บทถอดความฉบับเต็มของ Can the Power Grid Keep Up With AI? with Stephanie Cox

the grid of of power based on these peak events, weather events, anything like that, but the data centers and the demand that AI is creating is creating just a different type of shift than what we've traditionally seen before. 00:03:26.085 --> 00:03:30.631 One of these hyperscalers needs a gigawatt of energy for a new AI data center. 00:03:30.631 --> 00:03:33.737 What happens? Where does the plan meet reality? 00:03:33.737 --> 00:03:41.628 Yeah, so a lot of that is is hidden challenges that we see driving where some of the coordination early on is happening. 00:03:41.628 --> 00:03:48.971 So it's not just that, hey, we need power, we're having these new data centers, it's the amount of planning that's required. 00:03:48.971 --> 00:03:55.782 So in traditional energy usage, it's figuring out land permits, water, a lot of community involvement. 00:03:55.782 --> 00:04:06.367 So we're seeing that just at a bigger scale too, and kind of a new shift, but it's creating more of a dichotomy between is the power going to the data center or is the power going to communities? 00:04:06.367 --> 00:04:33.536 So from the early play stage of planning, we see you know, securing this power is requiring more than just a power purchase agreement. So the developers, the owners, all the parties involved need to navigate transmission lines, different interconnection cues,~ permitting water requirements, land availability and and local workforces and definitely that community support that I stressed on too. 00:04:33.536 --> 00:04:40.276 So we're we're seeing some changes going on with that and just a lot of different demand that we didn't see a decade ago. 00:04:40.302 --> 00:04:48.711 AI is shifting so fast with new models every few weeks and the chips they need and the requirements change every eighteen months or on a shorter cycle. 00:04:48.711 --> 00:04:51.076 But substations can take years to build. 00:04:51.076 --> 00:04:52.846 How do those two things kind of meet? 00:04:52.846 --> 00:04:55.858 Where AI wants to be fast, but infrastructure is slow. 00:04:55.858 --> 00:04:57.613 It just works at the pace of construction. 00:04:57.613 --> 00:04:59.495 Yeah, that's that's always the big challenge. 00:04:59.495 --> 00:05:07.927 I I'd say, you know, from my perspective,~ a bridging technology has been energy storage, especially in the last few years. 00:05:07.927 --> 00:05:18.394 It's it's been the bridging technology for just the differences between renewable and traditional energy. But then same with AI and data centers as well. 00:05:18.394 --> 00:05:26.620 So we're seeing that Software is kind of building at its own pace and a lot faster than what the grid and infrastructure can keep up with. 00:05:26.620 --> 00:05:40.399 But having energy storage kind of bridge that gap while while equipment catches up has g been a just a key factor because we're still facing supply chain, anything like that is is just can't keep up with the pace of of software growth. 00:05:40.427 --> 00:05:47.610 There's been a lot of talk of building data centers in unique locations, whether it's under the ocean or in outer space. 00:05:47.610 --> 00:05:52.418 How realistic is all of that? And is it just a distraction or is it a real possibility? 00:05:52.418 --> 00:05:54.209 You know, that's a a funny question. 00:05:54.209 --> 00:05:57.420 My my personal opinion is i it's a distraction. 00:05:57.420 --> 00:06:03.863 I mean, even even right now, these massive data centers, they're they're all simply science projects. 00:06:03.863 --> 00:06:11.158 We haven't seen something like this in previous decades and they've been deployed maybe, you know, they have been deployed but at a much smaller scale. 00:06:11.158 --> 00:06:15.139 But the scales that they're deploying right now are these larger hyperscalers. 00:06:15.139 --> 00:06:19.540 So I think having different areas to deploy, as you mentioned. 