Texas' Load Growth Challenges – And Opportunities, with Arushi Sharma Frank

Large load growth is one of the most urgent challenges and opportunities for the grid.

From data centers to factories to electrified oil and gas operations, demand on the ERCOT grid is expected to soar. If we don’t load forecasting and planning right, we risk higher costs, more outages, and a weaker grid. If we do it right, we can build a more resilient, affordable, and flexible system for decades to come.

On this episode of the Energy Capital Podcast, I’m joined by Arushi Sharma Frank, energy policy expert, consultant, and Substack writer, for a wide-ranging discussion about how Texas and other states can rise to meet this moment.

Arushi breaks down the four pillars of large load flexibility that can help avoid overbuilding, protect reliability, and strengthen the grid:

* Defensive Flexibility: Making sure large loads can ramp up and down smoothly without destabilizing the grid.

* Planning and Interconnection Flexibility: Allowing partial firm interconnections so that utilities can overbook capacity more efficiently, just like we already do with generation.

* Emergency Flexibility: Enabling operators to call on large loads for grid support during emergencies through dispatchable programs like Controllable Load Resources (CLR).

* Self-Limiting Interconnection Rights: Contractually capping how much a load can pull from the grid, with the load providing its own on-site generation or storage to fill the gap.

We also talk about why large load forecasting is critical and why the lack of accurate forecasting could compress maintenance windows, strain natural gas plants, and create reliability risks year-round.

Another major theme: cost allocation and fairness. Today, large industrial loads have far more tools to lower their transmission costs than residential customers do. As Arushi explains, unless we reform policies like Four Coincident Peak (4CP) rules, residential ratepayers will continue shouldering a disproportionate share of grid costs.

We also dive into:

* How House Bill 3970 and Senate Bill 6 at the Texas Legislature aim to manage large load interconnections. Plus, how combining the best parts of both bills could create a smarter, faster, fairer system.

* Why Texas’s "connect and manage" approach for generation could be adapted to load, allowing faster connections without unnecessary overbuild.

* How large data center loads could potentially fund demand-side reductions, like HVAC upgrades and home weatherization, that make the grid stronger and bills lower for everyone.

The stakes could not be higher. As we discuss, this isn’t just about transmission lines or interconnection queues. It’s about how we operate the entire system (generation, transmission, distribution, and demand side) as a coherent whole.

I wrote more about this systemic view in How Load Flexibility Could Unlock Energy Abundance.

If we get it right, we can build an energy system that supports economic growth, national security, and affordability for all Texans. If we get it wrong, we risk bottlenecks, blackouts, and runaway costs.

This conversation is one of the most technical we’ve done on the podcast, but it’s also one of the most important. Let me know in the comments if there are any concepts here that need additional explanation.

Timestamps

* 00:00 – Introduction

* 04:00 – Electric grids are vastly underutilized

* 08:00 – Managing the peak is crucial, with load flexibility we could have higher capacity factors and lower peaks

* 14:00 – Compressed maintenance windows for large outage periods for thermal plants, getting load forecasting right is absolutely necessary

* 18:00 – Challenges to transmission & distribution utilities in the high load growth environment

* 25:00 – The four pillars of large load flexibility, including grid forming capabilities and frequency support from large loads (very timely post-Spanish blackout)

* 32:00 – Connect and manage approach for generators could apply to loads

* 34:00 – continuation of four pillars: emergency flexibility, again includes discussion of instantaneous control to restore frequency

* 40:00 – SB 6 vs. HB 3970: Legislative proposals currently under consideration

* 42:00 – Prioritizing projects based on the shared benefits those projects bring

* 47:00 – Speculators “squatting” on prime spots on the grid

* 50:00 – Lack of policy structure is holding back signals to data center investors to bring shared benefits to the grid; HB 3970 checks most of the boxes

* 54:00 – Reasons why residential and small commercial bills are going higher

* 1:01:00 – Data centers without flexibility could purchase demand reductions from residential customers: reduce peak and help affordability

* 1:06:00 – Final thoughts & where to find Arushi

Show Notes

Resources

Arushi Sharma Frank

* Teach What I Know (Substack) — Arushi’s writing on energy policy, flexibility, and grid modernization.

* Article: Flexibility Compact

* Article: How Texas Wins the AI Race

* Article: Four Pillars of Flexibility

* Article: What’s Actually in the Data Center Flexibility Toolbox?

