Battery Storage Solutions For AI Data Centers And Power Grids

本站添加: 2026-05-22

[00:00:00]Tesla makes batteries. Yes, that's not a surprise. They've been making battery energy storage solutions for quite some time now using cells sourced from other manufacturers. And uh they still do that today. But a big chunk of the secret sauce that makes it so incredibly valuable is the software behind it. How come no one has copied it? Could it even be copied? Is there a way to understand what they're doing and try and figure it out in ways that can help other companies? It turns out there is uh our friends at uh Electric Era Charging in uh Washington State have been working on a solution just like that. We're going to get into it, find out what it is, what it does, and how I can steal it for myself. I'm Brian. Welcome to Futurasa.

[00:00:50]>> So, I'm here uh for very selfish, very greedy reasons. I'm going to uh effortlessly copy what you guys have spent years developing because that's how easy it is, I'm sure. So, you've got something new. Tell me what it is.

[00:01:05]>> Yeah, well, good to be here with you, Brian. And, uh, we're excited today to be announcing the COP power platform, which is our power capacity as a service offering for edge AI data centers that helps them unlock new behindthe-ter power capacity for their data centers.

[00:01:21]uh alleviating the grid in the process and ultimately saving a lot on electricity costs. Um this is a product that we've been building in stealth for about six to seven months and today we're announcing it to the public.

[00:01:34]>> Well, I know when I asked you guys maybe three months ago, is this is this something you're working on? You guys were like, uh, you know, who knows, right? Uh which could have gone either way, but here we are. So, this isn't a physical object per se. What does that mean? How does that work?

[00:01:56]>> Well, no. So, so ju just to be clear, it is it is a physical object. I uh I probably did a bad job describing this in the beginning. Um but let me share my screen real quick and I'll just show you. So at at a high level, you know, kind of what Electric Era, I would say at our DNA is really good at is using um really world-class software platform technology that we've developed to operate power infrastructure systems and and one examples of that is our EV fast charging stations. Many of your subscribers and viewers have probably used our EV fast charging stations and have come to discover it's a reliable experience with some advanced capabilities around energy management and our energy storage systems that we have on site. With that software system, we're able to pull together hardware assets that are brittle and fail and have a lot of unique behaviors and operate them in a very high reliable fashion that is delighting to drivers um and also uh in in this case delighting to the reliability requirements of data centers. So um we take that software, we pair it with uh some of our own proprietary hardware and a lot of thirdparty vendor hardware to build EV fast charging stations. And we're taking that same approach with the co-ower platform. Um, so we're we're using different battery providers, different larger power conversion system providers, different providers of uh medium voltage switch gear and transformers. We're rolling that whole hardware stack together with our uh software platform for power infrastructure that I just talked about in order to bring what is called behind the meter power capacity online for edge data centers. So you can see an example of that in the the photo here. This is a you know an edge data center with an array of roughly six energy storage systems. Those energy storage systems would have about five megawatt hours each and would each be able to discharge about 2 megawws. So call it uh you know 30 megawatt hours and approximately what is that 12 megawatts of total power that we can feed into the data center to support intermittent and uh peak usage ultimately helping them alleviate the electrical grid which is a really big part of the reason we did this.

[00:04:14]>> Well then where does the power come from? This is uh the kind of really underappreciated and perhaps understood part of how energy storage systems work, which is they just charge up during periods of low usage when the grid is not at its worst case strain and then they discharge during periods of peak usage when the grid is under a lot of stress. Uh this is what's called load shifting. Um so imagine a you know morning time period where there's not a whole lot of activity on the electrical grid. Um those are periods where you can charge up the battery for much lower cost without impacting um you know energy prices. You ultimately store that energy until later in the day typically during 4 to 9 p.m. when electricity spikes. Um, and in doing so, you actually alleviate the power pressure on the grid during the peak times because you're shifting the load from the morning to the evening.

[00:05:13]>> So, what I'm looking at here is a reservoir and what you're selling is the dam that controls when the power comes out in and out of it, when the water goes in and out of the reservoir. Uh, because it doesn't rain in the summer in a lot of places. Uh but thanks to uh reservoirs that's not uh going to kill all the crops and the people because you're storing it uh for times when it's more scarce. So what you're so what you're saying is the you you know put in the batteries, no big deal. Get them from wherever >> and then your system smartly controls the the ins and outs.

