1 00:00:00,560 --> 00:00:03,080 Speaker 1: This is Tom Rowland's Reese and you're listening to Switched 2 00:00:03,120 --> 00:00:06,440 Speaker 1: on the podcast Sports to You by BLOOMBERGHENIF. A massive 3 00:00:06,480 --> 00:00:09,680 Speaker 1: build out of data center capacity is underway. Bloombergenif is 4 00:00:09,720 --> 00:00:12,799 Speaker 1: tracking nearly twenty three gigawatts of new data center capacity 5 00:00:12,880 --> 00:00:16,120 Speaker 1: under construction worldwide, more than a third of today's installed 6 00:00:16,160 --> 00:00:19,400 Speaker 1: computing base. As AI and cloud workloads drive the next 7 00:00:19,440 --> 00:00:22,640 Speaker 1: phase of digital infrastructure growth, but deciding where to build 8 00:00:22,680 --> 00:00:26,000 Speaker 1: is becoming more complex. Energy availability is emerged as a 9 00:00:26,000 --> 00:00:30,000 Speaker 1: defining constraint, with developers prioritizing markets that can deliver large, 10 00:00:30,000 --> 00:00:33,519 Speaker 1: reliable grid connections quickly at the same time. Land permitting, 11 00:00:33,600 --> 00:00:36,960 Speaker 1: tax regimes, fiber connectivity, and the presence of existing data 12 00:00:36,960 --> 00:00:40,519 Speaker 1: center ecosystems all shape how competitive a market really is. 13 00:00:41,040 --> 00:00:44,800 Speaker 1: As AI workloads scale and sovereign ambitions grow, new regions 14 00:00:44,920 --> 00:00:47,199 Speaker 1: from the Nordics to the Middle East and Latin America 15 00:00:47,440 --> 00:00:50,240 Speaker 1: are positioning themselves to capture the next wave of development. 16 00:00:50,840 --> 00:00:53,400 Speaker 1: So what determines which markets win the race to host 17 00:00:53,400 --> 00:00:56,120 Speaker 1: the new generation of data centers? Today? I'm joined by 18 00:00:56,160 --> 00:00:59,360 Speaker 1: Lloyd Arnold, an analyst from BNF Technology and Innovations Team, 19 00:00:59,560 --> 00:01:02,840 Speaker 1: to review findings from their note data Center Development, Regional 20 00:01:02,920 --> 00:01:06,240 Speaker 1: Drivers and Hurdles. BNF clients can find this note, along 21 00:01:06,280 --> 00:01:08,880 Speaker 1: with other data center research by heading to BNF go 22 00:01:09,040 --> 00:01:12,000 Speaker 1: on the Bloomberg terminal or BNF dot com. If you'd 23 00:01:12,080 --> 00:01:14,840 Speaker 1: like to learn more about how BNF approaches strategy research 24 00:01:14,840 --> 00:01:18,160 Speaker 1: on the energy transition, including developments in commodity markets, trends 25 00:01:18,200 --> 00:01:21,000 Speaker 1: across different sectors, and the cross cutting technology is shaping 26 00:01:21,040 --> 00:01:23,960 Speaker 1: the future, you can find more information on BNEF dot 27 00:01:24,000 --> 00:01:25,800 Speaker 1: com and if you'd like to speak with a member 28 00:01:25,800 --> 00:01:28,400 Speaker 1: of our team about becoming a client, email US at 29 00:01:28,400 --> 00:01:31,720 Speaker 1: Sales dot BNF at Bloomberg dot net. But for now, 30 00:01:31,880 --> 00:01:39,320 Speaker 1: let's explore which market's a best position for new data centers. 31 00:01:43,120 --> 00:01:44,520 Speaker 1: Welcome to the show, Lloyd. 32 00:01:44,720 --> 00:01:46,160 Speaker 2: Morning, Tom, thanks for having me here. 33 00:01:46,720 --> 00:01:50,800 Speaker 1: We're here talking about a report that you recently published. 34 00:01:51,000 --> 00:01:54,760 Speaker 1: Just kick us off with explaining this report on data 35 00:01:54,760 --> 00:01:58,080 Speaker 1: center development drivers. What will we trying to accomplish there? 36 00:01:58,440 --> 00:02:01,880 Speaker 2: So we know there's at huge buildout of data center 37 00:02:01,880 --> 00:02:04,640 Speaker 2: capacity going on around the world. We hear a lot 38 00:02:04,840 --> 00:02:07,640 Speaker 2: about the US, and the US is an incredibly important market, 39 00:02:07,720 --> 00:02:10,520 Speaker 2: but this is happening in a huge number of markets 40 00:02:10,720 --> 00:02:13,080 Speaker 2: and the goal of thisss report was to think about 41 00:02:13,200 --> 00:02:15,799 Speaker 2: where is capacity now and where is it easiest to 42 00:02:15,840 --> 00:02:20,320 Speaker 2: bring new capacity online? So we evaluated seventeen markets across 43 00:02:20,400 --> 00:02:23,799 Speaker 2: the Americas and the mayor to understand what matters when 44 00:02:23,800 --> 00:02:26,120 Speaker 2: you're citing a data center and how do those markets 45 00:02:26,200 --> 00:02:26,720 Speaker 2: stack up? 46 00:02:26,960 --> 00:02:28,560 Speaker 1: Got it? So? I mean, I guess that leads us 47 00:02:28,600 --> 00:02:31,919 Speaker 1: straight into it, like what does matter when you've got 48 00:02:32,200 --> 00:02:34,400 Speaker 1: when you're trying to decide where to build a data center? 49 00:02:34,880 --> 00:02:38,480 Speaker 2: Lots of things, But through many, many conversations with data 50 00:02:38,480 --> 00:02:42,840 Speaker 2: center developers, we settled on five broad pillars. So this 51 00:02:42,960 --> 00:02:46,919 Speaker 2: is energy availability, this is land permitting, this is taxes, 52 00:02:47,280 --> 00:02:50,080 Speaker 2: fiber connectivity, that's a measure of how connected to the 53 00:02:50,080 --> 00:02:54,160 Speaker 2: Internet is a given market, and then existing live capacity, 54 00:02:54,240 --> 00:02:56,240 Speaker 2: So how many data centers are there in. 55 00:02:56,120 --> 00:03:01,200 Speaker 1: That market now? So those are the five factors energy, 56 00:03:01,280 --> 00:03:05,399 Speaker 1: land permitting, taxes, fiber live capacity already and are those 57 00:03:05,440 --> 00:03:08,200 Speaker 1: all sort of equally important to each other or would 58 00:03:08,200 --> 00:03:10,600 Speaker 1: we consider some to be more critical than others? 59 00:03:11,040 --> 00:03:15,120 Speaker 2: Some are definitely more critical. Energy was the most important 60 00:03:15,200 --> 00:03:18,560 Speaker 2: in our ranking because you might hear this phrase AI 61 00:03:18,840 --> 00:03:22,240 Speaker 2: is power constrained at the moment, The limiting factor in 62 00:03:22,280 --> 00:03:25,120 Speaker 2: most markets is how much power can a data center get, 63 00:03:25,320 --> 00:03:28,440 Speaker 2: but that's useless without land permitting. If you have nowhere 64 00:03:28,440 --> 00:03:30,639 Speaker 2: you can build a data center, doesn't matter if you've 65 00:03:30,639 --> 00:03:34,160 Speaker 2: got a gigawat of power. And so energy and land 66 00:03:34,440 --> 00:03:37,400 Speaker 2: where the two most important, with energy marginally more so. 67 00:03:37,680 --> 00:03:41,360 Speaker 2: Taxes used to be very important but has become less 68 00:03:41,400 --> 00:03:45,560 Speaker 2: important over time. And then fiber connectivity and existing market 69 00:03:45,680 --> 00:03:48,640 Speaker 2: size are quite correlated. So they were both scored in 70 00:03:48,680 --> 00:03:52,080 Speaker 2: the middle but made up a sizable chunk together, right. 71 00:03:52,040 --> 00:03:54,120 Speaker 1: Right, I get it. So you didn't want to overweight 72 00:03:54,200 --> 00:03:56,840 Speaker 1: them because you'll be double counting in that impact in 73 00:03:56,880 --> 00:04:00,480 Speaker 1: a cern sense. And the fiber connectivity does it also 74 00:04:00,560 --> 00:04:02,840 Speaker 1: depend on the kind of data center you're trying to build, 75 00:04:02,920 --> 00:04:06,320 Speaker 1: you know, the application, because I'm sure some applications fiber 76 00:04:06,400 --> 00:04:07,640 Speaker 1: is more important than others. 