* Article: Cost of Service Ratemaking Explained

* Arushi Sharma Frank on LinkedIn, Twitter

* CSIS Profile: Senior Associate, Energy Security and Climate Change Program

* LinkedIn: Bill Scorecard - Two Bills in Texas for Data Center Speed to Power

* LinkedIn: Breaking Down the Barriers to Flex IX Study

* Twitter: Stuff Utilities are Saying

Large Load Growth, Flexibility, and Planning

* How Load Flexibility Could Unlock Energy Abundance, Energy Capital Podcast with Tyler Norris – A deeper dive into how flexible loads like data centers can help integrate more renewables and improve reliability.

* The Name of the Game is Flexibility: Energy Capital Podcast with ERCOT's Pablo Vegas

* Large Load Task Force / Large Load Working Group (ERCOT) – The ERCOT group tracking the interconnection of large loads like data centers, and providing updates on projects in development.

* AI Compute Workload Orchestration (ERCOT)

* Report: DOE SEAB

* Report: Flexibility DC-Gen Colocation in FERC Markets (Enchanted Rock)

* Report: Understanding Non-Firm Connection Dynamics

* Slide Deck: Scale, Speed, and Supply Chain - Success Factors for Sustainable Electrical Energy Delivery (Quanta)

Flexible Interconnection Concepts

* Connect and Manage Framework for Generators (ERCOT) – How ERCOT allows generation to interconnect faster by accepting curtailment risk, a model that could be applied to large loads.

* House Bill 3970 (Texas Legislature) – Proposed legislation to enable faster interconnection for large flexible loads.

* Senate Bill 6 (Texas Legislature) – Proposed legislation focused on cost allocation, ratepayer protections, and emergency interconnection requirements for large loads.

* EPRI DC-FLEX Initiative – Research initiative working with utilities to pilot flexible load integration at large data center campuses.

* Press Release of Initiative

* ERCOT Guide to the Interconnection Process

* ERCOT Four Coincident Peak Calculations

Transcript

Doug Lewin (00:03.682)

Welcome to the Energy Capital Podcast. I'm your host, Doug Lewin. My guest today is Arushi Sharma Frank, the first returning guest of the pod. I interviewed her last year. We mostly talked about distributed energy resources. I highly encourage you to listen to that episode. It's one of the most listened episodes of the show ever, and it's a great one. On this one, we're talking about something a little different. We're going to talk about how we manage massive new load growth, especially from AI and data centers, while still improving reliability and affordability. This one builds really nicely on the episode with Tyler Norris I did just a couple weeks ago. So if you haven't heard that one, you could listen to that either before or after, probably before this one is probably a little better.

This one really kind of builds on that. Arushi explains here four pillars of large load flexibility and how they can be incorporated into grid planning. We also talked about legislation at the Texas Legislature as it enters its final month, especially House Bill 3970 and Senate Bill 6, two bills specifically dealing with the challenges of integrating large load. And we talked about how combining the best ideas in both of those bills could create a really good blueprint for Texas and ERCOT to successfully integrate large loads into the grid. Toward the end, we explore sort of a new and big idea. Could large data centers help fund efficiency upgrades for Texans, including low-income Texans, which would have the dual benefit of reducing peak demand and thus freeing up more headroom on the grid and reducing energy poverty in a state where 30 to 40% of folks struggle to pay their energy bills.

Arushi really brings clarity to some of these most complex debates in Texas energy policy. She has a wealth of experience having worked for the Energy Power Supply Association or EPSA, the American Gas Association, which is AGA, Tesla, and Exelon. She now runs Luminary Strategies, her own consulting firm, and writes a Substack under the title, "Teach What I Know." You can find that in the show notes and on my Substack recommendations page. As always, please like, share, leave a quick review, particularly a five-star review that really helps us keep this pod growing in the community of people growing who are trying to get energy policy right. Thanks so much for listening. Here's Arushi Sharma Frank.

Arushi Sharma Frank, welcome back to the Energy Capital Podcast.

Arushi Sharma Frank (02:29.474)

Thank you, Doug. I'm happy to be back.

Doug Lewin (02:31.842)

The last time we talked, I think there were at least five or six times in the last 10 minutes of that conversation when I said I need to have you back on the pod. So it's taken a while, but here we are. Thanks for coming back. I do encourage folks to listen to that first one. We talked a whole lot about distributed energy resources, distributed generation, that sort of thing. Today we're going to have a little bit of a different discussion. So Arushi, if you kind of want to start by filling people in on some of what you're doing, and I'll tee this up for our listeners. We're really going to get into talking about large loads and large load flexibility and large load interconnection. This is to me, Arushi, I think if we weren't arguing with all of these, I don't know how else to say it, let's be generous. With all these bills that I think are misdirected and creating a whole lot of distractions, people that are trying to punch down at renewables, which is taking a whole lot of effort and causing a lot of distraction.

