Jon Parrella  00:00

The grids can't operate fast enough to be able to respond to the volatility they're experiencing. They're creating a tremendous amount of harmonic distortion issues. And when you have that kind of power volatility, you can trip airbreakers at substations. It's reacting faster than even the generation assets can keep up with the fastest generation asset on the planet right now is natural gas. Is going to be a resip at about 25 to 30 seconds, we're seeing power volatility, where you're getting multiple 3040, 50, all the way up to 80% swings at the total size of the data center. So you talk a gigawatt scale, you could have 300 500 800 megawatt swings, multiple times per minute. They're breaking crankshafts on Gen sets. I mean, it's crazy. The new infrastructure requirements that are coming

Jon Parrella  00:42

out, I

Nico Johnson  00:42

we've moved into an era of giga scale. Ai campuses concentrated demand at levels that the grid really was not designed for. Frankly, the question isn't whether they're coming. The question is whether our energy architecture, generation, storage, market design is built for this level of scale that also includes a serious technical question. Today's dominant storage chemistries and architectures may not be sufficient for the scale we are asking them to supply, or perhaps the way in which this architecture needs to be responsive. Are we trying to power Giga scale future with tools designed for yesterday's load profile? That is the overarching question to explore that. I'm joined by John Perella, CEO of Terra flow, working at the intersection of storage deployment and Giga scale infrastructure, and Anna seifkin, Director of Policy and markets at the long duration energy storage Council. Better was better known as LDS and formerly the US Department of Energy, focused on how market design and policy align with the physics of our grid. Welcome you two.

Anna Siefken  01:42

It's great to be here. Thank you so much.

Nico Johnson  01:44

Awesome to see you. They told me that my the question lineup was challenging, so let's see how much we can make them squirm. Anna, from a system design perspective, I'm curious what changes when the load for Giga scale becomes highly concentrated,

Anna Siefken  02:00

when load becomes Giga scale and geographically concentrated, which is what happens with the data center, the real issue becomes reliability. So that whole equation. So what it used to be is that you had all of these different electrons coming into the system, and they would intersect and intermingle, but now you have these huge bursts of energy that are needed. And John, I know is going to talk about the intermittent data, the intermittent nature of the power that comes in. Intermittency is something where we understand because of renewables, but this is an entirely different ball game, yeah. So what we really have to do is focus on how to compress all of that diversity together,

Nico Johnson  02:38

yeah,

Anna Siefken  02:38

which is a different scenario. So planners have to rely on different sets of variables, and you have to raise the baseline so that you increase the probability that any extreme weather events or occurrences aren't going to top out what is already there in that elevated demand. By the way, that's where long duration energy storage comes in, right? And that's why I'm so excited to be here today, particularly with John and you mentioned the LDS Council. I do want to put in a quick note about the LDS council. So we are a global organization, a trade organization, that works with all types of long duration energy storage, eight hours and more. It could be days, weeks and seasons, across multiple different technologies, and just really excited to be here. John is one of our, our proud members, too.

Nico Johnson  03:26

Yeah. And one of the fastest growing as far as I can tell, John at the project level, talk to me about sort of legacy data center architecture compared with Giga scale AI campuses. And really what's changing in the way the the two approach, sort of the shape of power required.

Jon Parrella  03:43

So legacy data centers were very flat power curves. They were their load profiles were kind of very stable. The issue we're facing with these large AI data center loads is extremely volatile power and the issue is that when the data center starts to scale. They the whole data centers in unison, so that the volatility is gets bigger and bigger and bigger,

Nico Johnson  04:10

yeah, and

Jon Parrella  04:10

that starts to create all kinds of new infrastructure challenges that the data centers are facing, because they have to be able to determine in forecast for that load. First of all,

Nico Johnson  04:20

yeah,

Jon Parrella  04:20

the grids can't operate fast enough to be able to respond to the volatility they're experiencing, so there has to be something in there to buffer it. They're creating a tremendous amount of harmonic distortion issues. And when you have that kind of power volatility, you can trip air breakers at substations. Yeah, right. And it's reacting faster than even the generation assets can keep up with. The fastest generation asset on the planet right now is natural gas. Is going to be a recip at about 25 to 30 seconds, we're seeing power volatility where you're getting multiple 3040, 50, all the way up to 80% swings at the total size of the data center. So you talk a gigawatt scale, you could have 300 500 800 megawatt swings multiple times per. Minute, yeah, and they're breaking crankshafts on Gen sets. I mean, it's crazy the new infrastructure requirements that are coming out.

