1 00:00:00,040 --> 00:00:01,920 Speaker 1: All right, it's time now for our daily Wall Street 2 00:00:01,920 --> 00:00:05,680 Speaker 1: Week conversation, and the rise of AI is posing challenges 3 00:00:05,720 --> 00:00:09,440 Speaker 1: to data centers. Applied Digital, which designs, develops, and operates 4 00:00:09,520 --> 00:00:12,480 Speaker 1: data centers, is looking at how to adapt its infrastructure. 5 00:00:12,800 --> 00:00:15,400 Speaker 1: Wes coom and See is the company's co founder, chairman, 6 00:00:15,440 --> 00:00:18,400 Speaker 1: and CEO. He joins us now alongside Wall Street Week's 7 00:00:18,520 --> 00:00:21,160 Speaker 1: David Weston and data centers. I mean, it's one of 8 00:00:21,200 --> 00:00:22,639 Speaker 1: the hottest toppings out there, it really is. 9 00:00:22,680 --> 00:00:25,160 Speaker 2: Every day we're talking about data centers somehow, some way. 10 00:00:25,200 --> 00:00:27,440 Speaker 2: But we thanks so much for joining us because normally 11 00:00:27,440 --> 00:00:29,280 Speaker 2: when Katie and I talk with people who at data centers, 12 00:00:29,280 --> 00:00:31,159 Speaker 2: it tends to be the commercial real estate side of 13 00:00:31,160 --> 00:00:33,639 Speaker 2: it and investig in it. But you actually developed these 14 00:00:33,680 --> 00:00:36,720 Speaker 2: things and actually install and operate them. So give us 15 00:00:36,720 --> 00:00:39,199 Speaker 2: a sense of what's the big deal with data centers 16 00:00:39,240 --> 00:00:42,200 Speaker 2: now that didn't exist before the rush to AI. We've 17 00:00:42,200 --> 00:00:43,800 Speaker 2: had them around for some time, right. 18 00:00:44,479 --> 00:00:49,720 Speaker 3: Yeah, David, I think that the key to understand, the 19 00:00:49,760 --> 00:00:53,519 Speaker 3: primary thing to understand is the difference in the compute 20 00:00:53,520 --> 00:00:55,440 Speaker 3: workload that's going into data. 21 00:00:55,160 --> 00:00:56,960 Speaker 4: Centers now versus what it was in the past. 22 00:00:56,960 --> 00:00:59,360 Speaker 3: So in the past, you know, we were building data 23 00:00:59,400 --> 00:01:02,640 Speaker 3: centers that were in population centers that we're providing ultralo 24 00:01:02,840 --> 00:01:05,240 Speaker 3: latency communications so that. 25 00:01:05,200 --> 00:01:07,800 Speaker 4: We could get YouTube, so that we could get Netflix. 26 00:01:07,360 --> 00:01:09,080 Speaker 3: So we could get TikTok, right, so we can get 27 00:01:09,520 --> 00:01:11,560 Speaker 3: all of our things that have been driven mostly by 28 00:01:11,600 --> 00:01:12,600 Speaker 3: video applications. 29 00:01:13,000 --> 00:01:14,240 Speaker 4: Now, as we're moving. 30 00:01:14,200 --> 00:01:18,759 Speaker 3: From applications that need ultralo latency communications outside the data 31 00:01:18,800 --> 00:01:23,200 Speaker 3: center to applications that need massive amounts of compute. So 32 00:01:23,480 --> 00:01:26,000 Speaker 3: think of, you know, doing large math problems algorithms. This 33 00:01:26,120 --> 00:01:29,240 Speaker 3: is what AI is doing inside the data center. Compute 34 00:01:29,280 --> 00:01:32,640 Speaker 3: equals power, and then inside the data centers you need 35 00:01:32,720 --> 00:01:36,200 Speaker 3: a lot of compute close together, so you need really 36 00:01:36,280 --> 00:01:38,240 Speaker 3: high power density. And so these are kind of some 37 00:01:38,280 --> 00:01:41,600 Speaker 3: weird