If we weren't doing all that, this would probably be the biggest issue because we've got this load growth coming. It is challenging, sort of like integrating variable renewables. It's challenging, no doubt about it. It also gives you opportunity to make your system stronger if you do it right. So excited to get into that with you, but give everybody a little bit of an update. What are you working on? And then let's jump into talking about large loads.

Arushi Sharma Frank (03:51.448)

So since I left Tesla, I've been working a lot with clients and in the academic space on figuring out a number of different ways to utilize the grid we already have. So our last pod, we talked a lot about grid edge. And the reason we talked about it is frankly the reason we're talking about data centers today in the same vein. It's about figuring out who is going to motivate capital, which funds innovation, and where is it going to go to solve a problem that exists today. For us to be energy dominant, AI dominant, tech dominant, innovation dominant in this country. And once it's there, how are we going to deploy it to utilize the grid we have efficiently, smartly and better?

I was at Northwestern last week and a senior official from Commonwealth Edison in Illinois was delivering a keynote and reminded us that up in Illinois, the ComEd grid is utilized around a 55% load factor or capacity factor. And from Tyler Norris' pod with you, Doug, people know what that is now.

Doug Lewin (05:00.266)

And it's even less in the ERCOT grid, right? We're at like 53%.

Arushi Sharma Frank (05:03.922)

And remember, like every grid in the country, and not necessarily by grid, occasionally by state, has a peak that's designed for a different winter or summer weather condition and a certain duration. Really interesting example of this that I think resonates a lot with folks to really understand the scale of the issue with underutilized grid capacity is comparing states that are right next to each other. So Wyoming and Utah, two Rocky Mountain Power Berkshire Hathaway utility families. They're grids next to each other. They shape power across those two states, but their peaks are different. One's a winter peak state and the other is a June, July state. And when you look at ERCOT or you look at any other kind of single state grid, you do remember that at least in these regions, you're solving for a singularity.

So that's why I think ERCOT ultimately, much like New York, to a degree much like California. I think our single state grids, because they're designed for a net peak that is motivated by power sharing across that specific one grid, is really ripe for distributed energy solutions. It's ripe for what we're going to talk about today in terms of maximizing the efficiency of how we build data center load, now we're talking like high power GPU compute load, in a way that serves all of these objectives together. And it all starts with where are you going to put the capital? And then who's going to tell you to put it there, right? Are you going to do it yourself? Or is a regulatory program or a piece of legislation going to provide the signal to do that? Or is it going to be a URI type scenario, which for me, the UPP AIDR project was, which is it was catalyzed by the loss of money and people in Texas. And obviously now on the other side of that, doing this work in private practice and doing it as an academic who gets to talk with regulators who are trying to design tariffs and policies all around the country. It's almost like this has to be the one place I want to work. Where my work is not utilizing a crisis that's already here, it's getting there ahead of time and building the correct price signal to be in a state where we're motivating progress, not crisis.

Doug Lewin (07:25.902)

I love it. Yeah. I mean, that really is a good encapsulation of what we're working on. Like we could be in a paradigm where we're just managing crises and that is not necessary. There is a different way. So let's kind of start with, why are we having this conversation? Why is this important? In one of your blogs and your writing, both on LinkedIn and on Substack, I can't remember which one I pulled this quote from, but you quoted the CEO of Constellation, if I got this right off your website, hopefully I did, Joe Dominguez. Who said, I'm going to just read this quote, it's a short one, but "data centers are integral to our daily lives, economy, and national security. Our energy system is built to handle the extreme demands of our hottest summer days and coldest winter nights, but is often underutilized. The real challenge isn't a lack of energy for data centers, but managing the peak demand hours. The ability of data centers to flex during these critical periods is crucial." So obviously this is something like you just mentioned.

I talked a lot with Tyler Norris, but honestly, I don't think we can talk about it enough. We have plenty of capacity, 98, maybe 99, maybe 99.5% of the hours out of the year. And really one of the things I don't think we talk enough about, Arushi, is the difference between megawatts and megawatt hours. People say in Texas we're going have a 75% increase in power in the next five years, or there's going to be a tripling in the next 10 to 15 years. Are we talking about peak or are we talking about energy consumption? Because those are different things. And we could double or triple our energy consumption and have our peak not rise by nearly that much. That's good for all the generators on the system and would allow us to integrate these loads, which as Mr. Dominguez says are integral to our economy and our national security. To me, that's the why. Would you build on that? Why is this question so important for legislators, regulators, grid operators, stakeholders, and the general public?