[00:05:54]>> The ins and outs and the reliability of the power delivery. Um yeah, that that's right. and and doing this well is challenging because of what I mentioned earlier um around power infrastructure being brittle u and failing in a wide variety of modalities. Uh those the combination of those modalities also of which are unexpected and hard to uh navigate. So our system is able to kind of resiliently respond to that. Think of uh all of the things on site at a deployment like what you see on your screen. You have inverters, you have medium voltage switch gear, you have medium voltage transformers, you have the battery systems themselves, you have power meters, a whole array of telemetry systems that are gathering data. All of those are possible failure modalities.

[00:06:38]And with the kind of antiquated nature of a lot of power infrastructure, uh you ultimately see those systems fall out or not share data in the way that they're supposed to. The challenge for us as an operator of energy storage systems is using those systems in a manner that's reliable and that can uh you know deliver power to the data center or whatever end load you may have. You know again we do this for EV fast chargings every single day um in a very high reliable fashion so that the power is there when the user needs it. So, uh, is it because these kinds of installations, we're talking multimemegawatt installations, is it reasonable to say that customers can save a substantial amount of money right out of the gate?

[00:07:24]>> A shocking amount of money. Yeah. So, there's a a lot to unpack about how the utility prices electricity, but the two simple ideas are they charge you for the electricity itself, the the actual electrons. You know, imagine a single electron going into a single uh part of a battery inside your iPhone. That single electron has a cost. And then they charge you for the wires, all of the switch gear, the the the infrastructure itself. That's a separate cost structure. Both of those get rolled up into your energy bill. Um and if you're a data center, you're paying a lot for the infrastructure because it's so big. And so what we can do with this solution is we simply because we're moving power from point A to point B from the morning till the evening we're actually able to alleviate the cost of the power infrastructure in a really meaningful way and that's uh beneficial for everybody. Certainly data centers are under a lot of pressure to remove uh alleviate the strain that they place on the grid and and that's the price signal uh that helps them additionally save money and uh incentivize them to go and invest in solutions like the co-ower platform. So, what this means is as the grid adds more renewables, wind and solar, which are intermittent, this could be one of the many little reservoirs that hold that surplus as it comes in. Um, I'm not going to change my my electricity consumption habits by very much. I guess I could wait until nighttime to dry my clothes. Certainly, I could schedule my car to charge at night, but I really don't use that much juice. Uh, but with data centers, uh, the thirst is inquenchable, and you've got, uh, potentially a lot of storage.

[00:09:16]And if it is clear that in 3 hours, uh, the the power rates are going to be significantly lower, you could choose to discharge at a greater rate. Your software manages that.

[00:09:30]>> Yeah. programmatically and autonomously actually. And we're looking at a variety of other price signals um from the electrical grid every single day. And the software system actually autonomously decides how to interact with the grid in that day against those price signals in order to optimize the cost structure for the data center which ultimately is designed to alleviate pressure on the grid and drive down price prices for for all rateayers. So that's a really I think important part of the software systems that we've developed for our EV charging business that we actually deploy all across the country already. Um we're taking that same software and applying it to just a larger load um with bigger batteries. Uh so for us you know in designing this product we we started from kind of the first principles perspective of what do we have that is really good from a company competency perspective our software systems. What? And then we asked ourselves the questions, what do we need to do to uplevel our capability set uh for this new larger load customer? And that involved kind of just upsizing everything. Bigger batteries, bigger switch gear, bigger transformers.

[00:10:38]Um, and in doing those two things, we were able to kind of quickly stand this new product up in a very logical and streamlined fashion and have now been developing a large array of customer demand for that. Now, uh, for those who don't know, uh, Quincy's been on the show a number of times. Uh, Quincy and others at Electric Era are former SpaceX engineers. They worked on a variety of different uh different parts of the project. And what they really got out of it was the SpaceX speed, the the problem solving, and that's why when you've got a competency like that, you want to lean into it. Uh, so that's the background on on that. I realize that not everyone has seen every video we've done. We've done a few. Uh, another question I had for you about these data centers, uh, is there time of use for them in terms of their demand, not the cost of the power, but do they have more do they run full throttle 24/7 or are some data centers, you know, catering to the US where they're running higher loads during the day?