00:06:19.312 --> 00:06:28.278 ~ in in oceans, anything like that is is all helping to solve different problems, maybe from a heat standpoint, anything like that. 00:06:28.278 --> 00:06:35.896 But at the end of the day, it's it's still just newer technology coming in the ground that still needs to be hashed out between different parties. 00:06:35.896 --> 00:06:47.067 So if if there's more of a focus on how we figure out this land demand and the balance between just the grid on the grid itself versus getting water or space or anything like that. 00:06:47.067 --> 00:06:50.713 I think that's that's kind of the focus that I hope we we remain. 00:06:50.740 --> 00:06:53.540 We've already seen AI really affect some markets. 00:06:53.540 --> 00:07:01.637 Like the price of RAM is continually going up, up, up 'cause they're buying all of the chips for now and it's really affecting personal computers and other sectors. 00:07:01.637 --> 00:07:06.055 Even large companies like Sony can't get the chips for PlayStations anymore. 00:07:06.055 --> 00:07:08.899 So prices of everything like there's this fallen effect. 00:07:08.899 --> 00:07:14.187 And as AI needs more and more power, are we gonna start to see our power bills go up and up and up? 00:07:14.187 --> 00:07:17.795 And At what point are people going to decide, well, I don't want AI this much? 00:07:17.795 --> 00:07:23.331 And it's going to lead to that shift where it's the cost benefit analysis switches and people turn against AI. 00:07:23.331 --> 00:07:48.750 Yeah, I think I think two parts to your question is you know, one around pricing. I think from an energy market standpoint, there's there's been fluctuations, whether that's a political landscape, the different types of technologies, you know, we'll see different pockets, such as in California, Texas, kind of these hot spots that we've seen over the last couple decades, that newer technologies get brought onto the grid, cheaper power, anything like that. 00:07:48.750 --> 00:07:49.922 So I think we'll see that. 00:07:49.922 --> 00:07:53.605 replicate

acl across the data center landscape as well. 00:07:53.605 --> 00:07:56.697 So right now I think prices are still increasing. 00:07:56.697 --> 00:08:00.980 Everything's kind of matching the oil index as well and and how we see that. 00:08:00.980 --> 00:08:05.122 But I think there's enough healthy technology coming to the mix. 00:08:05.122 --> 00:08:07.423 And my personal opinion is is we need it all. 00:08:07.423 --> 00:08:16.641 You know, I've I've historically been in the renewable space, but agnostic to energy at the end of the day, because it's it's that healthy balance and load shifting that we do need. 00:08:16.641 --> 00:08:30.370 So pricing, my hope is will get driven down just as more people become receptive to receiving different types of technology and thermals making more of a comeback, nuclear is coming back too. So we'll see that shift. 00:08:30.370 --> 00:08:33.192 So I'd say that's that's where it's at with pricing. 00:08:33.192 --> 00:08:49.631 ~ your second point on communities kind of getting involved, and there's actually a good good ~ case study that was from the Nevada Energy is They had notified some communities of hey, this power is gonna go to data centers if if that needs to be the primary demand. 00:08:49.631 --> 00:09:08.167 So we are seeing a little bit of of shift from where communities might have a little bit more pushback. I was I was recently attended a webinar that there was some stat that I thought was astronomical but said 90% of folks in the US don't necessarily want the data centers, all these hyperscalers coming. 00:09:08.167 --> 00:09:16.856 coming into play because it's affecting where do you balance data center getting the power versus communities and livelihood getting the power too. 00:09:16.856 --> 00:09:24.115 So I think there's a hot topic debate that will still come out, but you know it's it's just coming down at the local community levels as well. 00:09:24.143 --> 00:09:30.279 Yeah, I think that in the past, when we've built infrastructure products, there's also been a lot of jobs that come with it. 00:09:30.279 --> 00:09:33.389 But data centers don't really have a lot of people inside of them. 00:09:33.389 --> 00:09:36.683 So there's this big cost, but not as much benefit to the community. 