numbers, but you know, historically data centers were let's 38 00:01:41,600 --> 00:01:44,959 Speaker 3: call it seven kilowatts for an entire rack of computing, 39 00:01:45,319 --> 00:01:48,360 Speaker 3: and Nvidia a one hundred server one by itself takes 40 00:01:48,400 --> 00:01:51,120 Speaker 3: ten ten kilo wants. So and then you need to 41 00:01:51,120 --> 00:01:53,280 Speaker 3: stack all of these together, so you have a massive 42 00:01:53,400 --> 00:01:55,880 Speaker 3: amount of power consumption that needs to be close together. 43 00:01:56,000 --> 00:01:59,200 Speaker 3: The latency inside the data center now versus outside, needs 44 00:01:59,200 --> 00:02:01,440 Speaker 3: to be ultralove, and so you have to build a 45 00:02:01,480 --> 00:02:04,560 Speaker 3: completely different style of data center, you need large amounts 46 00:02:04,600 --> 00:02:06,920 Speaker 3: of power in the same location, and so that's what 47 00:02:06,960 --> 00:02:10,520 Speaker 3: we're doing. We find large amounts of power, typically different 48 00:02:10,600 --> 00:02:12,080 Speaker 3: locations than you would do city center. 49 00:02:12,120 --> 00:02:12,840 Speaker 4: So it's no. 50 00:02:12,840 --> 00:02:15,960 Speaker 3: Longer New York or Virginia or Dallas or Los Angeles. 51 00:02:16,200 --> 00:02:18,760 Speaker 3: We're building in different locations where we use a lot 52 00:02:18,800 --> 00:02:22,120 Speaker 3: of renewable energy that have these hundreds of megawatts, if 53 00:02:22,160 --> 00:02:25,840 Speaker 3: not gigawat of power so that you can run these 54 00:02:25,880 --> 00:02:29,000 Speaker 3: workloads because they're just very different style workloads versus what 55 00:02:29,040 --> 00:02:29,480 Speaker 3: we're used to. 56 00:02:29,760 --> 00:02:32,720 Speaker 1: Right and when it comes to that adaption process really 57 00:02:32,840 --> 00:02:36,000 Speaker 1: sort of converting those existing data centers to be able 58 00:02:36,040 --> 00:02:40,120 Speaker 1: to handle the AI workload that people are trying to do. 59 00:02:40,200 --> 00:02:43,000 Speaker 1: I mean, how long the process is that? How long 60 00:02:43,080 --> 00:02:44,200 Speaker 1: is that timeline? 61 00:02:44,760 --> 00:02:48,680 Speaker 3: So I'm a big believer that you're better doing green 62 00:02:48,760 --> 00:02:51,720 Speaker 3: Field in that you just build an entirely new style 63 00:02:51,760 --> 00:02:53,959 Speaker 3: of data center versus trying to retrofit. 64 00:02:55,160 --> 00:02:56,160 Speaker 4: Cooling is different. 65 00:02:56,240 --> 00:03:00,560 Speaker 3: Again, the power density is significantly different, so there's there's 66 00:03:00,600 --> 00:03:01,280 Speaker 3: a big issue. 67 00:03:01,600 --> 00:03:04,519 Speaker 4: So the number one issue right now is finding power. 68 00:03:04,600 --> 00:03:07,960 Speaker 4: Where's near term power? We have we have a very. 69 00:03:07,960 --> 00:03:10,600 Speaker 3: Unique way of finding power, which is we find things 70 00:03:10,639 --> 00:03:13,040 Speaker 3: that are called stranded power, and we typically use a 71 00:03:13,080 --> 00:03:15,680 Speaker 3: lot of renewable energy that's more in the middle of 72 00:03:15,720 --> 00:03:16,360 Speaker 3: the country. 73 00:03:16,960 --> 00:03:18,920 Speaker 4: So number one is power, and then number two is 74 00:03:18,960 --> 00:03:19,640 Speaker 4: supply chain. 