Nico Johnson  05:07

Yeah, the graphic I saw that you sent over to us is something like six to 10 shunts of power throughout the minute, a one minute cycle. This is all related to just the nature of how an AI data center processes data and requires and consumes power in in an intermittent fashion, right? And that's versus like the nine nines of baseline power that a data center used to have, yeah,

Jon Parrella  05:30

so that they want maximum resiliency, as far as five nines reliability,

Nico Johnson  05:34

right?

Jon Parrella  05:35

But at the same time, they're, they're trying to ask the grid or generation assets to do something beyond the laws of physics in most cases. So it's a very interesting paradigm that they're creating, that storage, honestly, is really the only solution to be able to help bridge

Nico Johnson  05:52

that gap.

Jon Parrella  05:52

And that's why you're seeing a lot of increase in demand for storage solutions and and I think it was Darryl on your last podcast was saying there's a lot of new chemistries coming out. That's where long duration storage really is at is it's not lithium, it's other chemistries.

Nico Johnson  06:07

Yeah, let's talk about lithium. It dominates, obviously, storage deployment today at a giga scale. What are the architectural or chemistry constraints beyond just duration?

Jon Parrella  06:17

There

Anna Siefken  06:17

are actually quite a few. So it's interesting that short duration. Usually, lithium ion has become sort of a standard. It's like a building block. But what we're seeing is that it doesn't scale the same way. It's it can be modular, it can be standardized. It's also well understood, which is one of the challenges that we face with these different technologies.

Jon Parrella  06:38

Yes,

Anna Siefken  06:39

they're not even newer technologies, all of them, they're just different. So when you get into the the load, when it's concentrated in this way, you aren't just solving for that daily peak that you can get out of this short duration battery. You're actually looking at something that can go much longer, that has no degradation, and that is really built as infrastructure. So we talk about long duration energy storage as fit for purpose, because it's based on geography. It's based on use case. It's based on different things that you need. You can also get, not only power to power, which is the movement of electrons, you can also get power to heat, which is thermal, so you can use that power, or even the waste, to make heat and turn it back into electricity. So the degradation, the cycling are important. But I would also mention we have to talk about safety, because when you talk about lithium ion and you talk about runway for fire, those are real issues and real concerns, and most, if not all, of the long duration energy storage technologies that we work with are have addressed any fire issues. It's not even a part of the conversation. So you look for site specific, you optimize whatever system you're going to be looking at, and then it lasts for it can be put together in 1218, months, up to five years, depending on your system, and it can last up to 50 years. In some cases, these are infrastructure projects that are critical. And honestly, if we make that short term decision to just stack batteries, we aren't thinking about the long term solution that long duration energy storage can solve. For

Nico Johnson  08:19

John, one of the things that you and I have talked about offline is this concept around in concept around energy density. You all at Terra flow have a vanadium flow battery. By no means a new technology.

Jon Parrella  08:28

40 years old,

Nico Johnson  08:29

yeah, 40 years old. And, and, and it's, you know, it's, it's been tried and true, reliable. Somebody asked me yesterday, they were like, well, but nobody's ever done it at scale, that that would be wrong too. You could talk about that, yeah. But can you talk about the sort of the concept of energy density, some of the misnomers around vanadium flow, particularly and and the your perspective on specifically use of lithium ion at data centers, since that's the topic that we want to center

Jon Parrella  08:56

around

Jon Parrella  08:57

with lithium ion, and Anna spoke to most of the major points really well, with lithium ion, you've got a decent energy density. It's about 250 kilowatt hours per meter squared, yeah. And one of the challenges with lithium ion is that, because of the thermal runaway probability that could occur, especially in a data center situation, where you're exposing it to high volume of frequency cycles, right, you're exponentially increasing that probability with lithium ion, you have to it's a field of CONUS containers that you have to spread out and so that the challenge is that you need more floor plan to do it right, which is why they're doing shorter duration with traditional flow batteries. I heard a statistic at a conference we were at just recently for flow batteries, and they were trying to achieve 125 kilowatt hours per meter squared.