77 00:04:08,040 --> 00:04:11,800 Speaker 2: Absolutely, So we can kind of split data center workloads 78 00:04:12,000 --> 00:04:14,680 Speaker 2: into those which are latency sensitive and those which aren't. 79 00:04:14,960 --> 00:04:16,880 Speaker 2: You know, we're here in London, you're in New York. 80 00:04:17,240 --> 00:04:21,360 Speaker 2: We're connected virtually. If there was lag in that data center, 81 00:04:21,680 --> 00:04:24,520 Speaker 2: it would be a nightmare. A data center serving enterprise 82 00:04:24,520 --> 00:04:28,520 Speaker 2: workloads communication workloads has to be in a fiber hub 83 00:04:28,720 --> 00:04:33,440 Speaker 2: to get that rapid connection for emerging AI workloads, this 84 00:04:33,560 --> 00:04:36,200 Speaker 2: is less important. Maybe a year ago we were talking 85 00:04:36,279 --> 00:04:38,880 Speaker 2: about building data centers out in the desert far from 86 00:04:39,040 --> 00:04:43,040 Speaker 2: anything for AI training. That's kind of declined. But because 87 00:04:43,120 --> 00:04:45,960 Speaker 2: when we use AI model there is thinking time for 88 00:04:46,080 --> 00:04:48,839 Speaker 2: the model on top of maybe ten seconds for an answer, 89 00:04:48,920 --> 00:04:52,640 Speaker 2: we don't care that much about ten milliseconds of extra latency. 90 00:04:52,880 --> 00:04:55,920 Speaker 2: So AI facilities are being cited a little bit further 91 00:04:55,960 --> 00:04:58,000 Speaker 2: away from the traditional fiber hubs. 92 00:04:58,320 --> 00:05:02,039 Speaker 1: That's so interesting. It's it's opening up new possibilities in 93 00:05:02,080 --> 00:05:05,320 Speaker 1: a sense. And then finally, live capacity. I just want 94 00:05:05,360 --> 00:05:06,960 Speaker 1: to really, I mean, I know we talked about fiber 95 00:05:06,960 --> 00:05:09,360 Speaker 1: and live capacity together, but just explain a little bit 96 00:05:09,520 --> 00:05:13,040 Speaker 1: why live capacity matters so much? You know, why should 97 00:05:13,040 --> 00:05:15,640 Speaker 1: it matter how many data centers have already been built 98 00:05:15,640 --> 00:05:17,680 Speaker 1: in a region to building more, you know, because you 99 00:05:17,760 --> 00:05:20,440 Speaker 1: could look at it one way and say, oh, it 100 00:05:20,440 --> 00:05:22,880 Speaker 1: should be the other way around, because you know, all 101 00:05:22,880 --> 00:05:25,200 Speaker 1: the land, all the energy has been chomped up by 102 00:05:25,240 --> 00:05:27,839 Speaker 1: the ones that have proceeded. So why is live capacity 103 00:05:27,880 --> 00:05:30,880 Speaker 1: a positive factor for developing new data centers? 104 00:05:31,400 --> 00:05:35,000 Speaker 2: No? Absolutely, There's kind of two reasons live capacity matters. 105 00:05:35,400 --> 00:05:38,200 Speaker 2: The first is that it is an indicator of how 106 00:05:38,240 --> 00:05:40,760 Speaker 2: much demand there is in the market. Now, the biggest 107 00:05:40,800 --> 00:05:43,800 Speaker 2: data center markets in the world are largely financial hubs 108 00:05:43,800 --> 00:05:46,039 Speaker 2: and technology hubs, which need a lot of capacity, and 109 00:05:46,080 --> 00:05:50,159 Speaker 2: that's still true. But the other really important thing is 110 00:05:50,440 --> 00:05:52,560 Speaker 2: you need to be very skilled to build and operate 111 00:05:52,680 --> 00:05:56,680 Speaker 2: a data center. We wanted to include labor availability in 112 00:05:56,720 --> 00:06:00,760 Speaker 2: this ranking. Unfortunately that's really hard to quantify. Maybe a 113 00:06:00,839 --> 00:06:05,279 Speaker 2: country reports how many electricians work in the market, but 114 00:06:05,400 --> 00:06:07,560 Speaker 2: to build a data center, you only a subset of 115 00:06:07,600 --> 00:06:11,720 Speaker 2: those who have particular qualifications. We couldn't find systemic way 116 00:06:11,920 --> 00:06:14,799 Speaker 2: of tracking that that would be comparable between the markets. 117 00:06:15,080 --> 00:06:17,400 Speaker 2: But the fact that the data center was built last 118 00:06:17,440 --> 00:06:19,719 Speaker 2: week is a pretty good indicator there are people in 119 00:06:19,760 --> 00:06:21,599 Speaker 2: the market that know how to build a data center, 120 00:06:21,760 --> 00:06:25,080 Speaker 2: and so it was used as a proxy for that, unless. 121 00:06:24,800 --> 00:06:27,560 Speaker 1: It's a really shoddy data center. I don't know what 122 00:06:27,600 --> 00:06:30,440 Speaker 1: a shoddy DA data center looks like. So we've kind 123 00:06:30,440 --> 00:06:33,680 Speaker 1: of a service space. We've got these five different factors. 124 00:06:33,880 --> 00:06:36,719 Speaker 1: You guys made a beautiful little bar chart where you 125 00:06:36,760 --> 00:06:39,880 Speaker 1: stacked up the scores that you assigned to different markets, 126 00:06:40,160 --> 00:06:41,760 Speaker 1: but you put them all in order of you know, 127 00:06:41,839 --> 00:06:45,240 Speaker 1: the one that got the highest score as a market 128 00:06:45,240 --> 00:06:47,919 Speaker 1: for data centers versus the ones that got the lower score. 129 00:06:48,040 --> 00:06:50,000 Speaker 1: So tell us what was at the top of the list. 130 00:06:50,279 --> 00:06:52,600 Speaker 2: The top of the list was the state of Virginia 131 00:06:52,720 --> 00:06:54,720 Speaker 2: on the east coast of the US got it, and 132 00:06:54,880 --> 00:06:58,720 Speaker 2: second on the list Texas. So two US states coming 133 00:06:58,720 --> 00:06:59,920 Speaker 2: in ahead of many countries. 134 00:07:00,279 --> 00:07:02,480 Speaker 1: Gold and silver go to the US, And I suppose 135 00:07:02,480 --> 00:07:06,240 Speaker 1: people who follow this market shouldn't be surprised. And I 136 00:07:06,279 --> 00:07:09,040 Speaker 1: know that a lot of the point of this report 137 00:07:09,240 --> 00:07:14,240 Speaker 1: is to have people thinking beyond the US because we 138 00:07:14,280 --> 00:07:16,880 Speaker 1: already know so many of the world's data centers are 139 00:07:17,080 --> 00:07:20,000 Speaker 1: in the US and are being built in these two 140 00:07:20,080 --> 00:07:23,800 Speaker 1: particular states in particular. But just tell us the story 141 00:07:23,800 --> 00:07:26,480 Speaker 1: a little bit about Virginia and Texas. Why it is 142 00:07:26,520 --> 00:07:28,480 Speaker 1: that they come in so high on the list. 143 00:07:28,960 --> 00:07:31,760 Speaker 2: Yeah, so there's a couple of different dynamics going on 144 00:07:32,000 --> 00:07:36,120 Speaker 2: between the states. Virginia, which is home to over eleven 145 00:07:36,200 --> 00:07:40,000 Speaker 2: percent of the live capacity that we track globally, is 146 00:07:40,240 --> 00:07:45,560 Speaker 2: very very important because historically it was home to one 147 00:07:45,560 --> 00:07:49,000 Speaker 2: of the first ever Internet exchange points. So this is 148 00:07:49,000 --> 00:07:51,400 Speaker 2: a bit like an old fashioned phone switchboard for the Internet. 149 00:07:51,520 --> 00:07:55,960 Speaker 2: Virginia hosted one very very early. Virginia also introduced tax 150 00:07:56,000 --> 00:07:58,480 Speaker 2: breaks for data centers. That gave a one hundred percent 151 00:07:58,560 --> 00:08:03,040 Speaker 2: tax break on sales tax for data center infrastructure, and 152 00:08:03,480 --> 00:08:07,119 Speaker 2: that kind of gave the industry a foothold into the state. 153 00:08:07,440 --> 00:08:09,920 Speaker 2: And as we allude to the start, there's a snowball 154 00:08:09,960 --> 00:08:12,880 Speaker 2: effect and there's a network effect, so more capacity went there. 