[00:11:44]>> Yeah, that's a fantastic question and the answer is that it depends. There's some nuance here. So, when you see uh the word data center, you think of, you know, a kind of one-sizefits-all building that does the exact same thing as every other data center. But the truth is that they're all doing different types of workloads. They all have different infrastructure. And the way that the data center industry has been built out over the last 20 years has been to try to optimize the cost of running all that infrastructure for all of the workloads that are requested by these data centers. Now I mention that because it is kind of a leadin to where we decided to focus. Electric Air is not focused on alleviating the cost problems for large super uh factories, AI superactories uh mainly because they are kind of running full boore all day long and you really need like an on-site generation system like solar or natural gas or a nuclear power plant to ultimately alleviate uh the grid pressure in the way that we talked about. And that's because it's running full boore all the time. there isn't much time of use arbitrage that you can do. But for us in our focus on edge data centers which are smaller facilities that process real time low latency workloads uh think of an AI call center or a kiosk taking your order or a uh Whimo assisted fleet that's having inference running locally uh to help it drive. All of those are time-sensitive and actually have some sort of dural shape to them uh that matches the human circadian rhythm and the human daily activity schedule. And as you know, people are not awake typically at night.

[00:13:32]They're typically asleep and and that alleviates a lot of pressure on the grid and on these edge data centers during those times of the day. We're taking advantage of that time of use workload activity and leveraging energy storage systems to do the load shifting that we talked about earlier. So our focus is on the edge and helping those data centers ultimately alleviate a lot of their cost structure and uh strain on the electricity grid.

[00:13:59]>> So what I think I hear you saying is I probably shouldn't get one >> for you personally.

[00:14:07]Uh you can all Yeah, if you if you're interested in buying one of these uh let us know and um probably not a good fit though. Yeah, >> probably probably not right for my use case.

[00:14:18]>> Although you have Yeah, you have a lot of uh you know a lot of views and you know there is viewership happening during certain times of day. That's a good example, right? You probably have a lot of views happening during certain times of day and then views that are not happening in other times of the day.

[00:14:32]That's the exact kind of uh peak demand offsets that we are taking advantage of on just a much much larger scale with much more aggregated demand. Uh it's all of that getting added together.

[00:14:43]>> Now uh one thing I've heard about is that electricity is not as smooth as you might think. Uh it doesn't it doesn't have a beautiful laminer flow like you would get out of a out of a well engineered bit of uh water, you know, systems. Um, when the power comes into my house, if it's a little ragged, a little bumpy, uh, nothing in my house really cares. Uh, you can put a universal power supply between your computer and your wall that will smooth it out a little. Um, but with data centers, that's more important. Uh, how important is that?

[00:15:22]>> It's very important. And actually, it's important in the opposite opposite perspective that you than what you might imagine. In fact, it's AI data centers that potentially in the because of the way that they're using energy um that ultimately can cause what's called uh kind of power rippling on the local electricity grid. So, just like your example on the water from earlier, um there are kind of parameters within electricity that can change over time.

[00:15:58]One of them is voltage. Uh voltage is kind of like the pressure in water.

[00:16:03]Imagine if you had water pressure going hot and then off and then on and then off and then on and then off like a fire hose.

[00:16:10]>> That would >> Yeah, that would be equivalent to uh the voltage in the power going up and down and up and down. And that's called voltage flicker. Uh, a voltage flicker can occur as a result of AI data centers training where they're turning on a GPU and then turning off a GPU and then turning on a GPU and turning off a GPU.

[00:16:30]They have so many synchronous GPUs all activating together at the same time that if you don't smooth out that power usage before the grid sees it, it can actually cause the same voltage flicker behavior on the whole electrical service territory. So battery systems can be used as filters, electricity filters that actually smooth out that power impact on the electricity grid. This is another major use case that batteries are helping with in order to solve some of these AI load use problems um and in a manner that ultimately benefits every rateayer.

[00:17:07]>> So it's not just it is better for the data center, but it's also better for the entire everyone up to the substation level. uh and potentially even beyond.

[00:17:18]Uh because of course it didn't occur to me, but yeah, if it's rough going in, it'd be rough uh on both sides of the meter.

[00:17:26]>> That's right.

[00:17:28]>> Now, uh what's the actionable part of this video? Do we do we tell people if you have a data center, check out electric.te?