00:09:36.683 --> 00:09:41.832 So we're seeing a lot of communities think go, our prices are going up. We have power lines thrown over the city. 00:09:41.832 --> 00:09:44.914 There's all these negatives, but there's no positives for us. 00:09:44.914 --> 00:09:48.817 Maybe there's some taxes for the people in charge, but it's not really helping the community. 00:09:49.303 --> 00:09:50.865 And so we're seeing a lot of pushback. 00:09:50.865 --> 00:10:02.260 And one of the things that interests me is like we talk a lot about water and heat as a waste product is isn't there a way to turn the heat back into electricity to kind of get a benefit from that instead of just throwing it all away? 00:10:02.260 --> 00:10:05.873 Like I thought we have geo geothermal energy and these other resources. 00:10:05.873 --> 00:10:09.067 Is there a more efficient way to use that or something smarter we could be doing? 00:10:09.067 --> 00:10:23.168 Yeah, I I think there's some great opportunities for as we see this adva coming to the grid, it'll promote other technologies to pop up too and kind of have that different advancement and how we utilize the waste side of it. 00:10:23.168 --> 00:10:29.392 So to your point, the the thermal energy and water from that is great capture. 00:10:29.392 --> 00:10:37.267 But again, I think it becomes the issue of how do we get that captured energy back into transmission so that it's being used for something. 00:10:37.267 --> 00:10:40.268 Outside of just what's locally available on the site. 00:10:40.268 --> 00:10:59.663 But I've heard of yeah, some some different areas around the water being for data centers, as well as where we see these different microgrids for energy storage popping up in local communities and and folks that are close to these areas being able to push their power back to the grid too if they have some solar or anything like that. 00:10:59.663 --> 00:11:13.211 So it's it's kind of a mix of of where we see the current infrastructure, but I do think there's a lot of opportunity for how we can make these different types of technologies or different load demands be a lot more efficient than what they currently are. 00:11:12.759 --> 00:11:17.821 Biggest hottest topic right now is these hyperscalers, these massive AI data centers. 00:11:17.821 --> 00:11:27.019 And to me, then math never seems to work out because right now chips used in AI are obsolete within 18 months and that cycle keeps shortening. 00:11:27.019 --> 00:11:38.014 So I've seen smaller data centers spend like millions of dollars on chips that eighteen months later are considered obsolete and they don't recover the cost in that quickly. 00:11:38.041 --> 00:11:42.816 Right, because it's takes a long time in rental price, because the rental price is so much less than the purchase price. 00:11:42.816 --> 00:11:51.977 What do you think all these investors are kind of getting wrong when we look at this, whether it's the energy or the water or the land space or the locations or the power needs? 00:11:51.977 --> 00:11:58.000 Because we keep I think to me it seems like the thing we're forgetting is that we're constantly becoming more efficient. 00:11:58.028 --> 00:12:01.951 Like the model that can fit inside of your phone now, which I think is what everyone really wants. 00:12:01.951 --> 00:12:06.807 We want a local model, not something that's far away, not in a data center. That's the future everyone wants. 00:12:06.807 --> 00:12:09.527 And I feel like the shift is going to be in that direction. 00:12:09.527 --> 00:12:14.474 So my feeling is we're going build all these data centers, and by the time we build them, AIs will be so efficient we don't need them.