155 00:08:13,040 --> 00:08:17,160 Speaker 2: It also benefits from where it is geographically. It can 156 00:08:17,280 --> 00:08:21,880 Speaker 2: serve the eastern seaboard of the US and has subsea 157 00:08:22,000 --> 00:08:27,400 Speaker 2: cables connecting it to Europe. So that's really what drove Virginia, 158 00:08:27,480 --> 00:08:29,760 Speaker 2: and it's home to a huge amount of cloud and 159 00:08:30,000 --> 00:08:36,360 Speaker 2: enterprise infrastructure. Texas is a slightly different story, and really 160 00:08:36,559 --> 00:08:39,679 Speaker 2: the draw in Texas is there's a lot of land 161 00:08:39,880 --> 00:08:43,240 Speaker 2: and there's a lot of cheap power. We've seen significant 162 00:08:43,240 --> 00:08:48,400 Speaker 2: amounts of cryptocurrency mining infrastructure, so this is high performance 163 00:08:48,440 --> 00:08:53,520 Speaker 2: infrastructure deployed there. And as the AI build out is 164 00:08:53,559 --> 00:08:57,120 Speaker 2: happening and accelerating, Texas is a really great place to 165 00:08:57,240 --> 00:09:01,560 Speaker 2: build these huge power hungry aa I focused data centers 166 00:09:01,600 --> 00:09:03,920 Speaker 2: that can do very large workloads. 167 00:09:04,280 --> 00:09:08,280 Speaker 1: This is important to establish because this is what you know, 168 00:09:08,360 --> 00:09:10,760 Speaker 1: all of these other markets we're talking about, This is 169 00:09:10,760 --> 00:09:14,840 Speaker 1: what they're competing against two markets that have various factors 170 00:09:14,960 --> 00:09:17,520 Speaker 1: going in their favor, you know, on a sort of 171 00:09:17,559 --> 00:09:20,760 Speaker 1: an inherent level. I would say that the sort of 172 00:09:20,880 --> 00:09:23,880 Speaker 1: the factors such as energy, land permitting, and taxes, but 173 00:09:23,920 --> 00:09:27,920 Speaker 1: then also have the incumbency bonus that you know of 174 00:09:28,240 --> 00:09:31,240 Speaker 1: already having that head start. So the thing that's really 175 00:09:31,280 --> 00:09:35,640 Speaker 1: interesting about, you know, everything you've just described with Virginia 176 00:09:35,679 --> 00:09:38,480 Speaker 1: and Texas is not just the fact that they have 177 00:09:38,559 --> 00:09:43,880 Speaker 1: some inherent advantages that are definitely real but don't necessarily 178 00:09:43,920 --> 00:09:46,920 Speaker 1: stand out to completely set them apart from everyone else. 179 00:09:46,960 --> 00:09:49,560 Speaker 1: What really sets them apart from everyone else is that 180 00:09:49,720 --> 00:09:53,120 Speaker 1: they've had this head start. It's where stuff has already 181 00:09:53,160 --> 00:09:57,559 Speaker 1: been built. And then I reference this chart you made earlier. 182 00:09:57,720 --> 00:09:59,960 Speaker 1: When I look at this chart, I ignore the fine 183 00:10:00,280 --> 00:10:03,400 Speaker 1: bar in your sort of rankings, the live capacity bar, 184 00:10:03,679 --> 00:10:06,000 Speaker 1: the one that is all about the kind of historical 185 00:10:06,080 --> 00:10:09,360 Speaker 1: advantage different markets have. Virginia and Texts are still ranked 186 00:10:09,360 --> 00:10:11,880 Speaker 1: really high on the list, so they do have natural 187 00:10:11,880 --> 00:10:15,760 Speaker 1: advantages beyond incumbency, but they're not the top. The top 188 00:10:15,840 --> 00:10:20,679 Speaker 1: markets are Saudi Arabia, United Arab Emirates, and Brazil. So 189 00:10:21,559 --> 00:10:24,199 Speaker 1: maybe we'll start with the markets in the Middle East. 190 00:10:24,240 --> 00:10:27,079 Speaker 1: Tell us a little bit about those both in terms 191 00:10:27,080 --> 00:10:31,440 Speaker 1: of why they have such strong advantages and you know, 192 00:10:31,480 --> 00:10:33,160 Speaker 1: do we have any insights on what's actually going on 193 00:10:33,200 --> 00:10:35,480 Speaker 1: there in terms of data center development. Yeah. 194 00:10:35,640 --> 00:10:38,600 Speaker 2: So these are markets where there's a lot of land, 195 00:10:38,880 --> 00:10:43,800 Speaker 2: it's relatively easy to permit to use this land, and 196 00:10:44,160 --> 00:10:46,559 Speaker 2: there's a lot of excess power that's very, very cheap. 197 00:10:47,040 --> 00:10:50,600 Speaker 2: This all contributed to quite a big score, but there's 198 00:10:50,640 --> 00:10:54,640 Speaker 2: a gap in our scoring framework, and this gap does 199 00:10:54,760 --> 00:10:57,640 Speaker 2: create more risk and more uncertainty developing there. And that's 200 00:10:57,720 --> 00:11:03,760 Speaker 2: geopolitics we didn't include in our ranking system because it's 201 00:11:03,760 --> 00:11:07,840 Speaker 2: incredibly hard to quantify numerically and objectively, which was very 202 00:11:07,920 --> 00:11:10,839 Speaker 2: very important as we developed this. It's also much more 203 00:11:10,880 --> 00:11:12,920 Speaker 2: fast moving than a lot of the things. You know, 204 00:11:12,960 --> 00:11:16,600 Speaker 2: we were looking at the five year change in installed 205 00:11:16,640 --> 00:11:21,360 Speaker 2: capacity in a market, whereas geopolitical environments can change much 206 00:11:21,360 --> 00:11:23,760 Speaker 2: more rapidly. But the Middle East in particular that you've 207 00:11:23,800 --> 00:11:27,439 Speaker 2: brought up first. Under the Biden administration, there were export 208 00:11:27,480 --> 00:11:31,480 Speaker 2: controls on advanced semiconductors to the region. They were lifted 209 00:11:31,679 --> 00:11:35,200 Speaker 2: under the Trump administration, but there's no guarantee that that 210 00:11:35,280 --> 00:11:38,640 Speaker 2: will stay and so that creates extra risk for any 211 00:11:38,640 --> 00:11:41,400 Speaker 2: developer that wants to extend into the market, and so 212 00:11:41,480 --> 00:11:43,840 Speaker 2: the question of will you actually be able to get 213 00:11:43,880 --> 00:11:46,479 Speaker 2: off takers for that capacity remains. 214 00:11:47,360 --> 00:11:52,160 Speaker 1: That's so interesting as factors that penalize Saudi Arabia and 215 00:11:52,280 --> 00:11:56,640 Speaker 1: UAE are not to do with the inherent advantage of 216 00:11:56,920 --> 00:11:59,559 Speaker 1: building a data center there there, they're more reflective of 217 00:11:59,600 --> 00:12:03,520 Speaker 1: this recingly fragmented world we see in terms of the 218 00:12:04,200 --> 00:12:07,640 Speaker 1: global economy. What one thing I'm curious about is that 219 00:12:07,679 --> 00:12:09,960 Speaker 1: there seems to be two factors here. One is the 220 00:12:10,000 --> 00:12:13,200 Speaker 1: ability to get the equipment to build the data centers 221 00:12:13,520 --> 00:12:16,199 Speaker 1: and whether there are trade restrictions there and you mentioned 222 00:12:16,280 --> 00:12:18,600 Speaker 1: you know, some of them were lifted under Trump, but 223 00:12:18,679 --> 00:12:21,280 Speaker 1: there's always a chance they would come in. But I mean, 224 00:12:21,400 --> 00:12:24,240 Speaker 1: one question I have is is does this equipment have 225 00:12:24,320 --> 00:12:26,760 Speaker 1: to come from the US and there are other countries 226 00:12:26,920 --> 00:12:30,960 Speaker 1: interested in developing data centers in the Middle East, And 227 00:12:31,000 --> 00:12:35,640 Speaker 1: then is there scope for these Middle Eastern countries to 228 00:12:35,880 --> 00:12:38,720 Speaker 1: be the kind of the hub for non US data 229 00:12:38,760 --> 00:12:39,560 Speaker 1: center development. 