Arushi Sharma Frank (09:28.758)

No. So first of all, have to give a nod to Joe. Former constellation, myself before the spinoff. And I worked in the compliance group out in Baltimore on power plant compliance in multiple markets. So what that enables is the sort of depth that frankly, no policymaker has the time for. So it's really important to elevate that. It's really important that people like Joe Dominguez go out there and say the one sentence. Because even that itself is a revelation for a lot of people. And if you're in the space of being in the seat of making policy, and you already have 38 other problems to solve for, you kind of get to a point where you're sort of impatient and need to solve a big problem. Like, what is this? So it helps.

Doug Lewin (10:17.612)

If I may, right? Yes, there's 30 issues. There's probably more like 130 issues. And in Texas, they got 140 days every two years to work through all of those issues. It's insane. It's no way to run a state, but that's the way we do it, which means their time is very, very short. So it is important that, and of course, Constellation's a big player in the state of Texas. Yeah, go on.

Arushi Sharma Frank (10:40.482)

Yeah. Right. So spokespeople matter. Studies matter. The next piece of this, which I work on, and that's what we're going to talk about, and that's how I'm going to answer your question, is to truly understand the binding constraints for the power grid, the people who operate and run and study that grid, and the binding constraints on the entities that want the power. And following that statement that data centers can and should flex, we're going to get into what those constraints are and feel free to challenge me on them or help me help you understand it better so our audience gets these better.

But I want to start with this idea of why it's important to get this right. The criticality of it. Conventionally, when we build in a conventional vertically integrated monopoly area, when we build a new pole, a new wire or a new power plant, this is before the world of deregulation. We kind of have this cost plus model, the cost of building it plus some regulated return on top of it. You have the thing, it's there. And the reason you get that money to build it, the reason you can raise to build it and then recover the plus on top of the cost of the asset is this idea that you are building for the moment in time when everybody is using the seventh lane on an eight lane highway at the exact same time, right?

Doug Lewin (12:06.422)

Or the eighth lane, right? Or that. And maybe pushing into the ninth lane.

Arushi Sharma Frank (12:10.062)

Yeah, were there to be an eighth lane. The reason I'm not using the eighth lane is that I think grid edge EV CHP for CP congestion management all of the conventional tools we already have for grid flex that are baked into both economic and reliability programs, they're all living in another of the eighth lane, right? They're there. That's why I'm saying number seven.

Doug Lewin (12:35.308)

That makes sense. If we take the analogy further, and perhaps we shouldn't, but there's almost like a shoulder, right? Which is really like the reserve margin that's built in. I get what you're saying. You're not going to go into that eighth lane because that's your reserve margin. You have to keep some. That makes sense.

Arushi Sharma Frank (12:50.008)

So going back into why it's so important, there was testimony from ERCOT CEO, Pablo Vegas, a couple of weeks ago at a house hearing. And I don't really follow politics much, but I saw this clip and it was describing the fundamentals of the forecast. And there is right now disagreement between a utility forecast and an ERCOT forecast, right? And for me, the issue to bring out the importance of getting data center interconnection, load growth forecasting right comes down to this one thing is that we will have to force ourselves into the eighth lane which is our headroom, our reserve margin, the period of time in which we put all of our natural gas plants on outage or maintenance, those shoulders, we make the shoulders smaller and smaller and smaller.

The simplest way I have explained this in some of my writing too is that not getting the load forecast right for these large loads means that ERCOT has to plan for the N plus one or the redundancy system such that they have to assume a enormously constrained reliability metric for the grid all the time.

And that means every shoulder season, right? So the shoulder season, like the non-peak seasons, like before the summer starts, before the winter starts gets compressed. And the windows that ERCOT has to give permission to our generators, our seal in the ground, right? Our base load to take their outages and be ready to run for the next storm Elliott or Uri gets smaller and smaller and smaller. So hear that again. Not getting load growth planning, forecasting and flexibility right is a material risk to the daily seasonal and emergency operation capacity of the Texas natural gas fleet period.

Doug Lewin (14:48.01)

Yep.

