Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Power, Podiums & Processing: Nuclear Learning Curves, Brisbane's Olympic Push & France's AI Gambit
53 Plays18 days ago
Join Jason and Carlos in this episode of The Offsite Podcast as they explore three major developments reshaping the global infrastructure landscape:
⚛️ Nuclear Power's Learning Curve: Dive into the UK's nuclear ambitions through Sizewell C and Hinkley Point C projects. Discover how Hinkley's budget grew from £18 billion to £46 billion, the challenges of replicating designs, and what this means as 31 countries plan to triple their nuclear capacity by 2050.
🏅 Brisbane Olympics 2032 Infrastructure Challenge: An in-depth analysis of the $50 billion infrastructure pipeline needed for the games, including the demand for 10,000+ workers, 14,000 hotel rooms, and multiple transport projects. Compare this with London 2012's successful legacy and explore the risks of concentrated construction activity in Southeast Queensland.
🤖 Project France AI: Explore France's ambitious €109 billion AI infrastructure investment, including Europe's largest data center development. Understand how this positions France against the US's Project Stargate and examine different national approaches to AI infrastructure development.
Key Timestamps:
00:00 - Introduction
04:33 - Nuclear Power Discussion
16:44 - Brisbane Olympics Infrastructure Analysis
29:25 - French AI Investment Plans
Check out the Off Site newsletter here
Follow Carlos on Linkedin | Follow Jason on Linkedin | Check out Aphex
Recommended
Transcript
Nuclear Power Overview
00:00:00
Speaker
ah The other thing to mention, i didn't quite realize the number of operating like live reactors around the world. um say I didn't realize the US had as much nuclear as it does. There were 93 live nuclear power stations.
00:00:13
Speaker
yeahp um and Then we got sort of China and France in the 50s, and then Russia, to Japan, South Korea. so Some of these countries have serious investment into nuclear. 31 countries around the world plan to triple nuclear capacity by 2050. So if you're a contractor or if you're someone on a project now, ah getting experience and riding that learning curve yourself or as a company ah is a huge opportunity to then take that to the next one, take that to the next one. Yeah.
00:00:42
Speaker
Going forward.
Host Introductions
00:00:50
Speaker
Welcome back to the Offsite Podcast. I'm Jason Lansini and joined again by Carlos Cavallo. ah One of us sat in a beautiful snowing Germany and the other in a unfortunately not snowing London.
00:01:06
Speaker
How are Carlos? I'm very well, thanks. How are you? I am good. I'm just enthralled. It's the first time since we've been here that it's snowed and it's um like Christmas came late slash early.
Humor in Current Events
00:01:20
Speaker
It's beautiful. I saw ah couple of days ago in the news. I thought it was worth noting because it's like this week's episode of Black Mirror. um The Gulf of Mexico has officially been renamed in Google Maps as the Gulf of USA as per Donald Trump's request.
00:01:40
Speaker
So, um yeah, we always knew that um he had Zuckerberg, Musk and Bezos in his hand, but I wasn't so sure that he also controlled Google so strongly.
00:01:51
Speaker
So, yeah, pretty wild. Awesome. In other news, the content of the podcast has shifted hardly hard from construction to we're now just doing straight politics.
00:02:01
Speaker
We didn't think there was enough US politics podcasts out there. Yeah, it's too baffling to not mention. But yeah, we'll be back on track next week with some solid story about Aviva building a dam.
Podcast Focus Shift
00:02:16
Speaker
uh uh i didn't actually fact check that uh but i did get a message saying that it was maybe dubious in uh factual origin oh really yeah uh which i have refused to believe i think definitely still true there's at least four websites that have it albeit all four look a bit suspect but yeah between them they have 120 visitors a year or something yeah yeah I'd never to get let the truth get in the way of a good story anyway. Yeah, exactly. The truth starts now in the conversation. We're going to talk about a couple, we're talking about a lot of big projects that have been...
00:02:51
Speaker
I guess on radar, so some news, some in the news recently.
UK Nuclear Projects
00:02:56
Speaker
so we're going to kick off by discussing, um really triggered by an article in the Financial Times talking about big nuclear energy construction, specifically in the UK, and the trajectory of ah project delivery and this idea of, you know, the articles will all point to massive cost escalation of these projects.
00:03:15
Speaker
um and talk about whether that is whether there's ah pattern that's just going to continue happening or or what's going on behind the scenes. Then we're going to dive into, again, mega schemes. So there is an Olympics approaching for Brisbane in Australia.