230 00:12:39,920 --> 00:12:42,560 Speaker 2: Yeah. Absolutely, So there's a few things there to unpack, 231 00:12:42,640 --> 00:12:45,439 Speaker 2: and the first to spout supply chains, a trend we're 232 00:12:45,440 --> 00:12:50,559 Speaker 2: seeing especially in electrical equipment supply chains is incumbents are 233 00:12:50,760 --> 00:12:54,640 Speaker 2: acquiring to develop a regional footprint. We are also talking 234 00:12:54,679 --> 00:12:57,040 Speaker 2: to lots of companies who want to get in on 235 00:12:57,240 --> 00:13:01,199 Speaker 2: this data center trade and are looking at exploring internew 236 00:13:01,240 --> 00:13:04,600 Speaker 2: markets and maybe supplying data centers when that hasn't been 237 00:13:04,640 --> 00:13:09,280 Speaker 2: a primary market historically. In terms of silicon, So the 238 00:13:09,360 --> 00:13:13,720 Speaker 2: chips we need to do advanced computation, that still is 239 00:13:13,800 --> 00:13:18,960 Speaker 2: a primarily US story. In Nvidia has an absolutely huge 240 00:13:19,120 --> 00:13:23,160 Speaker 2: share of the market for AI INFRINCE chips that's the 241 00:13:23,240 --> 00:13:27,439 Speaker 2: chip used to build and run AI models. And whilst 242 00:13:27,440 --> 00:13:32,280 Speaker 2: there are competitors Google Amazon have in house solutions, are 243 00:13:32,360 --> 00:13:38,000 Speaker 2: still from American companies, so that will remain a constraint 244 00:13:38,120 --> 00:13:41,160 Speaker 2: at least in the coming years because it takes time 245 00:13:41,440 --> 00:13:45,280 Speaker 2: to develop the architectures to build this silicon in terms 246 00:13:45,360 --> 00:13:45,920 Speaker 2: of demand. 247 00:13:46,240 --> 00:13:48,079 Speaker 1: You're absolutely right. So many of the. 248 00:13:48,040 --> 00:13:52,560 Speaker 2: Companies developing foundational models. These are models like chat GPT 249 00:13:52,720 --> 00:13:55,760 Speaker 2: five point two, claud Opens four point six, the huge 250 00:13:55,800 --> 00:13:58,160 Speaker 2: air models that other tools are based on. Most of 251 00:13:58,160 --> 00:14:02,840 Speaker 2: those companies are based in However, as a growing interest 252 00:14:03,200 --> 00:14:08,040 Speaker 2: in sovereign AI, lots of countries Saudi Arabia and the 253 00:14:08,080 --> 00:14:13,320 Speaker 2: UA included have sovereign AI strategy that will involve developing 254 00:14:13,480 --> 00:14:18,040 Speaker 2: infrastructure on shore and developing software and tools on shore. 255 00:14:18,360 --> 00:14:21,880 Speaker 2: So whilst there are constraints on chip supply and there 256 00:14:21,880 --> 00:14:25,760 Speaker 2: are constraints on these huge American companies wanting that infrastructure, 257 00:14:26,040 --> 00:14:30,400 Speaker 2: midterm we'd also see a boost as more services are 258 00:14:30,440 --> 00:14:34,080 Speaker 2: brought on shore and more users in those markets make 259 00:14:34,200 --> 00:14:35,520 Speaker 2: use of advanced A tools. 260 00:14:36,480 --> 00:14:41,040 Speaker 1: There's so interesting hearing that as someone who's spent their 261 00:14:41,080 --> 00:14:45,080 Speaker 1: career being an energy analyst, and you know, as energy analysts, 262 00:14:45,120 --> 00:14:49,480 Speaker 1: we're having to learn all about artificial intelligence because it's 263 00:14:49,560 --> 00:14:53,440 Speaker 1: so impactful on the energy sector. But actually, leaving aside 264 00:14:53,480 --> 00:14:56,120 Speaker 1: its impact on the energy sector, there's this other angle 265 00:14:56,120 --> 00:14:59,520 Speaker 1: that's interesting is we've spent so many decades talking about 266 00:14:59,640 --> 00:15:04,200 Speaker 1: energy security, and it's why countries are so keen to 267 00:15:04,280 --> 00:15:07,200 Speaker 1: develop their own domestic energy supplies, even if you know 268 00:15:07,240 --> 00:15:09,960 Speaker 1: there are trade partners globally who could provide energy a 269 00:15:10,000 --> 00:15:12,640 Speaker 1: lot more cheaply. And you know, you've seen the importance 270 00:15:12,680 --> 00:15:16,120 Speaker 1: of that with Europe and its dependence on Russia for 271 00:15:16,240 --> 00:15:19,040 Speaker 1: gas and how that really has come to the head 272 00:15:19,080 --> 00:15:22,000 Speaker 1: in the last five years or so. And so now 273 00:15:22,040 --> 00:15:25,280 Speaker 1: we see some of the equivalent for artificial intelligence. This 274 00:15:25,360 --> 00:15:30,240 Speaker 1: idea of sovereign AI is this cognition and the power 275 00:15:30,280 --> 00:15:34,160 Speaker 1: that that brings is something that countries are wary of 276 00:15:34,280 --> 00:15:36,920 Speaker 1: just intrusting to one global partner because there's so much 277 00:15:36,960 --> 00:15:39,680 Speaker 1: power that comes with that. And then I suppose, you know, 278 00:15:39,760 --> 00:15:42,680 Speaker 1: my question is, is then given that in the same 279 00:15:42,680 --> 00:15:45,400 Speaker 1: way that bringing in the Middle East Again, the Middle 280 00:15:45,400 --> 00:15:48,160 Speaker 1: East is kind of the heartland for oil, and that 281 00:15:48,800 --> 00:15:51,480 Speaker 1: was a problem to so many countries being so dependent 282 00:15:51,560 --> 00:15:53,520 Speaker 1: on Middle East and oil. The sort of the Saudi 283 00:15:53,520 --> 00:15:58,360 Speaker 1: Arabia of artificial intelligence is the US. So at the moment, 284 00:15:58,680 --> 00:16:01,480 Speaker 1: sort of not being a a US market is a 285 00:16:01,520 --> 00:16:04,760 Speaker 1: penalty because so many much of the equipment, so much 286 00:16:04,760 --> 00:16:06,760 Speaker 1: the demand is coming from the US. But it's also 287 00:16:06,840 --> 00:16:09,600 Speaker 1: a motivating factor for these market markets to develop. So 288 00:16:09,720 --> 00:16:11,880 Speaker 1: even some of these ones in your list that don't 289 00:16:11,920 --> 00:16:15,080 Speaker 1: score as highly as Virginuine Texas, they have an inherent 290 00:16:15,400 --> 00:16:17,960 Speaker 1: still drive to build. Which is this idea of sovereign 291 00:16:17,960 --> 00:16:20,520 Speaker 1: AI is that what is behind a lot of this 292 00:16:20,920 --> 00:16:21,880 Speaker 1: In a sense. 293 00:16:22,120 --> 00:16:26,400 Speaker 2: I think that that's a factor, but it's complex, as 294 00:16:26,800 --> 00:16:28,160 Speaker 2: is pretty much every answer. 295 00:16:28,560 --> 00:16:31,640 Speaker 1: It's not the only factor, yeah, for sure. So then 296 00:16:31,760 --> 00:16:35,320 Speaker 1: bringing in the other market I mentioned that seems data 297 00:16:35,360 --> 00:16:38,640 Speaker 1: center ready but has relatively few built to date is Brazil. 298 00:16:39,000 --> 00:16:40,920 Speaker 1: So tell us a little bit about the story there. 299 00:16:41,240 --> 00:16:45,840 Speaker 2: Yeah, so Brazil is already a regional fiber hub for 300 00:16:45,920 --> 00:16:50,000 Speaker 2: Latin America, so a huge amount of the Internet traffic 301 00:16:50,080 --> 00:16:53,720 Speaker 2: that's being routed through Latin is going to be going 302 00:16:53,840 --> 00:16:58,480 Speaker 2: through exchange points in Brazil. This is expected to continue 303 00:16:58,560 --> 00:17:02,280 Speaker 2: because of these network effects and infrastructure, and a huge 304 00:17:02,280 --> 00:17:06,119 Speaker 2: tailwind is that the population in Brazil in particular and 305 00:17:06,200 --> 00:17:11,040 Speaker 2: surrounding countries more generally is young and is rapidly getting richer, 306 00:17:11,080 --> 00:17:15,000 Speaker 2: so it's spending more time online. That will drive domestic 307 00:17:15,040 --> 00:17:20,200 Speaker 2: demand for cloud services and for Internet services such as 308 00:17:20,400 --> 00:17:23,080 Speaker 2: social media. But the other thing we're starting to see 309 00:17:23,080 --> 00:17:27,560 Speaker 2: in Brazil are companies developing or planning to develop what 310 00:17:27,600 --> 00:17:31,080 Speaker 2: we might call the more speculative AI projects. And the 311 00:17:31,240 --> 00:17:34,600 Speaker 2: driving idea is there are constraints in the US. You know, 312 00:17:34,640 --> 00:17:37,480 Speaker 2: we've spoken that things are great in Virginia and Texas, 313 00:17:37,840 --> 00:17:42,000 Speaker 2: but there are still headwinds, and so as AI labs 314 00:17:42,080 --> 00:17:45,680 Speaker 2: scale rapidly, there is hope from some developers that they'll 315 00:17:45,680 --> 00:17:49,960 Speaker 2: be able to sell capacity in Brazil to American tech firms. 