Arushi Sharma Frank (14:50.42)

This is it. This is it. And then, of course, when you look at planning in the context outside of just Texas, the way that we've done interregional planning, the other big theme here is that transmission planning is a proactive thought process. It is the result of fairly sophisticated generator interconnection rules, regulation, transparency, and design. In the generation space, as much as there is noise around the cues are slow and all these dates and there's lots of clustering problems with restudying and there's squatting and we're trying to fix all these problems with generators. The reality is that the sophistication we have in planning for this redundancy and keeping shoulders open so other stuff can happen and we're not overpanicked about a reliability event because we understand the actual load that's coming or the actual generation set that's coming. It's dependent on having a very accurate planning picture for a generator.

So that doesn't exist anywhere in the country today for load. You don't have to take my word for it, but you could take DOE's word for it, or you could talk to the utility executives that are supporting every DCFlex initiative that we'll get into little more detail. And we all know collectively as an industry, as a power industry, that load connection to local power grids has always been a more bespoke process. It is non-transparent in design. It is a first-come, first-served process from the perspective of connecting sites when you know they're basically ready. And we've always had a lot more certainty with this type of load connection because typically we're not connecting 200 shopping campuses and OXIE facilities at the same time. So when people think, you know, they say like, well, load forecasting has always been a thing we've done. Like, yes, it is, but we never had the occasion anywhere in the country to be approached by, you know, say Tesla and Tesla saying we're building about 18 gigafactories at one go at this spot. It's one. And so these anchor tenants, if you will, that help a utility area plan a load forecast, there's been discrete tenants that kind of like set the full picture and provide them the wiggle room to understand how to connect incremental load and then work their brain as planner towards the outcome they know they need for a redundant and reliable transmission system. So EPRI...

Doug Lewin (17:28.138)

All that thought on EPRI, I do want to hear about DCFlex and what they're doing. I just want to put a finer point on this to your point of like what grid planners have been dealing with is sure there's large loads, but it's one here, two there. In the large load update given to the technical advisory committee at ERCOT TAC, the group that is one level below the ERCOT board, the large load task force, which is now becoming the large load working group. LLWG, I believe is the new acronym because we don't have enough acronyms. They gave a presentation just yesterday. We're recording on Thursday, April 24th. It was on April 23rd. They have a new slide. I'll put it in the show notes so people can see it. It is the Large Load Project Distribution. They are, and not all these will get built. There are 50 projects of a thousand megawatts or more looking to connect to the ERCOT grid. 50, 5.0, that are a thousand megawatts, to put that in perspective for folks that don't do this kind of work, or I do try to make this, I always like to make this accessible to folks that aren't in the industry.

Let's put that in perspective. The city of Waco, the city of Lubbock are around seven or 800 megawatts. There are 50 different loads looking to interconnect that are larger than Waco. So you start to get a sense of the enormity of the challenge that the grid operators in the utilities are facing and why what you're talking about is so critical. In the exchange, I believe you're talking about that it was state affairs where Ocasio Pablo Vegas was talking with representative Charlie Garan and Garan is saying, we need to build a grid that is for all these projects. I'm paraphrasing him as basically in Pablo kind of looked at him like, if we did that, the cost would be enormous because not all of these loads are going to come in. The ones that do come, if we know, and this is where I know the conversation is heading.

So you bring us back to EPRI DCFLEX or whatever you're going to say, but this is, I think, where I know this is where we're headed, is if the grid operator in the utility know that these loads have certain attributes, potential for flexibility. Whether that's because of the nature of the load behind that meter and its own flexibility, whether that's because they have storage or gas generation or whatever it is on site that they can reduce from maybe that thousand, those 50, 1000 megawatt loads, say 20 of them get built and each of them could get down to 500, right? They could be at 50% of their capacity. This is critical that we don't overbuild the system, spend way too much rate payer money on the transmission distribution side and on the generation side, to your point, that we're giving the gas generators, the coal, nuclear, the ability to go offline during those maintenance seasons, take the maintenance they desperately need to be ready for the winter and the summer peak seasons.

Arushi Sharma Frank (20:22.478)

You know, and this is not a fun and games for our poles and wires utilities either, right? Wherever you are in the country as a utility, whether or not you own generation or you contract it, procure it, whether or not you serve load. Poles and wires companies also do maintenance and they also plan OPEX. There's another Substack we'll drop explaining the essentials of cost of service rate making. And the theme there to be very, very sensitive to, which is why this is also important, is that whether you like it or not, and I mean, you, generically, not you, Doug, because I know what your answer will be, we have predominantly a cost plus or a cost of service business model for poles and wires in the US, right? So that means that every time you go to go raise capital, if you're a utility, you raise debt to fund your capital planning project, and your capital planning project has the super HV transformer times 35,000 because your load forecast says that. A, the cost of debt for you as a utility is atrocious. It makes no sense. Anytime your rating agency sees that they're going to question whether or not your utility commission would ever approve such a thing. When the answer is no, of course the rating agency will downgrade you, which will raise the cost of your debt.