00:03:33
Speaker
There is a lot of conversation across that country about ah infrastructure for the Olympics and a lot of conversation about the ability to deliver that infrastructure from construction workforce side from general contractor side.
00:03:50
Speaker
And we're going to talk a little bit about the complexity of delivering simultaneously games infrastructure, related infrastructure, housing, et cetera. And then final one is not to be, well, still to be outdone, but ah kind of coming in strong. ah the recent announcement from France in response and response, probably in response to the Stargate news that we talked about last week with a massive 109 billion euro investment into AI infrastructure and data centers across France, making France the kind of European AI hub.
00:04:30
Speaker
And again, what that means ah for the industry. So number one, nuclear power and specifically starting with c Sizewell in the UK and Include Point
Construction Challenges in Nuclear Power
00:04:41
Speaker
C. So a recent article in the Financial Times in the UK dives into the upcoming Sizewell C project. So a major large nuclear power station project planned um and soon to be under construction, if not Carls already under construction to some degree.
00:04:59
Speaker
Yeah, enabling works have started. um In the UK. And talks about the legacy of the current active project at Hinkley Point, so Hinkley Point C, about 280 miles away on the other side of the country.
00:05:14
Speaker
ah which has kind of unfortunately joined the trend of major nuclear power stations across the world where the end cost is double or in the realm of double and the schedule is late to very, very, very late compared to an initial estimates. And the article dives into how SizeWell is planning to, and has almost copied 85% of the design or the elements of Hinkley Point C with this idea of trying to ride the the learning curve of construction efficiency and design efficiency.
00:05:51
Speaker
So I guess it raises this ah question of, um is there this negative learning curve idea, an idea that they're they're actually getting worse and worse the more we learn about ah building and a major nuclear power stations?
00:06:07
Speaker
Is the idea of duplicating the, you know doing a carbon copy, copying bad ideas basically and um doomed to have the same outcome, or is it, okay, let's start writing the learning curve yeah what's the What are your thoughts on on the topic, being a ah UK resident and presumably not diving into a car, but presumably a taxpayer that's somewhat contributing to it?
00:06:31
Speaker
You weren't meant to ask me that on this podcast. but um I think just to give some context, ah so the the latest figures for Hinckley was originally going to be completed this year for 18 billion.
00:06:44
Speaker
It's now 2029 for 46 billion is the estimate in that article. So it's quite an increase in line with other similar buccalipal stations being built in other countries at the moment. Yeah, it's it's interesting in terms of the the the copy between Hinkley and SizeWell.
00:06:59
Speaker
I'd like to think some someone somewhere grabs a file, hits create a copy, and renames it SizeWell, and that's how easy it is to copy a design. I'm sure it's not that easy. But um no, and it it definitely makes a lot of sense. um So there was an interesting stat in that article, which was Hinkley's got two reactors. The second one's 30% quicker.
00:07:19
Speaker
So you'd like to think that that's learnings, that's efficiencies, that's all sorts of things that you're... working out on the first, replicating on the second. and If we can take that learning onto Sizewell, then hopefully it can and it it can improve with each one.
00:07:32
Speaker
Sizewell, they are forecasting 20% cheaper build than Hinkley um on that basis. it seems to be and You can imagine it's a safety critical project, not to suggest other projects aren't safety critical, but it's nuclear. There's tens of thousands of people um on a site.
00:07:46
Speaker
There's a lot of moving parts, heavy machinery. um and ah In terms of UK regulation, um it's an interesting point around as as risks emerge, rules get put in place, as we all know on construction projects. so Over time, things are getting tighter and tighter and more difficult to build and more stringent with more rules in place. so You can see why you've got this balance between we're getting the hang of this, we know what we're doing, we're getting more efficient,
00:08:13
Speaker
But then at the same time, you're being sort of restricted in your ability to do things because every time there's an issue or a risk emerges, the rule set's getting tighter. It's a really difficult balance between those two around doing things or repeating things quicker, but also... if You're in the shoes of the people with decision-making capability around the planning of this. Are are you looking globally and going, let's kind of find what's working somewhere else and bring that in, even if it's not the same as kind of what we did on Inkly?
00:08:42
Speaker
Or is it like we've kind of picked a design or technology, you know, partners and supply chain, and let's just duplicate that and do it as, you know, as quick and efficient again and and start to build that like know-how?