75 00:03:19,760 --> 00:03:23,560 Speaker 3: So this is electrical components, it's just you know, it's transformers, 76 00:03:23,639 --> 00:03:27,000 Speaker 3: it's chillers, it's high voltage switch here. So the supply 77 00:03:27,080 --> 00:03:30,400 Speaker 3: chain because of the demand for this has become really 78 00:03:30,440 --> 00:03:31,160 Speaker 3: stretched as well. 79 00:03:31,200 --> 00:03:33,239 Speaker 4: So number one is power. Number two supply chain. 80 00:03:33,400 --> 00:03:35,400 Speaker 2: So wes are you building your own power? I mean 81 00:03:35,400 --> 00:03:38,280 Speaker 2: you're building those wind farms or solar event or even 82 00:03:38,400 --> 00:03:40,520 Speaker 2: nuclear for that matter, or are you tapping into some 83 00:03:40,560 --> 00:03:43,120 Speaker 2: other part of the grid. And similarly on the supply chains, 84 00:03:43,440 --> 00:03:45,200 Speaker 2: how long is the lag time on all this? It 85 00:03:45,200 --> 00:03:46,480 Speaker 2: sounds like it's going to take an awful lot of 86 00:03:46,480 --> 00:03:48,840 Speaker 2: time to build that power and also come up with 87 00:03:49,040 --> 00:03:50,040 Speaker 2: all that supply chain. 88 00:03:50,840 --> 00:03:53,440 Speaker 3: So right now we're not building the power that's something 89 00:03:53,480 --> 00:03:56,120 Speaker 3: for the future for us. So there's there's a unique 90 00:03:56,120 --> 00:03:59,520 Speaker 3: setup in North America with that we take advantage of 91 00:03:59,760 --> 00:04:01,360 Speaker 3: what I call it is stranded power. 92 00:04:01,760 --> 00:04:04,680 Speaker 4: So there is power that is generated and not used. 93 00:04:04,720 --> 00:04:06,280 Speaker 3: And I know a lot of people would think that 94 00:04:06,320 --> 00:04:08,560 Speaker 3: if you can generate an electron anywhere in the US, 95 00:04:08,560 --> 00:04:09,720 Speaker 3: it can be used anywhere else. 96 00:04:09,560 --> 00:04:11,000 Speaker 4: In the US, and that's just not the case. 97 00:04:11,360 --> 00:04:14,720 Speaker 3: You need the transmission infrastructure to actually be able to 98 00:04:14,720 --> 00:04:18,400 Speaker 3: do that. So we find places where, you know, for Ellendale, 99 00:04:18,440 --> 00:04:22,240 Speaker 3: North Dakota, for example, there's a massive amount of wind 100 00:04:22,320 --> 00:04:25,760 Speaker 3: power that feeds into a substation, and oftentimes they have 101 00:04:25,800 --> 00:04:28,920 Speaker 3: to curtail the wind farms because there's not enough demand, 102 00:04:28,960 --> 00:04:32,039 Speaker 3: there's not enough load to use the power that's generated. 103 00:04:32,320 --> 00:04:33,799 Speaker 4: So we go to that location. 104 00:04:34,279 --> 00:04:36,320 Speaker 3: And we have several of these locations where we use 105 00:04:36,360 --> 00:04:39,880 Speaker 3: power that's currently not being used now to go to 106 00:04:39,920 --> 00:04:43,000 Speaker 3: the other part of your question is for new power, 107 00:04:43,080 --> 00:04:46,640 Speaker 3: so you can build power gen on renewables over you know, 108 00:04:46,720 --> 00:04:49,000 Speaker 3: let's call it a twenty four maybe thirty six month 109 00:04:49,000 --> 00:04:49,440 Speaker 3: time frame. 110 00:04:49,520 --> 00:04:52,039 Speaker 4: The bigger issue is transmission. 111 00:04:52,160 --> 00:04:55,520 Speaker 3: So transmission to transmit that power that typically is built 112 00:04:55,720 --> 00:04:58,520 Speaker 3: somewhere in the middle of the country to end users. 