00:12:14.474 --> 00:12:18.764 Like it's this ~ problem that the second we solve it, we don't need that solution, anyways. 00:12:18.764 --> 00:12:22.307 But what maybe I'm wrong, and I want to get your opinion. 00:12:22.307 --> 00:12:32.971 Like, what do you think investors are getting wrong by pouring all their money into this idea where there's all these data centers, but we don't really know if there's a long term it's very exciting, I guess, in that field. 00:12:32.971 --> 00:12:38.236 Like everyone wants to jump in on it, but is it the right long term move and what are they kind of missing, in your opinion? 00:12:38.071 --> 00:12:50.782 Yeah, that's a that's a super interesting topic because it's you know, especially is if I if I can rel relate it back to the renewables market is a lot of renewables today have been driven by tax credits. 00:12:50.782 --> 00:12:54.384 So it's investors jumping in to kind of get a piece of that. 00:12:54.384 --> 00:13:02.692 And to your point too, is just where they can get a piece of the data centers and you know, capitalize on this twelve to eighteen months plan. 00:13:02.692 --> 00:13:06.955 So I think there's a good mix of of folks just jumping in right away. 00:13:06.955 --> 00:13:21.900 But some of the investor folks and the clients that we work with, they are looking five years out and trying to model that decline that I think I agree with you is that you know maybe there'll there'll be a decline in and where we see this because the technology becomes obsolete. 00:13:21.900 --> 00:13:23.741 We might not have as much demand. 00:13:23.741 --> 00:13:34.046 Software is getting smarter so that we won't have to build this intense amount of chips and processing. And the hope is that as we innovate, everything does become a little bit more efficient. 00:13:34.046 --> 00:13:42.499 So I think there's a healthy amount of companies and investors that are looking for more than just, you know, a two-year out, they're looking five to ten. 00:13:42.499 --> 00:13:47.649 And they're really looking to diversify where they have their portfolio. 00:13:47.649 --> 00:14:07.932 So I think investors today, at least from the clients we work with, are being a little bit smarter, a little bit more savvy about comparing cost benefit of the energy side of it, where their portfolios invested in that, in addition to capitalizing on this near term opportunity in in anticipation of that decline. So that's my opinion. 00:14:07.932 --> 00:14:18.972 I share the same with you of just let's hope things become more efficient as we've seen. I mean, three years ago I I know companies talking about policies around chat GPT and here we are today. 00:14:18.972 --> 00:14:22.976 So what is an accelerated path of the next three to five years look like? 00:14:23.004 --> 00:14:51.664 There's a lot of abandoned businesses, whether it's malls that are closed all over America or old hotels and all these data centers we're building, do you think there's a po what happens when we build more than we need and they're kind of the pullback happens and suddenly there's these abandoned massive mega centers? Will the improvements we've made to the grid then benefit All the people in those communities that had the initial negative effect will kind of switch. Well, now we have a stronger grid and there's so much more capacity we need. So the price of electricity will plummet. 00:14:51.664 --> 00:14:52.798 Is that a possibility? 00:14:52.798 --> 00:14:54.308 I I think that's a possibility. 00:14:54.308 --> 00:15:13.066 I I smirk because we could go into some rabbit holes around, yeah, sustainability and these abandoned spaces because that that's gonna be an issue coming into that. But I I think that's a a major possibility of what happens when these massive pieces of land, equipment, building space is no longer needed. 00:15:13.066 --> 00:15:18.600 And I think with anything, you know, I try to look at the positive of what's what's the opportunity that's created. 00:15:18.822 --> 00:15:36.126 years down. And so I think there'll there'll be a lot of space for secondary~ the sustainability side. I know there's folks looking into and researching even everywhere from the the abandoned malls to have the vertical agricultural growth because it's already a compliant building. 00:15:36.126 --> 00:15:39.148 You already have the HVAC system kind of in the mix with that. 00:15:39.148 --> 00:15:42.048 So hopefully people are planning ahead too. 00:15:42.048 --> 00:15:52.870 But I think that that definitely broaches a new topic of ~ sustainability in that sense too, which is close to the heart, especially on where we see all the waste of equipment, aging, anything like that. 00:15:52.897 --> 00:16:04.428 Yeah, it really seems like we are so excited about the technology that we're not you very rarely hear people talk about sustainability anymore, where it was a really important term a few years ago. 00:16:04.428 --> 00:16:10.010 Now it's like I mean, we haven't built a nuclear facility in America since the 70s. 00:16:10.010 --> 00:16:11.553 I think that's when we started the moratorium. 