Nico Johnson  09:44

Oh, that's a goal. That's

Jon Parrella  09:45

their goal. So which means they're not even there yet.

Nico Johnson  09:47

Half of what lithium can do, we've

Jon Parrella  09:49

actually been able to achieve 654 kilowatt hours per meter squared with our terminology. So we've made some major breakthroughs in what we call large tank format flow batteries, specifically designed for grid scale batteries and data set. Dollars,

Nico Johnson  10:00

right?

Jon Parrella  10:00

But we've been able to achieve it in a much smaller footprint and 150%

Nico Johnson  10:03

greater density,

Jon Parrella  10:05

greater density, no thermal runaway. And actually, what's crazy with the data centers now, and it's it's sad to say this, is that they're turning the batteries and the genocide cores into consumables, and that's a very high capex expense for infrastructure to have to it's not that it doesn't work

Nico Johnson  10:24

right.

Jon Parrella  10:24

It's just that you're going to have to replace them very frequently if you use them. And if that's your architecture path, that's where a lot of new technologies are emerging to try and help solve that problem, to be able to handle the volatility, to be able to solve for the the the harmonic distortion issues that they're experiencing. They're experiencing, and then be able to take them even beyond that with turning them into responsive loads. And that's really what's getting exciting.

Nico Johnson  10:49

Yeah, on the note of responsive loads, the state of Texas is as as per usual, sort of leading the charge on how we think about the value stack, how we how we can monetize these assets. Anna, you spend a lot of time around market and policy, certainly prior to LDS, and now thinking specifically about your around your customer set, are we sending clear signals that the flexibility and resilience that these assets can offer to scale are being valued appropriately?

Anna Siefken  11:21

So we're not, we're not seeing exactly what we need in the market signals, but they're coming up in certain places. We're seeing it in California, we're seeing it in Texas, and as of yesterday, we're seeing it in Virginia.

Nico Johnson  11:33

Oh yes,

Anna Siefken  11:34

state bill and House Bill. It's going to the governor, which has a carve out for long duration energy storage, and we were a part of an exercise to help bring that along.

Nico Johnson  11:45

Fantastic.

Anna Siefken  11:45

So the point of it is, is that it's not where it needs to be, because what we need are specific carve outs, because we need to be able to compete for our lithium ion or four hour short duration is not enough. We need capacity accreditation that works. We need to be able to do revenue stacking. We need for financiers and investors to not feel the risk that they're currently feeling right. These market signals tell an investor that this is the right thing to do, right and it's not universal right now, so we're not necessarily seeing it coming from the federal government. We're seeing at the state level, so it's a bottom up approach. There are still blind spots, and what they're not seeing is a way to compensate for duration as well as resilience. So resilience becomes a really important thing here. If we can keep a data center from going down, that's millions and millions of dollars saved.

Nico Johnson  12:42

Yeah,

Anna Siefken  12:42

right, but that risk resilience reward is not necessarily being seen, but we're making really good progress, state by state, ISO by ISO.

Nico Johnson  12:52

John, one of the risks that I've heard you say that you are able to help mitigate is the now feels like age old, interconnection queue problem. You pointed to a Princeton study around how to accelerate the motivating factors that help not only the utilities but the developers, integrate projects better. Can you talk about that element of incentivizing resiliency and also, one of the things that I've heard you describe around storage is that it's sort of the shock absorber of the grid. What specifically are we absorbing? But talk about those two topics for us, it was interesting. I read an article yesterday after we did

Jon Parrella  13:30

our interview, and what was amazing is that the article came out and said, If the data center is not a response large load is not a responsive load, they're not going to get interconnected at all. And the regulatory is pushing that way. Senate Bill six in Texas is an example of that, where they're now requiring them to be controllable loads

Nico Johnson  13:48

right

Jon Parrella  13:49

in the very near future, that's going to continue to push beyond that, where it's going to turn into responsive loads,

Nico Johnson  13:54

yeah,

Jon Parrella  13:54

if the data centers want to succeed, they have to become assets to the grid, not liabilities if they want grid interconnection,

Nico Johnson  14:01

right?