113 00:04:58,720 --> 00:05:01,080 Speaker 3: You know, that's more of a tend thirteen year process, 114 00:05:01,120 --> 00:05:03,960 Speaker 3: and so we're kind of shortcutting that by taking these 115 00:05:04,000 --> 00:05:06,160 Speaker 3: workloads directly to the point of generation. 116 00:05:06,800 --> 00:05:09,080 Speaker 1: I want to talk about the competitive landscape here because 117 00:05:09,120 --> 00:05:12,200 Speaker 1: you think about AI and it's just really made its 118 00:05:12,200 --> 00:05:14,919 Speaker 1: way into every part of the market. It seems like 119 00:05:14,960 --> 00:05:17,960 Speaker 1: every investor in every asset class is trying to figure 120 00:05:17,960 --> 00:05:20,920 Speaker 1: out what's the AI narrative in their industry. When you 121 00:05:20,920 --> 00:05:23,559 Speaker 1: think about your own industry and data centers right now, 122 00:05:23,760 --> 00:05:25,120 Speaker 1: how fierce is the competition. 123 00:05:27,279 --> 00:05:29,320 Speaker 4: So it's an interesting time for sure. 124 00:05:29,400 --> 00:05:32,120 Speaker 3: So if you go back a year ago, everyone is 125 00:05:32,120 --> 00:05:34,640 Speaker 3: trying to figure out on the data center side, what's 126 00:05:34,680 --> 00:05:37,679 Speaker 3: happening in this market right Why is all of this demand? 127 00:05:37,800 --> 00:05:39,320 Speaker 4: Why for the high power density? 128 00:05:39,680 --> 00:05:43,200 Speaker 3: Now in this industry figured it out, and now it's 129 00:05:43,520 --> 00:05:47,000 Speaker 3: trying to make this solution because again, getting the amount 130 00:05:47,000 --> 00:05:50,160 Speaker 3: of power that's needed and operating the style of data 131 00:05:50,200 --> 00:05:54,840 Speaker 3: centers is extremely difficult. And so think of on the 132 00:05:54,920 --> 00:05:57,640 Speaker 3: data center side, we're the picks and shovels. This is 133 00:05:57,720 --> 00:06:01,240 Speaker 3: digital infrastructure. This is the digital infrastructure was built out 134 00:06:01,240 --> 00:06:03,840 Speaker 3: to run everything that we're you know, everyone's used to 135 00:06:03,920 --> 00:06:05,719 Speaker 3: running on their phone or on their TV at home, 136 00:06:05,960 --> 00:06:09,000 Speaker 3: and now it has to be rebuilt. It's not a replacement, 137 00:06:09,000 --> 00:06:11,599 Speaker 3: by the way, this is an add on, but rebuilt 138 00:06:11,720 --> 00:06:15,320 Speaker 3: for compute and AI workloads. So we're just we're at 139 00:06:15,360 --> 00:06:18,640 Speaker 3: the building blocks. We're the infrastructure that runs all the 140 00:06:18,680 --> 00:06:20,400 Speaker 3: AI that you know, everyone wants to do and we 141 00:06:20,520 --> 00:06:22,880 Speaker 3: talked about you know, everyone knows chat GBT, but if 142 00:06:23,080 --> 00:06:26,360 Speaker 3: you're doing anything in in uh, you know, in video 143 00:06:26,640 --> 00:06:30,919 Speaker 3: and visual anything generative AI runs on this infrastructure and 144 00:06:30,960 --> 00:06:33,159 Speaker 3: a massive amount of this infrastructure used to be built, 145 00:06:33,480 --> 00:06:35,520 Speaker 3: you know, and when Jensen and in Video talks about 146 00:06:35,520 --> 00:06:38,880 Speaker 3: AI factories. So so what we've done we actually designed 147 00:06:38,880 --> 00:06:42,159 Speaker 3: our data center and you know, in partnership with in Video, 148 00:06:42,160 --> 00:06:44,560 Speaker 3: we work with their