn’t be worse.

Aubrey Bishai [00:03:54]

For a company looking to develop a data center, I’ve heard you break the regulatory challenge into three basic questions. How do you connect it? Where does the power come from? And who can sell that power? Why is each one of these questions more complicated than it might sound?

Jeff Jakubiak [00:04:13]

Each of these issues is a combination of law and engineering in some respect. On the connecting part of it, there are questions of where to connect and how to connect. To drill down a little more, the where is very important, because the data centers need to be where there is power to supply, or where you can build power to supply the data center. I’ll get to that in a second. And in terms of where to supply the power, you need resources, whether that be wind or sun or natural gas pipelines or coal, if you’re going to go old school.

And the connection is not necessarily readily apparent. There are entities that want to connect to the grid to have the grid supply the power, or even if they have their own generation, to have the grid as a backup. And there are other entities that want to go in what we’re calling an island mode, in which they have the generation paired with the data center or load, but it’s not connected to the rest of the grid. So that connection question alone has a lot of permutations, options.

And then relatedly, there is the second issue of the source of the power. As I noted, you could buy power from the grid or you could build generation, but the answer of which to choose isn’t readily apparent either. Much of the power on the grid is already spoken for and already being utilized, particularly at the sizes that we’re talking about that the data centers need. The generation just isn’t there like it was maybe 15, 20 years ago. And so, yes, there are places where you might be able to find power and buy it from the grid, but a lot of the time you’re going to have to generate. And that also has to factor into the decisions of where and how to build.

And then thirdly, there’s an issue of selling the power. It may seem to some people that if you build a generating facility and a data center, you can just have the generating facility sell to the data center, but it’s not that simple. In a lot of the United States, utilities still have a retail franchise, which is an exclusive right to make a retail sale. And in those instances, the generation may have to sell to the local utility, which then in turn sells to the data center. So there’s this regulatory overlay that can complicate things in certain parts of the country. And all three of these issues have to be worked on and managed in tandem in order to get something to fruition.

Aubrey Bishai [00:06:32]

Certainly one of the biggest challenges is timing. Data center developers want power quickly, while new generation and transmission can take years. How is that mismatch changing the way projects are being planned and the choices companies are making between buying power and building their own generation?

Jeff Jakubiak [00:06:54]

Well, you’re exactly right on the timing issue. The data center developers, as I mentioned, want something up and running within 18 months. It may be possible to have generation up and running in that time, although to be blunt, it all depends on whether you can get your hands on generation resources. There’s a long lead time for new generation just to purchase the equipment, and it’s getting longer. So in some instances, you can buy either used equipment or get in the front of the queue and get it up and running, but that’s already a challenge.

The transmission, though, is really the temporal pole in the tent, usually. It takes years for the engineers to determine what transmission has to be built to accommodate the new data center load and generation, and then construct it. And there’s a lot of demand on the system that’s already in place wanting to build this transmission.

Utilities are by their nature conservative entities, understandably so. If there is a problem on the grid, the lights go out, and nobody wants to be in a position where they allow new load or new generation to interconnect, and not study it correctly, and then the grid goes down and you have a massive blackout. So the utilities understandably take their time to study these, study it correctly, and on top of that, it just takes a while to build transmission.

So one of the solutions, and I alluded to this before, is islanding. A lot of entities are now trying to avoid the long lead time for transmission by islanding their generation with their load and not even connecting to the grid, or more likely these days, starting the situation with a plan to connect later, so then they can hedge and buy either from the onsite generation or from the grid. And that seems to be the way the industry is converging around how to get speed to power and get these things built.

Aubrey Bishai [00:08:51]

Jeff, you’ve argued that regulators increasingly need to think about the grid not simply as a utility system, but as an economic development platform. At the same time, they have to protect existing customers from unnecessary costs. How should regulators balance those interests? And what could those decisions mean for where future data center investment goes?