316 00:17:50,200 --> 00:17:55,320 Speaker 2: Whether or not that manifests is uncertain, and Brazil has 317 00:17:55,480 --> 00:17:58,239 Speaker 2: this huge pipeline. So if we look at what's been 318 00:17:58,240 --> 00:18:01,800 Speaker 2: announced in Brazil versus what there is, it has by 319 00:18:01,880 --> 00:18:05,640 Speaker 2: far the biggest ratio of pipeline to current but about 320 00:18:05,720 --> 00:18:09,280 Speaker 2: ninety percent of that is in two mega projects that 321 00:18:09,320 --> 00:18:12,760 Speaker 2: are trying to serve USAI in the way I've mentioned, 322 00:18:13,040 --> 00:18:16,840 Speaker 2: and so that's just some uncertainty as the market progresses, 323 00:18:16,960 --> 00:18:20,479 Speaker 2: although the growing domestic demand is likely to be very strong. 324 00:18:20,720 --> 00:18:23,600 Speaker 1: So you mentioned two projects, so it's quite lumpy at 325 00:18:23,600 --> 00:18:26,960 Speaker 1: the moment. It's not necessarily looking like it's growing organically, 326 00:18:27,000 --> 00:18:29,520 Speaker 1: but it could be a huge boost of these mega 327 00:18:29,520 --> 00:18:32,680 Speaker 1: projects come through and provide real momentum there for. 328 00:18:32,680 --> 00:18:36,159 Speaker 2: Sure, and we were definitely considered medium term. There'd be 329 00:18:36,200 --> 00:18:38,760 Speaker 2: a lot of growth in a similar way to some 330 00:18:38,840 --> 00:18:41,679 Speaker 2: markets that weren't included in this note but will be 331 00:18:41,760 --> 00:18:43,680 Speaker 2: profiled in upcoming features. 332 00:18:43,880 --> 00:18:45,040 Speaker 1: Through Southeast Asia. 333 00:18:45,119 --> 00:18:48,439 Speaker 2: You know, as the population used Internet more heavily for 334 00:18:48,840 --> 00:18:52,639 Speaker 2: all parts of life, that stimulates demand for cloud services 335 00:18:52,880 --> 00:18:57,120 Speaker 2: and enterprise services, which will be correlated with and will 336 00:18:57,200 --> 00:19:00,639 Speaker 2: drive lots of data center development in Brazil. 337 00:19:00,760 --> 00:19:02,720 Speaker 1: A couple of years Ago. I had to give a 338 00:19:02,760 --> 00:19:06,520 Speaker 1: presentation for an event we were doing internally, and I 339 00:19:06,560 --> 00:19:09,679 Speaker 1: had to like read up and kind of get smart 340 00:19:09,680 --> 00:19:12,240 Speaker 1: on Brazil really quickly. But at the time I was 341 00:19:12,240 --> 00:19:15,960 Speaker 1: really struck by how well positioned Brazil was, in a 342 00:19:16,080 --> 00:19:20,960 Speaker 1: large part because it has an almost zero carbon grid 343 00:19:21,200 --> 00:19:24,919 Speaker 1: because of the huge amounts of hydro that are already 344 00:19:24,920 --> 00:19:28,439 Speaker 1: built in Brazil, and so therefore the ability to develop 345 00:19:28,480 --> 00:19:32,879 Speaker 1: low carbon industries. You know, Brazil extremely well positioned. Do 346 00:19:32,920 --> 00:19:35,320 Speaker 1: you think there's this one thing that might be in 347 00:19:35,359 --> 00:19:38,960 Speaker 1: the back of some of these US companies developing AI 348 00:19:39,440 --> 00:19:41,879 Speaker 1: is that when climate change does rise up on a 349 00:19:42,000 --> 00:19:45,600 Speaker 1: gender again and people already question the environmental impact of AI, 350 00:19:45,840 --> 00:19:48,520 Speaker 1: these Brazilian data centers, you may be able to claim 351 00:19:48,640 --> 00:19:52,280 Speaker 1: that this is this is green or greener AI than 352 00:19:52,320 --> 00:19:54,080 Speaker 1: the rest or Do you think that's not really a 353 00:19:54,119 --> 00:19:55,880 Speaker 1: motivating factor at the moment. 354 00:19:55,960 --> 00:19:59,720 Speaker 2: I think it's not nothing the largest That's a very 355 00:20:00,040 --> 00:20:01,840 Speaker 2: traumatic answer, That's what I was saying. 356 00:20:02,800 --> 00:20:06,280 Speaker 1: Your question was wide of the mark. That's a long 357 00:20:06,320 --> 00:20:07,600 Speaker 1: time asking it as well. 358 00:20:09,119 --> 00:20:11,840 Speaker 2: No, it is a thing. These the big tech companies 359 00:20:12,160 --> 00:20:16,400 Speaker 2: have very ambitious corporate stainability goals, and you know, many 360 00:20:16,440 --> 00:20:20,119 Speaker 2: of them who invested huge amounts in bringing renewable capacity 361 00:20:20,320 --> 00:20:25,359 Speaker 2: online through PPAs. Google, for example, have invested a lot 362 00:20:25,520 --> 00:20:28,280 Speaker 2: and worked a lot on twenty four to seven matching 363 00:20:28,760 --> 00:20:33,600 Speaker 2: for renewable generation and their data center capacity. That historically 364 00:20:33,600 --> 00:20:37,240 Speaker 2: has been important to these companies, and it's been important 365 00:20:37,240 --> 00:20:39,679 Speaker 2: in the tech sector more broadly. We spend a lot 366 00:20:39,680 --> 00:20:43,480 Speaker 2: of time speaking to data center developers who maybe want 367 00:20:43,520 --> 00:20:46,200 Speaker 2: to rent out capacity to a smaller tech firm. Most 368 00:20:46,200 --> 00:20:50,919 Speaker 2: of those historically have prioritized clean energy, but we've really 369 00:20:50,920 --> 00:20:53,719 Speaker 2: seen a shift in the last couple of years because, 370 00:20:53,760 --> 00:20:55,479 Speaker 2: as I kind of said at the start, the sector 371 00:20:55,560 --> 00:21:00,879 Speaker 2: is increasingly power constrained, and there's this zero sum that 372 00:21:01,119 --> 00:21:04,720 Speaker 2: more compute will mean bigger market share and better air 373 00:21:04,800 --> 00:21:08,800 Speaker 2: models now, and the first company to hit some kind 374 00:21:08,840 --> 00:21:12,600 Speaker 2: of benchmarks and have a threshold will win very big 375 00:21:12,760 --> 00:21:17,520 Speaker 2: and the others will fall away. That belief is causing 376 00:21:17,760 --> 00:21:21,880 Speaker 2: the cleanness of energy to drop down in ranking's short term, 377 00:21:21,960 --> 00:21:25,080 Speaker 2: and we're actually seeing a lot of on site gas 378 00:21:25,480 --> 00:21:28,400 Speaker 2: and foster generation more generally to support this build out. 379 00:21:28,840 --> 00:21:30,119 Speaker 1: And is that also true in Brazil? 