Jon Parrella  14:01

Even for the ones that want to do behind the meter generation, if they don't put some sort of buffer or shock absorber between the data center and the generation assets, they're going to burn out their cores much faster.

Nico Johnson  14:11

Yeah, and

Jon Parrella  14:12

that's a very expensive and very long lead item,

Nico Johnson  14:15

and if it is a truly controllable load, then they relinquish some control over how quickly it gets degraded

Jon Parrella  14:23

absolutely and there are companies that are trying to use lithium ion batteries to basically charge when the data center drops off and discharge when the data center peaks, to flatten that curve out as much as possible. The only issue is, we've heard from some of the customers that have done that where they burning out in like six months. So it's, it's scary. There was an article out of South Korea where one of them caught fire inside the data center. Yeah, and they had to bulldoze it out while on fire just to be able to put it out and not catch the rest of data center on fire. So you're going to start to see that more, I'm sure, with lithium ion now, that's not to say all lithium ion batteries are equal. There are some safer chemistries and non i. So I don't want to just bash on lithium ion, because I'm a big believer in all storage needs to be pushed more. But I don't believe it's the right solution for the data centers when it comes to this volatility issue, and I don't believe that short duration batteries are the solution to the data center issue. I think it's got to be long duration storage, which allows them to extend the life of all their assets. It allows them to become responsive loads, and it solves all of their complex power control system. And I mean, it just goes on and on and on when you implement the solution the correct way.

Nico Johnson  15:31

It's interesting to me because, you know, we just saw alphabet acquire intersect. One of the things that intersect and bright night and many other developers are pushing is this hybridized system approach. What do you see as adaptations that are being required in the marketplace to accommodate the just the sheer volume of of load growth that we're seeing in the market, and how? How? How does that adaptation favor long duration storage?

Jon Parrella  16:00

So what we've seen is that, obviously, with the lead time on natural gas generation assets, a lot of a lot of the data centers are turning to solar because it's faster to deploy. And so now you're introducing a hybrid of generation sources, which creates a whole nother challenge of managing power and then the power control system. So you have, you know, potentially, a mix of solar with either fuel cells or natural gas, depending on if they're in a non attainment zone or not. And then beyond that, then they're adding batteries and grid eventually, right? So they all have different intermittency they all have different ramp rates. They all have different control systems, which creates a whole nother complexity of how do you match and balance what the data center is drawing versus what the Gen sets or the generation assets are creating. And that's where batteries become that hub or that buffer, and that that shock absorber, if you will, because it decouples the data center from the generation assets and allows them to have reaction time and really extend the life of all their assets. So

Anna Siefken  16:57

one thing that you could mention, though, as well as Co Location. So we're seeing quite a bit of Co Location,

Jon Parrella  17:02

absolutely.

Anna Siefken  17:03

So there's this hybridization of the systems where we're stacking to get where we need to go right, as we're getting all of these technologies and bankable right. And we have a new bankability report where we talk about all the barriers and things that you need to have in place in order for a project to be bankable, we really need to have something that we can put out there. And we're doing it many cases on site. I know that you're doing that in Katy, Texas. You're doing a co location with a data center. Well,

Jon Parrella  17:31

we're doing a number of different things. There's in Texas, we're seeing a lot of data centers co locate with renewable assets that are that are stranded or congested.

Nico Johnson  17:39

Yeah.

Jon Parrella  17:39

So 500 megawatt in wind farms, they're only deploying 125, megawatts because of congestion on the grid. So now they're trying to figure out how to use that stranded asset and monetize that stranded asset.