Then you look silly in front of your regulator and then you generate everything else that comes next, which is consumer ire, rate payer pressure, and just sheer confusion around where is really the best place to spend a marginal dollar that has a rate of return tied to it. To add to that, this sort of forecast then also creates another pressure on utilities, which is the OPEX problem. Trucks, people, maintenance, sensors, software, all this stuff is not part of the CapEx book that has a rate of return tied to it. It's part of just operating the darn enterprise. And if you look at any private sector competitive enterprise, the way that you would set it up is to not run things to failure and order lots of things and hope that you'll use them, is to use what you have well and efficiently and sensibly in light of the actual demand on your equipment and your capex. And you'll put your optics budget in the right spot in that framework so you're maintaining this stuff that you already have. So you don't need to go get more of it. So for utilities too, this is a huge issue inherently and how they actually hang on to a business model that conventionally has a lot of safety on the cost recovery side.

Maybe utilities might disagree with me on that, but obviously they're not a competitive business. But on the other side, it's like their entire way of procuring assets, right? The planning side is like, when am going to order the next 23 large transformers? This problem massively discombobulates the CFOs. A great resource we'll drop in the chat on this is a plenary session presentation done by Quantas Services, which does a lot of contracting work with and for utilities. They presented at ECIG in Austin just a few weeks ago and mentioned that even utility procurement practices are going to have to change dramatically, which puts them as procurement entities in the spot of making speculative buys for this equipment, which is this crazy thing no CFO in their right mind, private or public sector, wants to do.

Doug Lewin (24:04.314)

So I want to come back to the to the eight lane highway and how really what we're trying to do is make sure that you know if we have to add a ninth, make sure we really need that ninth and not add a tenth and 11th. But also make sure that the fourth and fifth and sixth lanes are getting used more in order to meet that goal to keep that peak in megawatts down while we bring up the megawatt hours, the actual energy consumption. I want to ask you this, and I know you've written about this a lot. What are the kinds of flexibility that would actually help grid operators? I want get into the nitty gritty here, not the, right, I know that's what you do. This is what your blog is so good for. There's a lot of general discussion around this, but specifically, what are the kinds of flexibility that would help grid operators?

Arushi Sharma Frank (24:57.784)

Sure. So I've categorized it into four pillars in my taxonomy. And I think you'll start to see more reports come out that have similar ways of describing this work and these concepts. The first is defensive flexibility. Defensive flexibility means that your site, whether it's just a colo, so like some kind of power producer, and the load together, a consumer, together, if it's a colo or whether it's just a load, you have the ability as that connected site.

Doug Lewin (25:28.462)

So, co-lo is short for...

Arushi Sharma Frank (25:31.502) Co-located generator and data center.

Doug Lewin (25:34.19) Okay, I have not heard anybody say co-lo yet. I like it. Okay, we can start using.

Arushi Sharma Frank (25:40.582) I

t's very much a substation, EPC, planning, construction vocabulary term. I like it. So yeah, you'll learn new...

Doug Lewin (25:49.518)

I've been saying it the long clunky way. I just keep saying co-location. All right, I gotta change my, update my lingo. Co-lo, go on.

Arushi Sharma Frank (25:58.136)

So all right, so what is this? What is this? This is all the stuff you hear about the most in terms of instantaneous reliability issues, right through ungraceful behavior by the data centers and whatever is paired with them. So that's defensive flexibility, right? So whether or not there's legislation, a grid operator needs to make sure that if you're going to get up from the grid and go take a bath and come back later, that you don't take the entire bathtub with you, right, the entire electric grid. So if we can't solve for this type of flexibility, you can't have data centers period. So this is not one of those flexibilities that is riding on the idea that data centers have to be 99.99% firm and they need their big full power. That's not the issue. This is a different type of flexibility. It is you have to design these sites to be able to gracefully connect, disconnect, ramp, turn up, turn down from a grid without destroying everything, including your own gensets. So that's number one.

Doug Lewin (26:54.038)

And so on that, what we're talking about there is especially the larger the load is, the more this matters. If you are one of those 51,000 megawatt projects that are looking to locate on the Texas grid, if a 1000 megawatt load drops off, that is similar to if a 1000 megawatt generator drops off. If we lose one of our nuclear units, that causes frequency deviations, which can be very damaging.