380 00:21:30,119 --> 00:21:34,000 Speaker 2: Would you say in Brazil not yet, So that's typically 381 00:21:34,040 --> 00:21:38,280 Speaker 2: happening in the markets which have really really hot demand, 382 00:21:38,520 --> 00:21:40,120 Speaker 2: which are going to be indicator that as a market 383 00:21:40,240 --> 00:21:43,520 Speaker 2: scoring well on fiber and well on live capacity that 384 00:21:43,800 --> 00:21:48,160 Speaker 2: are struggling to energize and meet that demand. Brazil does 385 00:21:48,240 --> 00:21:51,040 Speaker 2: have this excess when you're learn to do that you've highlighted, 386 00:21:51,200 --> 00:21:54,120 Speaker 2: and that's one of the key value propositions of these 387 00:21:54,160 --> 00:21:57,520 Speaker 2: mega projects that are being planned there. But the real 388 00:21:57,640 --> 00:22:02,160 Speaker 2: question is will attech company the US or elsewhere want 389 00:22:02,160 --> 00:22:04,679 Speaker 2: to lease that capacity or would they rather have it 390 00:22:04,840 --> 00:22:08,280 Speaker 2: at home fossil powered. For now, the at home fossil 391 00:22:08,280 --> 00:22:11,240 Speaker 2: powered seems to be winning. The other thing is as 392 00:22:11,240 --> 00:22:14,000 Speaker 2: we look at how the emitsionality of these companies as 393 00:22:14,119 --> 00:22:17,919 Speaker 2: being reported, the emissions or like there are from the 394 00:22:17,960 --> 00:22:20,919 Speaker 2: power they're purchasing for data centers wouldn't be coming into 395 00:22:21,160 --> 00:22:23,920 Speaker 2: their reporting. That's gonna be their scope three, which isn't 396 00:22:24,040 --> 00:22:26,680 Speaker 2: really what most of them are going to be interested in, 397 00:22:26,960 --> 00:22:29,720 Speaker 2: and so it's a relatively low priority. 398 00:22:30,280 --> 00:22:32,840 Speaker 1: Let's move on to some other markets, because you know, 399 00:22:32,960 --> 00:22:35,359 Speaker 1: I think you reviewed seventeen markets in your report, and 400 00:22:35,359 --> 00:22:37,560 Speaker 1: we obviously are not going to talk about every single 401 00:22:37,600 --> 00:22:40,680 Speaker 1: one in all of them Menu Shai and we've talked 402 00:22:40,720 --> 00:22:42,400 Speaker 1: about the ones that there were really at the top 403 00:22:42,440 --> 00:22:45,720 Speaker 1: of the list, So Virginia, Texas, Saudi Arabia, Ua, Brazil. 404 00:22:45,800 --> 00:22:49,280 Speaker 1: That's what was the top five. Nine. I think of 405 00:22:49,320 --> 00:22:52,440 Speaker 1: the seventeen markets you reviewed were in Europe, so none 406 00:22:52,480 --> 00:22:54,400 Speaker 1: of them were in the top five. But it's still 407 00:22:54,440 --> 00:22:58,880 Speaker 1: quite significant that there's so many that you included in Europe, 408 00:22:58,920 --> 00:23:01,960 Speaker 1: and so maybe none of them is quite the sort 409 00:23:01,960 --> 00:23:03,879 Speaker 1: of home run that some of the ones in the 410 00:23:03,880 --> 00:23:06,480 Speaker 1: top five are. But tell us a little bit about 411 00:23:06,560 --> 00:23:08,800 Speaker 1: just in general terms that what the picture in Europe 412 00:23:08,840 --> 00:23:10,560 Speaker 1: is like, and then maybe we can dive into some 413 00:23:10,720 --> 00:23:14,400 Speaker 1: of the more interesting or more significant markets that you've noticed. 414 00:23:14,800 --> 00:23:18,840 Speaker 2: Yeah. Absolutely, so, Yeah, we'll start at this conversation. We 415 00:23:18,840 --> 00:23:21,680 Speaker 2: were saying Virginia and Texas. Part of where they're really 416 00:23:21,720 --> 00:23:24,800 Speaker 2: attractive and doing really well is because they have huge 417 00:23:24,800 --> 00:23:29,040 Speaker 2: amounts of existing capacity. In Europe, we're kind of seeing 418 00:23:29,280 --> 00:23:33,879 Speaker 2: the converse, and so the traditional tier one markets in Europe. 419 00:23:34,000 --> 00:23:37,600 Speaker 2: The top data center markets have formed in the metropolitan 420 00:23:37,680 --> 00:23:42,800 Speaker 2: areas around Frankfurt, London, Amsterdam, Paris, and Dublin which has 421 00:23:42,920 --> 00:23:47,719 Speaker 2: lovely academ Flat d These are major technology and financial hubs, 422 00:23:48,040 --> 00:23:51,960 Speaker 2: and data center capacity has built there because those industries 423 00:23:52,000 --> 00:23:55,240 Speaker 2: needed it, and they needed it early. They've now evolved 424 00:23:55,440 --> 00:24:01,000 Speaker 2: into key centers for Internet exchange and for computing, so 425 00:24:01,000 --> 00:24:04,520 Speaker 2: they're still really attractive, but it's hard to get energy 426 00:24:04,640 --> 00:24:08,800 Speaker 2: in these markets. It's often hard and complicated to get 427 00:24:08,880 --> 00:24:13,640 Speaker 2: permission to build, and local attitudes to projects are really 428 00:24:13,720 --> 00:24:16,600 Speaker 2: quite hostile compared to other places in Europe, and so 429 00:24:17,359 --> 00:24:20,639 Speaker 2: flat D is kind of still very attractive, but challenging. 430 00:24:20,880 --> 00:24:24,040 Speaker 2: And then we spoke about our European growth markets. So 431 00:24:24,480 --> 00:24:28,960 Speaker 2: in the Nordics and in Iberia we see good land availability, 432 00:24:29,280 --> 00:24:34,840 Speaker 2: relatively easier permitting processes, and relatively better energy availability. They 433 00:24:34,880 --> 00:24:39,080 Speaker 2: don't yet have the fiber infrastructure that the flap D 434 00:24:39,440 --> 00:24:43,640 Speaker 2: hubs intier One do, but they have many other structural advantages. 435 00:24:43,880 --> 00:24:46,480 Speaker 2: And so you know, one line summary of what we'd 436 00:24:46,560 --> 00:24:49,880 Speaker 2: expect to see in Europe. Cloud compute and enterprise compute 437 00:24:50,080 --> 00:24:54,560 Speaker 2: stays in or near flap D, AI compute and large 438 00:24:54,600 --> 00:24:59,720 Speaker 2: scale deployments move more towards Iberia or up into the Nordics. 439 00:25:00,520 --> 00:25:02,600 Speaker 1: So I take your point that the sort of the 440 00:25:02,640 --> 00:25:05,240 Speaker 1: traditional data centers are likely to get built in flat D, 441 00:25:05,640 --> 00:25:09,919 Speaker 1: and then the sort of the emerging applications like the 442 00:25:09,960 --> 00:25:13,840 Speaker 1: Nordics and Spain look pretty promising. And actually, you know, 443 00:25:14,160 --> 00:25:17,159 Speaker 1: I head up our global power markets work. One of 444 00:25:17,160 --> 00:25:19,720 Speaker 1: our analysts is currently working on a report and I'm 445 00:25:19,760 --> 00:25:21,600 Speaker 1: sure you've spoken with her a lot about this is 446 00:25:21,880 --> 00:25:24,560 Speaker 1: about data centers being built in the Nordics, or it's 447 00:25:24,560 --> 00:25:26,560 Speaker 1: about power demanding the Nordics, but you can't e come 448 00:25:26,600 --> 00:25:29,080 Speaker 1: are the data centers, So that's really interesting. What I'm 449 00:25:29,119 --> 00:25:32,960 Speaker 1: kind of curious about is is France, which came kind 450 00:25:32,960 --> 00:25:34,879 Speaker 1: of out of all the European countries, came top of 451 00:25:34,880 --> 00:25:38,480 Speaker 1: your rankings. Is France position at least to kind of 452 00:25:38,520 --> 00:25:42,720 Speaker 1: be both Apart from Finland, it has the highest energy 453 00:25:42,760 --> 00:25:46,280 Speaker 1: availability according to your rankings, due in large part, I'm 454 00:25:46,320 --> 00:25:48,960 Speaker 1: sure because of all the surplus nucleon and France. For 455 00:25:48,960 --> 00:25:51,119 Speaker 1: those of you not familiar with the way that you 456 00:25:51,119 --> 00:25:54,520 Speaker 1: know power works in Europe, France is just constantly exporting 457 00:25:54,600 --> 00:25:57,440 Speaker 1: power to the rest of Europe. So if you use 458 00:25:57,560 --> 00:26:00,000 Speaker 1: up some more power in France, they just have to 459 00:26:00,040 --> 00:26:01,800 Speaker 1: we export a little bit less to everyone else. But 460 00:26:01,840 --> 00:26:04,320 Speaker 1: France is always going to be fine, And there's the 461 00:26:04,359 --> 00:26:06,760 Speaker 1: fiber network is there? So do you think that France 462 00:26:06,840 --> 00:26:09,600 Speaker 1: could be kind of both? Yeah? 463 00:26:09,720 --> 00:26:12,720 Speaker 2: And we are already seeing this. So we spoke a 464 00:26:12,760 --> 00:26:16,600 Speaker 2: little bit earlier on about foundational AI models. These companies 465 00:26:16,640 --> 00:26:20,240 Speaker 2: are building huge AI tools and I said most of 466 00:26:20,280 --> 00:26:23,480 Speaker 2: them are in the US. There's also a group in China. 