Small Nuclear Reactors in the UK
00:08:56
Speaker
Yeah, so like the major player in Europe is EDF and they're they're trying to do a copy and paste across ah major reactors. And I guess the secondary to that, we have the small reactors where like Rolls-Royce and organizations are like that are getting involved and you imagine they are shit hot at a repeatable sort of factory mindset in terms of building parts, delivering these schemes. They're much smaller, so less kind complexity, but um yeah, their idea is they're going to do this copy and paste exercise. And i think there's nine planned for the UK.
00:09:25
Speaker
So it needs to get to that point. And you'd imagine there's then fringe benefits like offsite manufacture, because we have standardization and everything's not full customer bespoke.
00:09:36
Speaker
So what would you do? What would you do if you were in the, in the, you if I was going to say if you're in the position of the politician, or the but you've you've you've actually done exactly what the politician would do, which is just answer.
00:09:54
Speaker
There's no hiding in this podcast. I would not put contractors on target cost arrangements. Okay.
00:10:03
Speaker
Well, i'll I'll put a stake in the sand. I'd definitely do the same thing again. i think i think the idea of a learning curve is exactly what construction does well.
Learning Curve in Construction
00:10:15
Speaker
ah The idea of production, you know, from the very first project that I worked on, the idea that you can ah do something for the first time, measure how it long it took and all of its component parts, turn that into kind of a repeatable process that every time you iterate, we go, what's the cycle time of this thing? Where do we lose time?
00:10:35
Speaker
Where can we you know tighten up handoffs? Where can we um you know optimize resources? and repeat repeat, repeat, repeat, I think is exactly playing to the strength of what construction as an industry does.
00:10:51
Speaker
you not think they're already doing that? Well, I think the problems are that, you know, even if you try to copy and paste things, the problem is it's like the first big one that's been built in the country. The workforce has never done these things before.
00:11:04
Speaker
The supply chain has never done these things before. Some of it requires new tooling it like a supplier. Some requires like building whole facilities in terms of the manufacturer.
00:11:15
Speaker
And so there's, I think there's like, there's definitely a cycle times on the actual project that they will be doing. But then by the time you hit the construction site, there's so many things that are kind of locked in place. Your supply chain set, especially Hinkley, like where it's located, the option set is yeah some of the stuff you are kind of locked into.
00:11:33
Speaker
So there's like tight cycle times in the project, but then there's like broader cycle times and you know learning curves to happen at higher and higher levels on the way up. And then there's this other trend you kind of touched on it, which is like the regulation around this industry that forces each one to be somewhat different than the other.
00:11:54
Speaker
And the like enemy of a learning curve is like bespokeness between the the the things that you're trying to repeat. you know, if you're trying to do a bridge and every bent or every section of the bridge is different to the one before, you can never get the cycle time like dialed in exactly.
00:12:11
Speaker
But there's a... Yeah, just to like close that thought, sorry to interrupt, but like the, so you end up with these, you know, the increasing regulations, the obligations, the around environmental, et cetera, et cetera, that force each of these things to be so different to the one before.
00:12:28
Speaker
This trend that's happening around countries around the world, we'll see it in the news in France and the things we're talking about next, we just saw it in the announcement in the US and the UK is just similarly pushing again around reducing regulation to allow more of this nuclear power to be built.
00:12:44
Speaker
will allow more of the elements to be the same as the previous and more quickly ride the learning
Global Nuclear Investment Opportunities
00:12:52
Speaker
curve. So I think I would be doubling down into, you know, unless there's something fundamentally flawed, but I don't, from my knowledge it isn't, I think I would be doubling down into cycle time and as much copy paste as possible.
00:13:06
Speaker
Yeah, the difficult thing is, um like take the yeah UK for example, there isn't a huge pipeline of major nuclear reactors. Like these are the two we know of and that's that's it. If we exclude the smaller ones.
00:13:17
Speaker
As a contractor, how much time are you really going to invest on that sort of mindset to future-proof the next? When, if you're a Lang O'Rourke who's on Hinkley um and Sizewell, after Sizewell, are you really going to be thinking about like how can we really tighten this up and make it repeatable? So they're not really incentivized to do that.
00:13:36
Speaker
It's small. It's small ball though. Like, like realistically, the UK is a very small part of Europe, like more broadly. And there's, I think there's something like 60 something active nuclear power stations being built around the world right now.