Nico Johnson  17:50

Yeah,

Jon Parrella  17:50

but again, any one of those resources is going to require more than just the wind or the solar to be able to accomplish the goal. So yes, actual Co Location and hybridization is isn't a must for these data centers, especially if they want that resistance. Want that resiliency and reliability.

Nico Johnson  18:06

If we look out over the next decade, we've got to architect a grid that really feels like it's reliable and resilient. What does that architecture look like? And what's absent today?

Anna Siefken  18:18

So it's it's got integrated generation and storage as a single planning unit. So I think that's one of the things that's really missing, is that we're bolting different things on. We're cluging it together right now. It'll be planned in advance, and resilience will be embedded in that design. We'll have capacity accreditation that's fully recognized duration and all of the value sets that it offers. And then that long duration capacity is going to be embedded in the infrastructure. It's not just optional capacity. And I would say by 2035 if I was predicting, I would say that a system won't be looking whether it can do eight hours or 24 hours. It's just going to be duration as needed we're going to get to that space.

Jon Parrella  19:04

I think it's interesting, because there's a bunch of articles that have come out where they talk about, if you're 95% reliable, that you could fit most of the data centers that are being planned today, I think that there's a huge amount of stranded asset from a standpoint of time, and so that's where longer duration allows you to shift that power around and flatten that curve quite a bit and fit more on the existing infrastructure that you have today. So I think storage is going to be a key player. You know, I heard the mention of SMRs. I'm a big believer in it. We're very involved in multiple SMRs. But as everybody knows, nukes are on or off. There's no variability to them whatsoever. So I think when you start to incorporate new technologies, it also has to be able to deal with the exposure to that kind of volatility, and have to have something that can buffer it. So I think no matter which way you go, storage is definitely going to be the solution. Now that doesn't mean we don't. Need more generation assets or other things as well. I think what you're going to see over time is that all these data centers that are building behind the meter generation will get grid connected, and then they will become generation assets that become peaker plants. So if you put the right storage in place where the data center can be responsive load, then they can inject that power, and you have a huge new wave of infrastructure that's able and available to the market. So I think, you know, I've heard a lot of people talking about how data centers are increasing the price. I think you're going to see the opposite, once they get built at scale. They actually help drive the price down, and they're going to be some of the Savior to the grid, upgrading the infrastructure to meet the demand. I

Nico Johnson  20:39

would, I'm gonna, I'm gonna play devil's advocate. I would love to believe that that is true, but most of the people building the data centers are capitalists, and when they become utilities, they will behave like utilities and and capture the available dollar per kilowatt hour that utilities capture well,

Jon Parrella  20:57

but I could be wrong. If you look at Senate Bill six, for example, they're they're forcing the cost of the infrastructure upgrades on the data centers. Yeah, so the data centers are paying for those upgrades and not passing it through to the rate payers.

Nico Johnson  21:08

No. But when once they have the opportunity to charge rate payers, or in some way be be post market compensated for it, my hypothesis is that I don't feel as altruistic about the data centers when they become utilities, as perhaps you do, but you've had more conversations with them than

Jon Parrella  21:26

I do. They're gonna have to operate within the grid. Rules.

Jon Parrella  21:29

So,

Nico Johnson  21:29

rules, Yeah, true. I've so John mentioned the interview that we did. We do have an interview that come out for all of the questions that are popping up in your head, as they did mine when I first started hearing John tell me these outlandish claims around how we're breaking assets having built data center. So please, if you haven't yet subscribed, subscribe to SunCast. That episode will come out in March, and you absolutely want to queue that up. I wanted to ask you a few quick questions and then one final wrap up question, what's the most under misunderstood aspect of long duration storage today,

Anna Siefken  22:06

that it's just a longer battery,

Nico Johnson  22:08

that

Anna Siefken  22:09

it isn't a whole sector, an entirely different way of thinking about how we store electricity,

Jon Parrella  22:16

right? That it's a new technology,

Nico Johnson  22:19

that

Anna Siefken  22:20

it's not ready,

Jon Parrella  22:21

that it's

Anna Siefken  22:22

not

Jon Parrella  22:23

ready. How many gigawatts? How many gigawatts is ronke deployed? They're over 5000 megawatts. They're deploying a gigawatt per year right now,

Nico Johnson  22:32

batteries.