So you have to make sure that they are capable behind that meter of if something happens, they don't just drop from 1,000 down to zero. There's some sort of, as you put it, a graceful ramp, I believe is how you describe that in your writing. Yeah. OK, cool. What's the next one? That's number one.

Arushi Sharma Frank (27:37.678)

So that's defensive, right? So you're defending the grid against a catastrophe and you're also defending your site against the mirrored impact of that kind of catastrophe. I want to mention one thing really quick here. NPRR 1100, which is the second NPRR I worked on before I worked on VPP. A piece of that that's very important to understand is that we have been working for a very long time in complex manufacturing and industrial process industries on achieving this kind of graceful ramp. The way we do it is UPS, uninterruptible power supply. The reason you need that is it's not just about the grid staying whole as large loads leave. It's also about the finnickiness that the load itself. If you've got stamping machines and aluminum robots, and my favorite at the Gigafactory and Tesla was Bumblebee, I didn't get to name them, but I contribute to the theme. There's things to be able to also ramp down slowly if there's an outage or a loss of power.

Doug Lewin (28:39.628)

Yeah. What is Bumblebee?

Arushi Sharma Frank (28:41.496)

Bumblebee, it's one of our robots at Gigafactory Texas.

Doug Lewin (28:44.554)

Got it. OK, sorry. Go on. OK, keep going. Yeah.

Arushi Sharma Frank (28:47.468)

Right? So data centers are not that different in their need for stable power from the kind of complex engineered load processes at machine and industrial campuses that do things like have robots make cars. Right? So the reason I mentioned the NPRR, it's important to understand that we've been trying to create mechanisms for uninterruptible power supply, flexibility, grid forming behavior for quite some time. the data center boom, I hope, will push that innovation to the next level, because now we're trying to solve the problem for quality of the grid and quality of continuing production or continuing computer, whatever it is simultaneously. Does that make sense? So then the second kind of flexibility is planning and interconnection.

Doug Lewin (29:36.536)

It does.

Arushi Sharma Frank (29:45.386)

What this means is enabling the utility company, the wires company, and the grid operator. And I separated them, Doug, because in some cases around the country, they are the same entity. And in others, they are different, right? Like the competitive areas of Texas, there's differentiation there. Enabling them to study only a portion of a large load as firm, and thereby enabling overbooking of the system. That is planning and interconnection flexibility.

So before you question like, the heck would you ever do that? That's how we connect generators in Texas, right? The connect and manage framework for generators in Texas is essentially overbooking transmission farm rights relative to the size of the generator. I had a great conversation with colleagues at EPRI earlier today actually, and put the thesis out that the vision of flexible load interconnection and some of the markets where EPRI will be doing these demo sites with partner utilities is to effectively achieve the same thing for loads, connect and manage. What does that mean? It means that there will be hours of the day of the month of the year where the load cannot do its full draw. And in those hours, the load can figure out whatever it wants to do, self-supply, back down, shift, turn off, but there will be hours where you will not get your firm load rights. The reason this is so revolutionary is that you take the connect and manage regime and apply it in a space where historically we have always had, we as in utility planners and grid power people had this idea when load comes, you just connect it, whatever, they pay the charges and you just do your thing. Like there's never been the need to have connect and manage as a regime for load. You just connect load.

And if you're a utility, you can decide if that load has a unique impact on your system and add, you know, an access charge of some kind. If it's really a problem to connect it that, you know, it has an outsized impact on the general rate base. So that is planning and interconnection together. And I have been talking about it with the connect and manage logic at the back of my head because it resonates so well with folks in Texas and outside of Texas as well. As you know, Texas connects generators the fastest because we have enabled the price signals to generators to do two things. Either be ready to curtail in exchange for getting connected faster or put your own stuff on your site. Invest in a demand response, invest in CHP, establish a pun, whatever it is. We've created the price signal for them to figure out that they can't get it all from the grid all the time in terms of giving themselves the highway to sell constantly to ERCOT. So enabling that on the load side is this type of flexibility.