team to build the style of 149 00:06:44,640 --> 00:06:47,400 Speaker 3: data center that we're doing because it's this is the 150 00:06:47,400 --> 00:06:49,520 Speaker 3: issue in the data center industry is it's just a 151 00:06:49,760 --> 00:06:52,960 Speaker 3: very different style from data centers that were done in 152 00:06:52,960 --> 00:06:56,200 Speaker 3: the past. But we're running all AI workloads and you 153 00:06:56,200 --> 00:06:59,360 Speaker 3: know these are really split between training and inference. So 154 00:06:59,440 --> 00:07:03,000 Speaker 3: the terms that are used in AI, but the entire 155 00:07:03,000 --> 00:07:06,599 Speaker 3: industry is i would call it scrambling to build the 156 00:07:06,640 --> 00:07:09,720 Speaker 3: amount of infrastructure that's needed to house the compute to 157 00:07:09,840 --> 00:07:11,280 Speaker 3: run these applications. 158 00:07:11,960 --> 00:07:13,640 Speaker 2: Well, it wasn't that long that we hadn't even heard 159 00:07:13,680 --> 00:07:16,280 Speaker 2: about all these data centers. What's going to prevent absolutcence? 160 00:07:17,800 --> 00:07:19,440 Speaker 4: So that's a great question. 161 00:07:19,520 --> 00:07:22,160 Speaker 3: So when we build our facilities, we think about what 162 00:07:22,200 --> 00:07:25,280 Speaker 3: we call future proofing. Right, so we think the power 163 00:07:25,280 --> 00:07:29,480 Speaker 3: density goes higher cooling, but I think you know, the 164 00:07:29,680 --> 00:07:33,880 Speaker 3: last twenty thirty years were again really driven by applications 165 00:07:33,920 --> 00:07:37,440 Speaker 3: that required call comms, high speed communications, and as we 166 00:07:37,480 --> 00:07:39,800 Speaker 3: go for the next twenty or thirty years, applications are 167 00:07:39,840 --> 00:07:42,120 Speaker 3: really going to be driven by compute and that's what 168 00:07:42,200 --> 00:07:45,080 Speaker 3: AI is, and so we're working to future proof that. 169 00:07:45,160 --> 00:07:49,600 Speaker 3: And by the way, the older style data centers, those 170 00:07:49,640 --> 00:07:53,200 Speaker 3: are not obsolete. We still need all of that capacity 171 00:07:53,280 --> 00:07:55,720 Speaker 3: for the applications that we continue to use. This is 172 00:07:55,960 --> 00:07:59,480 Speaker 3: absolutely a complete add on to what's been done in 173 00:07:59,520 --> 00:08:02,720 Speaker 3: the past and so but trying to not be obsolete 174 00:08:02,800 --> 00:08:05,600 Speaker 3: is making sure that your future proofing your facilities as 175 00:08:05,680 --> 00:08:08,320 Speaker 3: much as possible. With power density, with cooling, we're trying 176 00:08:08,320 --> 00:08:11,440 Speaker 3: to run an extremely high efficiency so that we lower 177 00:08:11,440 --> 00:08:15,080 Speaker 3: our carbon footprint. You know, we build where there are 178 00:08:15,400 --> 00:08:18,680 Speaker 3: renewable we mostly use renewable resources. But those are the 179 00:08:18,720 --> 00:08:21,000 Speaker 3: things you try to do obviously, you know, not being 180 00:08:21,040 --> 00:08:23,280 Speaker 3: obsolete in the world is one of the keys here. 181 00:08:23,400 --> 00:08:25,600 Speaker 1: All right, future proofing, that's a good place to leave it. 182 00:08:25,680 --> 00:08:28,560 Speaker 1: Really enjoyed this conversation our thanks to wes Coomins. He 183 00:08:28,720 --> 00:08:31,720 Speaker 1: is Applied Digital co founder, chairman and CEO