Jeff Jakubiak [00:09:20]

I appreciate you asking that. This is one of the issues facing the industry — how regulators deal essentially with this new world order and this new environment of the data centers. Just like the utilities and the developers are facing unprecedented situations in terms of demand required and the time in which people want it built, the regulators are watching this all and trying to figure out how to balance their need to protect consumers and protect grid integrity with making sure that things actually get built and that the utilities operate as they’re intended to.

I have come to believe that the utilities can build resources in anticipation of the demand more than they’re already doing. The only way that I see that the demand is going to be met in anything close to the level that we’re seeing called for is if the utilities get in front of the wave and start building the infrastructure before it is requested.

Historically, utilities have operated on a reactionary model. A big customer comes to them, says they want to site a new industrial facility, the utility studies it, it builds it. And that’s frankly how it’s still going on now with data centers, but that takes a long time. It’s a long process. I believe that if utilities start building, the load will come and it will be utilized, and the utilities can compete with one another to attract load in this manner.

But utilities, by their nature, are financially conservative — historically been called the stocks of widows and orphans, as we know. And that is because they are guaranteed cost recovery, as permitted by state and federal law. And the states, historically, which regulate the retail service, haven’t been willing to give utilities essentially free rein to build things in what might be considered a speculative environment. The regulators understandably don’t want to be seen blessing what may become a bridge to nowhere.

But I think the time has come for the utility commissions to start thinking a little differently, and stop thinking about themselves as merely restrictors on costs and entities that are charged with keeping down costs and prices, and to think of themselves as engines of change and economic development. I think that if the utility commissions recognize a role, or adopt a role, as a broader engine of economic development, they could encourage the utilities to build in anticipation of load to a certain extent, attract the load, which would bring tax dollars.

In a sense, it’s no different than municipalities either giving tax breaks to stadiums or even issuing bonds in the municipality’s name in order to fund stadiums. Those situations put risk on taxpayers, but with a recognition of an economic benefit on the back end. And utilities and state utility commissions could do the same thing. And I think that is the best way to see the development at the speed we need to see it.

Aubrey Bishai [00:12:27]

Clearly, there’s a lot that’s still evolving. If you’re a developer, investor, utility, or hyperscaler planning projects today, what should you be doing now to make long-term decisions when the regulatory framework itself is still taking shape?

Jeff Jakubiak [00:12:47]

Your question contains a very good point that is worth emphasizing, that the regulatory framework is still taking shape. And that’s true at both the federal and state levels. There are multiple proceedings at the federal level, at the state levels as well, in which regulators are trying to figure out how to deal with these issues — how to get the load built while maintaining grid integrity, while ensuring that the people or entities that benefit pay, and that there aren’t ancillary knock-on effects that they don’t want.

Any entity that’s interested in this space should get involved in that. These proceedings are still very active, though they are moving quickly. And entities who are looking in the space should get involved, should comment, or at the very least follow what’s going on.

That said, I think there is a trend in the industry to what we are calling bring-your-own-generation. As I mentioned, generation is in short supply, and certainly there is power available off the grid or from existing generators that can supply the data centers, but it’s getting fewer and further between. And the rights of a new large load like a data center to take power off the grid are being looked at in all these proceedings.

The trend, though, is that if you bring your own generation, the world may not be your oyster, but it will certainly give you a leg up and give you more rights to power and firm supply of energy than you otherwise would if you were just pulling power off the grid. The generation is really the ticket to the party now, as I see it. And anybody looking into the space should very quickly scope out where they’re going to secure new or existing power supply, as I see that being the scarce resource that’s going to determine where, how, and when all this new load can come online.

Aubrey Bishai [00:14:36]

Jeff, thank you so much for these insights today. I appreciate you being on the show.

Jeff Jakubiak [00:14:42]

You’re very welcome. I appreciate being here. Thanks so much.

Aubrey Bishai [00:14:45]

And thank you all for joining us on Powering Progress. To learn more about the insights on AI infrastructure and the legal frameworks behind it, visit us at velaw.com.