467 00:26:23,560 --> 00:26:26,600 Speaker 2: There is one European champion, really and that is Meestrel, 468 00:26:26,760 --> 00:26:31,879 Speaker 2: a French company. Mistral is developing an AI training facility 469 00:26:32,040 --> 00:26:34,960 Speaker 2: in the Cambrai region. Apologies for any French speakers for 470 00:26:34,960 --> 00:26:40,000 Speaker 2: my pronunciation. There just northeast of Paris. And so yeah, 471 00:26:40,040 --> 00:26:44,000 Speaker 2: we're seeing deployments. We've also seen the end of the 472 00:26:44,400 --> 00:26:49,520 Speaker 2: ARNH mechanism for EDF, the French state energy provider selling 473 00:26:49,600 --> 00:26:53,560 Speaker 2: nuclear that's ended, which means it's now looking for private 474 00:26:53,600 --> 00:26:57,640 Speaker 2: sector of commercial offtakers. Data centers have been identified as 475 00:26:57,880 --> 00:27:00,800 Speaker 2: a good target customer and so EDF is in the 476 00:27:00,800 --> 00:27:04,200 Speaker 2: process of selling four possibly going to be extended to 477 00:27:04,280 --> 00:27:07,959 Speaker 2: six plots of powered land which will provide you through 478 00:27:08,040 --> 00:27:11,479 Speaker 2: energy for a data center on site at AI scale. 479 00:27:11,920 --> 00:27:15,239 Speaker 1: So yeah, tailwinds in France to two things. I'm going 480 00:27:15,320 --> 00:27:18,480 Speaker 1: to ask you to defind their powered land and put 481 00:27:18,480 --> 00:27:20,639 Speaker 1: a number around what we mean by AI scale. 482 00:27:21,200 --> 00:27:25,960 Speaker 2: Yeah, absolutely, so powered land is you ask twenty people. 483 00:27:25,680 --> 00:27:26,200 Speaker 1: In this space. 484 00:27:26,200 --> 00:27:28,720 Speaker 2: What they mean, you'll probably get training different answers. What 485 00:27:28,760 --> 00:27:32,560 Speaker 2: we mean, here are plots of land that are pre permitted, 486 00:27:32,840 --> 00:27:35,720 Speaker 2: so they're ready to build, and they have got any 487 00:27:35,840 --> 00:27:38,240 Speaker 2: source of power, in this case a grid connection to 488 00:27:38,280 --> 00:27:41,120 Speaker 2: a nuclear asset, ready to go. This is a plot 489 00:27:41,119 --> 00:27:43,960 Speaker 2: of land. The DF is saying, however, many football fields 490 00:27:44,000 --> 00:27:47,040 Speaker 2: of space. Build a data center, plug it in, and 491 00:27:47,480 --> 00:27:49,919 Speaker 2: all of that's taken care of for you, And then 492 00:27:49,920 --> 00:27:54,320 Speaker 2: I think the other question was AI scale. Similarly training answers. 493 00:27:54,760 --> 00:27:58,399 Speaker 2: Really we're talking about the hundreds of megawaps, if not 494 00:27:58,480 --> 00:28:03,280 Speaker 2: low gigawatts of capacity. Four plots edf have offered up 495 00:28:03,520 --> 00:28:07,080 Speaker 2: come to somewhere around two gigawatts together, So that kind 496 00:28:07,080 --> 00:28:09,400 Speaker 2: of gives you a sense of just how big we mean. 497 00:28:10,520 --> 00:28:14,919 Speaker 1: It's huge. It's small compared to Virginia, but it kind 498 00:28:14,960 --> 00:28:17,359 Speaker 1: of similar to what we were talking about with Brazil. Once 499 00:28:17,400 --> 00:28:20,440 Speaker 1: you get some data centers of that scale built, it's 500 00:28:20,480 --> 00:28:24,040 Speaker 1: almost like you've you've got over the hill in terms 501 00:28:24,040 --> 00:28:27,040 Speaker 1: of you know, can you build data to that scale. 502 00:28:27,080 --> 00:28:31,840 Speaker 1: So we assume that if those large scale projects happen suddenly. 503 00:28:32,240 --> 00:28:35,600 Speaker 1: Now France has an incumbency advantage in this race. Is 504 00:28:35,600 --> 00:28:37,359 Speaker 1: that is that a fair way to think about it. 505 00:28:37,760 --> 00:28:41,840 Speaker 2: Yes, caveatd on the network effect and this noble effect 506 00:28:42,120 --> 00:28:45,600 Speaker 2: on large scale AI projects is probably going to be 507 00:28:45,640 --> 00:28:49,400 Speaker 2: lower than for traditional data centers because, as we say 508 00:28:49,440 --> 00:28:52,160 Speaker 2: at the start, AI facilities care that little bit less 509 00:28:52,480 --> 00:28:56,880 Speaker 2: about low latency connection. But France does have structured advantages 510 00:28:57,120 --> 00:29:01,240 Speaker 2: deploying this will make it more attractive. And it's also 511 00:29:01,680 --> 00:29:05,840 Speaker 2: the government is hugely supportive of AI infrastructure and has 512 00:29:05,880 --> 00:29:08,640 Speaker 2: helped secure one hundred and ten billion euros, which is 513 00:29:08,640 --> 00:29:12,040 Speaker 2: one hundred and twelve billion US dollars of private sector 514 00:29:12,040 --> 00:29:15,920 Speaker 2: investment to help build and develop this data center capacity. 515 00:29:16,120 --> 00:29:19,720 Speaker 2: That's vastly more than in the UK, which has similar ambition, 516 00:29:19,960 --> 00:29:22,960 Speaker 2: that has only secured around thirty one and a half 517 00:29:23,080 --> 00:29:25,920 Speaker 2: billion US dollars of private sector investment to support its 518 00:29:26,040 --> 00:29:26,959 Speaker 2: data center sector. 519 00:29:27,840 --> 00:29:32,080 Speaker 1: Firstly, I'm really impressed with your currency conversions on a fly. 520 00:29:33,760 --> 00:29:38,200 Speaker 1: If you know, a few really huge projects could be significant, 521 00:29:38,240 --> 00:29:41,600 Speaker 1: Although you know, I do take on board your caveat. 522 00:29:41,760 --> 00:29:44,640 Speaker 1: Are there any other really large scale projects across Europe 523 00:29:45,240 --> 00:29:47,720 Speaker 1: or not even necessary projects like the equivalents of this 524 00:29:47,800 --> 00:29:49,800 Speaker 1: powered land you have in France? Are there any other 525 00:29:49,840 --> 00:29:53,800 Speaker 1: examples of countries where there's a sort of a nicely 526 00:29:53,880 --> 00:29:56,480 Speaker 1: set up opportunity to get ahead on this game. 527 00:29:56,840 --> 00:29:59,880 Speaker 2: Yeah. Absolutely, so to bene f subscribe as people access 528 00:29:59,880 --> 00:30:02,640 Speaker 2: to this note start of section five. We've got a 529 00:30:02,680 --> 00:30:06,440 Speaker 2: map of the European pipeline and you can see where 530 00:30:06,600 --> 00:30:10,080 Speaker 2: the large announced projects are. But really what we're seeing 531 00:30:10,360 --> 00:30:13,160 Speaker 2: are a number in the northeast of England and up 532 00:30:13,240 --> 00:30:18,040 Speaker 2: into Scotland around the UK's AI Growth Zone policy. We're 533 00:30:18,080 --> 00:30:21,560 Speaker 2: seeing projects in the north of the Nordics, and we're 534 00:30:21,560 --> 00:30:25,320 Speaker 2: seeing projects into Spain and also the Stark Campus in Portugal. 535 00:30:25,600 --> 00:30:29,600 Speaker 2: Things I'd really flag are Stargate Europe, so this is 536 00:30:29,880 --> 00:30:34,440 Speaker 2: a project tied to USAI firm OpenAI company makes Touch 537 00:30:34,480 --> 00:30:39,680 Speaker 2: GPT in Norway, and then the hyperscalers, so Google, Microsoft, 538 00:30:40,000 --> 00:30:44,360 Speaker 2: Meta and Amazon also have large scale deployments, notably in 539 00:30:44,400 --> 00:30:46,600 Speaker 2: the Nordics and the Nordics. 