Jon Parrella  22:33

And I've been to their factory, and it's impressive, yeah, and they're out. They're friends of ours. So it's, it's, we're big believers in it. We're standing up five gigawatts of manufacturing capacity of our batteries right now, right

Nico Johnson  22:43

here in Texas, in Texas? Yeah, so not here.

Jon Parrella  22:46

We're not in Texas,

Nico Johnson  22:48

right here in

Jon Parrella  22:48

America. But yeah, no, we're standing up manufacturing at scale here in the US. What's

Nico Johnson  22:52

one architectural assumption that needs to evolve?

Anna Siefken  22:55

I would say that four hours should be the benchmark. Four hours isn't the benchmark. It's It was what we could do at the time stretch

Nico Johnson  23:03

goal.

Anna Siefken  23:03

It's an art. It's, it's, it's a it's a relic,

Nico Johnson  23:09

interesting.

Jon Parrella  23:10

So not all batteries are equal. A lot of times we get paired with just best,

Nico Johnson  23:15

yeah, and

Jon Parrella  23:16

so you know that there's a lot of best moratoriums and things coming out right now because of lithium ion, unfortunately, and there are alternate, safer, cleaner solutions for storage that are long duration. So we're having to constantly educate the policymakers on, you know,

Anna Siefken  23:37

and the public and

Jon Parrella  23:38

the public and that, yeah, you know, yes, building theirs may be at risk, but building ours is not final

Nico Johnson  23:43

question. If we architect all of this properly, what becomes possible at Giga scale? That's not possible today, and I'll toss my answer in there. Turns out things like vanadium flow batteries need, like many other technologies, need scale to prove how to scale up and become cost effective. So ronke is a great example of how if we actually get to giga scale, we can have tech, not new technology options afforded to us. So if we architect this all properly, what becomes possible John and then and then Anna.

Jon Parrella  24:13

So we've been able to drive the price of vanadium flow batteries through the floor compared to our competitors, just because we're going at a whole nother scale,

Nico Johnson  24:21

density through

Anna Siefken  24:21

the roof.

Jon Parrella  24:22

Energy density has gone up, and we're solving for a different problem than historical so a lot of it is it was something I was reading this morning about bankability. It wasn't your report, but it was a different comment on bankability. Once you get to that scale, bankability becomes a lot easier. So there's a lot of factors that go into building at that scale. Fortunately for us, the solution for data centers and demand for data centers is really helping drive that demand and allowing us to scale faster than I think anybody could even possibly imagine

Nico Johnson  24:55

in a last word.

Anna Siefken  24:55

So if we if we architect it correctly, there's. Going to be build out that doesn't need to happen. We're going to be able to use our existing assets. Customers are going to be happier. Things are going to be less volatile, and we're going to have resilience in the wake of storms and events that might happen. So there's going to be and maybe, if people don't know that, it was the reason. That's a good thing, right? Because the event never happened.

Jon Parrella  25:24

Yep,

Anna Siefken  25:24

they didn't have a blackout or brownout. They just never had it, and they don't even know why, but we, but we did that. Yeah?

Nico Johnson  25:31

John peralez, the of Terra flow, and Anna seifkin is the Director of Policy and markets at the long duration energy storage Council. If you are going to one of our other upcoming events. Please swing by our live stage and say hello. I do love meeting you in person. You can actually find out which events we'll be at next at SunCast, dot media and click on the Events tab twice weekly, we deliver conversations with founders and leaders on the front lines of the clean energy transition, and we're here bringing this content to you each and every week free of charge. Of course, it's not free. Our sponsors help pay the bills and keep the lights on so that we can help you build your legacy in the clean energy transition. And so if you'd like to say thank you to them, or learn more about what it is that they bring into the world, or perhaps see how you could reach 1000s of listeners twice a week, just like they do. Check it out at SunCast. Dot media. Forward slash sponsor, remember you already listened to thanks again for showing up. Solar warrior, it's half

Jon Parrella  26:30

the battle.