[00:17:38]I think um yeah, I think I think um when this video goes live, we will have that web page that I brought up earlier um on our website. So, you can go to electricair.te and you will have two options now to select from. If you're interested in our EV fast charging products, you'll click on that option.

[00:18:00]And if you're interested in our data center solutions and technologies and what our team is building of Xbasics engineers in the data center sector, click on that option. Um so you know for us these are kind of intertwined and and um problems and and as a result these are intertwined technologies and we actually share a lot of our uh kind of secret sauce from our EV charging product into the data center product. So I think over time you know these two things will become further and further intertwined from a technical perspective and there'll be a lot of interesting announcements as a result of that.

[00:18:33]>> Anything I'm forgetting to ask?

[00:18:35]>> Um no but I I just will add that uh you know batteries are just absolutely paramount for us to solve the future of the grid. There's hyperinmcy for load generation with renewables.

[00:18:49]There's hyper intermittency with consumption of electricity uh with things like data centers. That means you have kind of two mismatching intermittent supply demand curves and you need a buffer, an electrical buffer, a reservoir to dampen that all out. Um we as a country have about a terowatt of grid capacity. Our average consumption on it at is about 500 gigawatts.

[00:19:17]Batteries help ultimately use more of the grid capacity, boosting the average higher to the peak. And it's a foundational technology that we're going to need a whole lot more of. And I expect, you know, this as an industry will cause us to deploy teratt hours of energy storage over the next uh you know, dozen or so years. This is going to be a major major shift in how the grid works and I'm super excited to be participating. How are battery prices doing at the wholesale level, sell prices?

[00:19:50]>> Yeah, that's a great question. Um, you know, I think folks that are interested in that should kind of keep an eye on the lithium spot market price. Uh, if you go to I'll just share my screen.

[00:20:01]This is a really interesting look for folks. So, there's a bunch of commodity inputs obviously in battery cells. Um, there's a lot that goes into that. Uh but at a high level um lithium is a major cost driver right and right around what was that 2022 november lithium spiked a massive amount and then it fell off a cliff and then it went up a little bit and then it fell off a cliff and now it's kind of stabilizing. Um so but but I would say stabilizing and going up in a more sustainable fashion. I would say this is a leading indicator for what the price is at the battery cell and battery pack level. As a result of the dynamics playing out here and the massive amount of manufacturing that's been stood up, battery prices have gone down materially. When I first started building batterybacked power infrastructure, battery pack prices was about 200 or more dollars per kilowatt hour. Today you can buy packs at 70 to90 uh dollars per kilowatt hour. It's a massive reduction in cost approximate approximate numbers there. But that's the level of decrease that we've seen over the intervening years.

[00:21:12]>> Wow. And that's at the pack level, not the cell level.

[00:21:15]>> Yeah. Uh the $70 per kilowatt hour figure is more of a cell level number. I would say at the pack level um you always want to add about I don't know probably 20 to 30% more cost. So packs are getting substantially cheaper as well. And overall the whole industry is declining in its cost curve.

[00:21:36]>> Mhm. And we are seeing a lot more a at least in this hemisphere uh factories being built. That's still moving forward reasonably well.

[00:21:45]>> Yeah. Um I can't announce yet. Uh but actually you know we will have this announced in the uh the news when it comes out. Uh we we have selected as our uh vendor um LG Energy. are of course open to talking to other vendors and like all the other solutions like the mega pack but LG Energy as an example does produce in America and I think that's really good right you know like as a country we we want to have a lot of highpaying manufacturing jobs for one and then for two it is useful to have that technology here in America in case there's things like the uh kind of the uh you know Ukraine war popup you know we we want to have resilient um supply chains as we kind of all learned during co >> I'm of the opinion that energy independence is national security it just and forget about any kind of political tensions which is impossible but just having a supply chain disruption having a shipping disruption uh having a pandemic we saw all kinds of supply lines that we thought were very secure uh absolutely crumble uh under just the least amount of pressure. Uh so guys, uh in the comments, what did we miss? What do we misunderstand? Leave it. Leave it. I beg of you. Uh bunch of other great videos on the channel with Electric Eura Tech. You can uh look them up. You can just search them. Uh who knows? Maybe we'll put some in the description. Probably not. I don't know.

[00:23:13]Everybody else, you know, uh, like, subscribe, do what you do, and stay tuned, stay juicy, and I cannot wait to hear from you, clever robots, when next we