540 00:30:46,200 --> 00:30:48,920 Speaker 1: Makes sense to me because there's a huge amount of 541 00:30:49,120 --> 00:30:52,680 Speaker 1: either hydro or nuclear or hydro and nuclear, and I 542 00:30:52,720 --> 00:30:55,600 Speaker 1: know those data centers under development in the parts of 543 00:30:55,840 --> 00:30:59,280 Speaker 1: the Nordics where there's low demand but high generation. I'm 544 00:30:59,360 --> 00:31:02,320 Speaker 1: kind of curious Spain and Portugal as well. I mean, 545 00:31:02,320 --> 00:31:04,120 Speaker 1: I realized there's a lot of hydro and Portugal, but 546 00:31:04,360 --> 00:31:08,040 Speaker 1: a lot of the kind of the energy abundance emerging 547 00:31:08,080 --> 00:31:12,640 Speaker 1: in Spain is relating to solar and excess solar. Does 548 00:31:12,640 --> 00:31:15,080 Speaker 1: that carveat the advantage that Spain has in terms of 549 00:31:15,200 --> 00:31:17,920 Speaker 1: energy availability? Or am I oversimplifying it all sort of 550 00:31:17,960 --> 00:31:19,160 Speaker 1: missing a factor there? 551 00:31:19,520 --> 00:31:21,320 Speaker 2: So I guess there's a couple of things we could 552 00:31:21,320 --> 00:31:24,080 Speaker 2: think about in answer to that. One that I'm just 553 00:31:24,280 --> 00:31:26,120 Speaker 2: not going to be able to get into here, but 554 00:31:26,480 --> 00:31:30,480 Speaker 2: is around data center flexibility. This is something lots of 555 00:31:31,120 --> 00:31:33,640 Speaker 2: groups are increasingly thinking about and exploring. So how can 556 00:31:33,760 --> 00:31:37,000 Speaker 2: data centers make themselves more flexible and move their load 557 00:31:37,200 --> 00:31:41,440 Speaker 2: to better match generation curves? But yeah, you're exactly right, 558 00:31:41,600 --> 00:31:45,680 Speaker 2: solar in Spain, and this is what's drawing a lot 559 00:31:45,720 --> 00:31:50,600 Speaker 2: of these projects into the market. So auctions the bid 560 00:31:50,680 --> 00:31:54,120 Speaker 2: process for a grid connection if it's contested. Data centers 561 00:31:54,160 --> 00:31:59,120 Speaker 2: are systemically disadvantaged against decobilization and clean industry as a 562 00:31:59,160 --> 00:32:04,000 Speaker 2: result of Banish government policy, But via Royal Decree one 563 00:32:04,120 --> 00:32:07,360 Speaker 2: one eight three in twenty twenty, you can use up 564 00:32:07,400 --> 00:32:10,560 Speaker 2: to fifty percent of the CAPASC you generate on site. 565 00:32:10,920 --> 00:32:14,280 Speaker 2: That has created a market opportunity for data centers to 566 00:32:14,360 --> 00:32:19,440 Speaker 2: co locate with solar, and we're seeing solar developers in 567 00:32:19,480 --> 00:32:23,040 Speaker 2: Spain providing powered land a bit like EDF in France 568 00:32:23,280 --> 00:32:25,360 Speaker 2: to a data center saying, look, you can get a 569 00:32:25,360 --> 00:32:28,959 Speaker 2: grid connection here. You can use our solar really cheaply 570 00:32:29,160 --> 00:32:31,360 Speaker 2: when the sun is shining, and then you've got that 571 00:32:31,400 --> 00:32:34,360 Speaker 2: grid connection to bring power in when it's not. So 572 00:32:34,720 --> 00:32:38,200 Speaker 2: that's a win for solar developers because it helps them 573 00:32:38,240 --> 00:32:41,600 Speaker 2: to close the missing money gap that being f writed 574 00:32:41,600 --> 00:32:44,320 Speaker 2: about in the last Spanish power outlook. And it's a 575 00:32:44,320 --> 00:32:46,800 Speaker 2: win for data centers because it means they can get 576 00:32:46,840 --> 00:32:47,600 Speaker 2: a grid connection. 577 00:32:48,160 --> 00:32:52,320 Speaker 1: So it's a win win for both sides of the party. Lloyd, 578 00:32:52,360 --> 00:32:55,840 Speaker 1: this has been a fascinating conversation, and I think you've 579 00:32:56,080 --> 00:32:59,240 Speaker 1: said on multiple times in response to my questions that 580 00:32:59,360 --> 00:33:03,960 Speaker 1: it's a bit more complex than that, which is fair, 581 00:33:04,720 --> 00:33:07,760 Speaker 1: and I suppose what I'm learning here is I mean 582 00:33:07,800 --> 00:33:10,640 Speaker 1: you've kind of distilled it down into five pillars in 583 00:33:10,680 --> 00:33:12,840 Speaker 1: terms of, you know, where we expect data centers to 584 00:33:12,880 --> 00:33:15,680 Speaker 1: get built in the markets that you've covered, or you 585 00:33:15,680 --> 00:33:18,040 Speaker 1: know what factors will determine that. But even doing that, 586 00:33:18,160 --> 00:33:22,120 Speaker 1: there's still more factors. You know, we talked about sovereign AI. 587 00:33:22,840 --> 00:33:25,920 Speaker 1: We talked about the emissions perspective and sort of the 588 00:33:26,280 --> 00:33:29,240 Speaker 1: role that government can play, and so it really paints 589 00:33:29,240 --> 00:33:33,080 Speaker 1: a picture of an interesting space in the next five 590 00:33:33,160 --> 00:33:35,480 Speaker 1: years or so that there are a lot of markets 591 00:33:35,480 --> 00:33:39,440 Speaker 1: that maybe aren't on the radar as places where AI 592 00:33:39,640 --> 00:33:42,920 Speaker 1: is getting built, but there are many that potentially could be. 593 00:33:43,200 --> 00:33:45,479 Speaker 1: I'm sure that not all of them will be, and 594 00:33:45,520 --> 00:33:47,360 Speaker 1: so I think it's going to be really interesting seeing 595 00:33:47,480 --> 00:33:50,560 Speaker 1: what happens in the next five years, what emerges as 596 00:33:50,640 --> 00:33:53,880 Speaker 1: the new Virginia, you know for Europe, or the new 597 00:33:53,920 --> 00:33:56,120 Speaker 1: Texas but in the Middle East, and so we'll be 598 00:33:56,240 --> 00:33:58,040 Speaker 1: keeping our eye on that. And so this has really 599 00:33:58,080 --> 00:34:01,120 Speaker 1: open my eyes as to why I'm based in the US. 600 00:34:01,240 --> 00:34:03,520 Speaker 1: I look at power markets so so much. My fixation 601 00:34:03,800 --> 00:34:06,680 Speaker 1: is on the impact of data centers in those two 602 00:34:06,800 --> 00:34:09,840 Speaker 1: markets Virginia and Texas. But you've actually really opened my 603 00:34:09,880 --> 00:34:12,800 Speaker 1: mind up to why this could be so much bigger 604 00:34:12,800 --> 00:34:15,120 Speaker 1: than that and impact so many more markets. So thank 605 00:34:15,160 --> 00:34:17,520 Speaker 1: you so much. It's been a real pleasure having this 606 00:34:17,560 --> 00:34:18,920 Speaker 1: conversation with you. Now. 607 00:34:18,920 --> 00:34:20,680 Speaker 2: It's been great to be here. Thank you so much 608 00:34:20,719 --> 00:34:30,239 Speaker 2: having me on Tom. 609 00:34:30,440 --> 00:34:33,560 Speaker 1: Today's episode of Switched On was produced by Cam Gray, 610 00:34:33,760 --> 00:34:37,480 Speaker 1: with production assistants from Kamala Shelling. Bloomberg NEIF is a 611 00:34:37,520 --> 00:34:40,520 Speaker 1: service provided by Bloomberg Finance LP and its affiliates. 612 00:34:40,560 --> 00:34:43,280 Speaker 2: This recording does not constitute, nor should it be construed, 613 00:34:43,320 --> 00:34:47,200 Speaker 2: as investment advice, investment recommendations, or a recommendation as to 614 00:34:47,239 --> 00:34:50,080 Speaker 2: an investment or other strategy. Bloomberg ANIF should not be 615 00:34:50,160 --> 00:34:53,960 Speaker 2: considered as information sufficient upon which to base an investment decision. 616 00:34:54,040 --> 00:34:57,000 Speaker 1: Neither Bloomberg Finance LP Nor any of its affiliates makes 617 00:34:57,040 --> 00:35:00,799 Speaker 1: any representation or warranty as to the accuracy or completeness 618 00:35:00,800 --> 00:35:03,799 Speaker 1: of the information contained in this recording, and any liability 619 00:35:03,840 --> 00:35:06,480 Speaker 1: as a result of this recording is expressly disclaimed.