Doug Lewin (32:39.224)

So, Arushi, this was actually one of the most revelatory things in your writings that I learned a lot from reading your stuff. But this one really stuck out to me because it seems so obvious in retrospect, right, we do have this connect and manage. This is something that Texas is pretty proud of. This is like the third time we're mentioning Tyler Norris, but Tyler had this really great exchange when he was in Congress. He was making a presentation and Congressman Randy Weber Republican from Houston was asking about Connect and Manage and he was saying ERCOT does it better. It's this point of pride that we have that we're bringing these generation resources on quicker. takes a year or two in Texas and five years in PJM or other places. And you're right, there's a framework for how that happens and it does include some curtailment. And now as these loads come on, we're going to be curtailing generation resources less, particularly if we cite the loads in the right place.

But it makes sense that you could potentially apply that principle. And so you're still going through the, are you calling them pillars or principles? You're still going through those. But I do want to bring up that there are active policy discussions going on at the legislature around this, particularly with House Bill 3970, which if I'm reading it right, is really kind of doing that. It's kind of putting a connect and manage to the loads, like, hey, here's a speed to interconnect, a speed to power if you're willing to have a few conditions and that could be reducing your demand, could be switching to standby power. Can you talk a little bit about, and if you want to go through the others and then come back to 3970 and Senate Bill 6 and what's going on at the ledge, that's fine too.

Arushi Sharma Frank (34:20.661)

Yeah, well the last two I think are easy, so I'll quickly run through them. Perfect. The next one is emergency flexibility. So that is the ability to simply drop load under operator discretion with verifiable control. The way that we did that in Texas today is CLRs, but CLRs, Controllable Load Resources, is a voluntary framework in which a load can register. It will appear as a dispatchable asset in ERCOT SCED, which is their engine for five minute dispatch. And CLR control is so refined that even if ERCOT gives a CLR a five minute instruction to behave a certain way and signal that it's time for you to start ramping. If ERCOT finds an instantaneous frequency problem, they can stop the ramp. So it really is total control. Really quickly, cause I know people who are listening to this pod less listen to the other one for the most part, maybe there'll be new audience this time. AIDR. The whole point of AIDR, the DPP program, is to do the same thing. It has become other things over time. It has become those things after I have left the leadership of the task force. But the point of AIDR was to turn power walls, end phase batteries, pick your favorite residential DER and turn it into an asset dispatchable in sketch.

So you can imagine for thinking about why these solutions have mirrors, it's not about looking at data centers as VPPs and all that like that. It's there are the peripheries of what's remotely possible and necessary. But the thing that you do need this emergency management, any grid operator that wants to manage an emergency is only going to rely on resources that it actually has some capacity to signal a control signal to. That doesn't mean that they show up and run the data center down for you, nor did they ramp a single power wall down. But because you're in sked, there's a tool for that. Whether you're in ERCOT or you're in outside of ERCOT and say, you know, salt river project or dominion or wherever there's going to need to be talking, right? Between a site controller, cloud, the operator of that utility system and its communication tools so that the site can respect the emergency signals that it gets and so that the site has advanced notice to be in a state to be able to deliver during an emergency. This is another one of those dogs that's like, it's like a non-negotiable flexibility issue. It really is. And if you don't want to deal with that as a load of any kind, you got to build it yourself. And when you build it yourself, you're also going to build your redundancy and that's really hard.

The last one type of flexibility is simply self-limiting interconnection right. That means you're going to contractually enforce at the point of interconnection a limit on what you can draw from the grid and you'll work with your utility partner or your grid operator or both to just fundamentally flatline what you're going to ask the grid for and everything else you're going to do to augment base load at your campus, you're going to do it a different way.

Doug Lewin (37:35.118)

Arushi would this be, so this would be a situation where maybe I'm a large data center. I'm 500 megawatts and the utility or ERCOT or the combination of them says, uh, this site, you really can't do more than like 350. And you say, but I need speed to power, can't wait for you to do years of upgrades. So I'll take the 350 you've got and I'll put enough power on-site combination of power sources that I'll always be able to cover 150. that what we're talking about or is it something different?

Arushi Sharma Frank (38:06.286)

Mm-hmm, exactly right.

Doug Lewin (38:24.738)

Got it. four things are, is it something like you don't need, I'm just trying to think through this question even makes sense or if it's just so obvious I shouldn't ask, but no, I don't think it is. So the four things are, it's not necessary that a data center have all of them or is it? Is it like you could have two or three out of four or do you need all four of those things to really sort of strengthen the grid as you enter it?

Arushi Sharma Frank (38:34.382)

You need the defensive one and then you need the emergency one because there's not going to be too many utility operators that are excited to put you on their system if you can't deliver on those. You're going need those, right? There are basic, basic operational capabilities. The others, which is planning and interconnection and then self-limiting