00:13:48
Speaker
And so there might be a small pipeline in the UK, but if, ah you know, the developers, you know, in this case, EDF, they're obviously French. So their, their consideration is much broader than, than the u k A lot of the contractors bidding on work and winning work on these power stations are much broader than the UK.
00:14:06
Speaker
And it gives them a skillset that they can export to build these projects. You do a good job on one, you get to go on to the the next. yeah you selling It's your selling point. You know, we did this, we learned, you know, these things went well, these things didn't go well. We're the one with the learning. So but even if there were some things that didn't go well, do you want to take us to take those learnings forward? Or do you want to go to someone totally new who could totally shit the bed on the thing again?
00:14:29
Speaker
Yeah, no, 100%. I guess I was thinking more like Langerwog doesn't operate in Europe. So that specific example is that future plans, is that something they plan to do in Australia where they do have a presence and things like that?
00:14:41
Speaker
ah Do you think that they wouldn't? Like if they had the opportunity to joint venture with, a I don't know, a Vinci on a honor project or an iFrage on a project.
00:14:52
Speaker
It's a train ride away. You can actually drive across the train. It's not like you're building on Mars. So I think like, yeah, definitely. I think if ah if anyone is thinking that way, I'd say get out of the like small island thinking.
00:15:08
Speaker
um Yeah, yeah. For sure. Yeah, yeah, yeah. There's definitely an aspect to that. the The other thing to mention, i didn't quite realize the number of operating like live reactors around the world.
00:15:19
Speaker
um So I didn't realize the US had as much nuclear as it does. There were 93 live nuclear power stations. yeahp um And then we got sort of China and France at the 50s, and then Russia, to Japan, South Korea. So some of these countries have serious investment into nuclear.
00:15:36
Speaker
31 countries around the world plan to triple nuclear capacity by 2050. So if you're a contractor or if you're someone on a project now, ah getting experience and riding that learning curve yourself or as a company ah is a huge opportunity to then take that to the next one, take that to the next one. Yeah.
00:15:55
Speaker
Going forward. Yeah, absolutely. Countries like Finland, they've got five major nuclear power stations running at only 5 million population. You can see how clean that energy is for almost everyone. yeah Some have doubled down on it and it's obviously working well.
00:16:08
Speaker
The other thing is, if you can, as and as a country, if you can build nuclear, build it cheap, and therefore you get cheap energy, you must be more attractive for investment in AI and data centers and all the other things because it's a really attractive proposition if you can offer cheap energy. It's a really worthwhile investment. um yeah sure ah Speaking as someone that is in Germany, ah you know, this will segue onto actually the next topic, which is around like data centers, the data the energy source needed for a data center.
00:16:36
Speaker
Actually, it's probably worth even just prepping and diving straight into, oh, that's topic three. Sorry, let's let's close that thought. let's We'll go to topic two and then we'll we'll dive into that in the third. um There we go.
00:16:50
Speaker
Everyone with me? Good. Okay.
Brisbane's Olympic Infrastructure Challenges
00:16:55
Speaker
There is, um there's ah there's an Olympics coming up, right? um And it is in 2032 in Brisbane, in Australia, and there is a growing unrest and awareness of a looming problem which is that there's something like seven odd years to deliver ah infrastructure for the project.
00:17:18
Speaker
There is the need to deliver direct venue infrastructure, um like stadiums, support venues, athletes, villages, et cetera. And then there is the the usual thing that cities will try and do is to capture that and build you know a legacy around infrastructure. So rail infrastructure, hotels, you know road infrastructure, parks and public realm.
00:17:43
Speaker
And then that usually drives a population boom, both from the construction activity and from the you know the awareness of the city, which then suddenly drives the need for more commercial or residential ah ah housing, et cetera.
00:18:00
Speaker
And that can create this perfect storm of everything getting built at the same time um in the same place, which draws massively on the workforce, the resources, ah which has the next consequences like skyrocketing prices of everything.
00:18:20
Speaker
And one of the legacies can be, well, a massive bill to the whole region and to companies and stuff in the region. unrealistic and unsustainable ah wages and salaries because of the you know the competition.
00:18:35
Speaker
And then also this like dearth of maybe there's not enough infrastructure work or construction work afterwards. And then suddenly the whole thing cycles ah back through. So it's almost like the You know, this idea of construction is a cyclical industry. It always is some form of this kind of ah boom and bust. But something like ah an Olympics can, you know, has the opportunity to make that into like a mega cycle.