00:16:11.553 --> 00:16:13.864 And now suddenly everyone's buying reactors. 00:16:13.864 --> 00:16:17.517 They're firing up Three Mile Island again, they're doing all of these things. 00:16:17.517 --> 00:16:25.543 You're talking about micro reactors and So the mindset about nuclear has shifted so fast from it's the worst thing ever, never do it again to let's build a bunch. 00:16:25.543 --> 00:16:27.660 There was no middle transitional period. 00:16:27.660 --> 00:16:36.953 And we're seeing the same thing where I think at first it seems you see the benefits, but communities are starting to see, wait, you're gonna take all the power, the power's gonna go up, there's no jobs, it's gonna take a lot of infrastructure. 00:16:36.953 --> 00:16:41.239 And we're starting to see kind of the cost benefit. 00:16:41.239 --> 00:16:44.157 When we look five

or ten years out, what do you think? 00:16:44.047 --> 00:16:49.529 The future really is going to be like when AI, data center, storage, and the grid actually converge. 00:16:49.529 --> 00:16:51.957 What do you think the future will be like in five or ten years? 00:16:52.217 --> 00:17:29.432 I I would say I expect and hope for a more diversified and resilient ecosystem. You know, ever everything learning from the past,~ to your point where people are just kinda chasing a shiny ball for lack of better terms, too. And but I think now there's a lot more collaboration that's known and needed to get projects in the ground efficiently, to have a right execution plan in place, the right partners. So~ five to ten years. Yeah, I hope we're relying more on more than just one technology. the distributed data centers, like I mentioned, not these hyperscalers. 00:17:29.432 --> 00:17:32.673 I think it'll be based on where we see higher demand. 00:17:32.673 --> 00:17:35.615 Right now it's based on land availability from what I'm seeing. 00:17:35.615 --> 00:17:44.999 And then just a broader mix of generation source sources with storage, smarter grid controls, and I'd say just AI itself. 00:17:44.999 --> 00:17:47.648 helping to optimize the electricity demand. 00:17:47.648 --> 00:17:54.461 So that that's my hope. ~ where we see opportunities like that and I think these secondary markets that are coming up. 00:17:54.461 --> 00:18:01.981 ~ one one piece of of GridEQ side that we're leaning into is all the secondary equipment that is unused. 00:18:01.981 --> 00:18:08.053 you know, projects get canceled because interconnection cues and timeline, supply chain shifts, anything like that. 00:18:08.053 --> 00:18:15.229 And so part of GridEQ business is is helping to find homes for this brand new equipment that hasn't come up. 00:18:15.229 --> 00:18:24.573 So I think there's again opportunity for helping to balance that out and just people being creative with with where we can be a little bit more efficient and a lot more sustainable. 00:18:24.599 --> 00:18:26.441 I think this is a really important conversation. 00:18:26.441 --> 00:18:28.565 It sometimes gets lost in all the excitement. 00:18:28.565 --> 00:18:33.934 Either people are very pro AI or very anti, but we forget all the secondary markets infrastructure behind it. 00:18:33.934 --> 00:18:35.236 So this is really, really useful. 00:18:35.236 --> 00:18:39.436 So you tell us a little bit about what you do and where people can find you in GridEQ? 00:18:39.351 --> 00:18:44.292 Yeah, so so GridEQ at our core, we're a an advisory intelligence firm. 00:18:44.292 --> 00:19:18.457 so we provide consulting work, working with investors, developers, different owners. started this company a little over a year ago with a business partner, and we saw just a a gap in all this growth and demand and the lack of clarity really when it comes to finding partners, the right teams to execute. So ~ some of the advisory services we provide as a shining a lot more light on the companies that investors might be looking to put their money in, how projects are being built with the different partners out there. So that's our core. 00:19:18.457 --> 00:19:48.885 And then ~ as I mentioned previously, we we recently launched a new business line about five months ago for this secondary market look focusing on transformers and breakers and all this long lead equipment time that data centers different types of technologies all need to to actually get their projects in the ground. So we're hoping to just fill the gaps in in this crazy energy provide some clarity and and hope to bring a lot more projects to speed a lot faster and with a lot more intelligence behind it. 00:19:48.913 --> 00:19:55.476 It's amazing. Thank you so much for sharing and being with us today for another amazing episode of the Artificial Intelligence Podcast. 00:19:55.476 --> 00:19:56.775 Thank you, Jonathan. Appreciate it.