00:19:00
Speaker
So before diving into some of the stats, no, two audibles, let's dive into some of the stats. And then I'll ask you, Carlos, about, you know, reflecting of, say your experience with London Olympics, you know, is this a cycle that's doomed to repeat? So Brisbane is expected to have about 50 billion worth of total infrastructure in pipeline.
00:19:21
Speaker
They expect to need over 10,000 workers for the Olympics projects alone, excluding ancillary related infrastructure. They're expecting on top of that to need to build something like 14,000 hotel rooms, 20 plus, you know, residential towers for athletes village.
00:19:38
Speaker
And there is this predicted giant like crash of a demand between about 2027 and 2029, where the cost of everything is going to go up, the demand for everything is going to go up.
00:19:51
Speaker
Yeah, there's the contractors and projects will be scrambling for people, ah for materials um and for resources. And so i guess, yeah, first question, you know, does this is this something that you've seen before?
00:20:06
Speaker
Yeah, so it's going back a few years now. didn't quite realize it was 12, 13 years ago, but in general terms, it was seen as a success, the investment in the Olympics in London and the fringe benefits have we got from that investment.
00:20:20
Speaker
So the the major things or the major aspects was infrastructure, So they used a as an opportunity to effectively fast track spend. It gave the mayor of London quite a strong card to play to to get things done quicker.
00:20:34
Speaker
So they did a capacity project on the Jubilee line, particularly around the Stratford area. Yeah, it was an extension as well. Did they extend the the line up through to Stratford? Basically, it finished before Stratford before. So it went straight into Stratford and increased capacity in a few other stations leading up.
00:20:52
Speaker
yep There's the London Overground. So, ah you know, Shoreditch High Street station and all those overground stations, they were completely new just for that. And they are very well used systems now.
00:21:03
Speaker
um There was the DLR extension, um which sort of sort of extended some of the lines there. So there was a lot of infrastructure which was very much needed anyway, but it was an opportunity to accelerate that and make sure it's done for the Olympics. So that was that was a big win from an infrastructure point of view.
00:21:20
Speaker
There was the Olympic Village, obviously had to be built, um but after the Olympics, they turned it into thousands of flats, proper residential flats.
00:21:31
Speaker
So it wasn't a build and demolish or yeah end up with something that wasn't particularly useful. um It became, I think it was 3,000 homes built in relatively central London. The sporting venues, so the main Olympic stadium is now West Ham's football club stadium. So they they resold it onto the Premier League club or they hire, they lease it.
00:21:50
Speaker
Like the Aquatic Centre, the Velodrome, they are all now proper sports facilities. They weren't temporary structures. It's not like a Qatar World Cup where they just demolish them after because who's going to use them? and Over that period, I think it's they estimated 46,000 workers were used to to build everything Olympics related.
00:22:09
Speaker
and it was kind of a i don't know if someone back in the day tried to plan it this way. You probably think not. But the fact that Crossrail started concurrently and really started to ramp at the end of when the Olympics happened, loads of those workers would have shifted straight across. and I worked with a ton of people that worked in the Olympics. so It almost helped us avoid that lull where we then had a bunch of unemployed workers that have relocated to the city or or or lived there in the first place and um those schemes ended. Yeah, I think London is held up often by Olympic committees and countries bidding as like the gold standard of what it was able to achieve.
London Olympics Infrastructure Legacy
00:22:44
Speaker
I personally ah saw exactly that. Like when I lived in London, I lived out near Stratford, right near the Olympic Park and that area used to be a total like marshland.
00:22:55
Speaker
ah You would never... Very industrial. Yeah, very industrial, a lot of open, unusable space. um And all of that transport I would use on a ah weekly a weekly basis and go walking around those facilities. it's like It was a totally rejuvenated ah part of the city that gave a total bunch of access to people for housing, et cetera. So that's kind of like the- it's still growing Yeah, yeah.
00:23:22
Speaker
It's flats flats flats. a hub. Yeah, yeah. And and um the idea that I don't know if you've got stats on like what what the investment or the cost was ah there, but you know coinciding with like the GFC when a lot of that was happening, I'm sure that is like crazy good value now.
00:23:42
Speaker
I've got so six and a half billion on the transport related infrastructure. Yeah. On top of that, you can't dig a hole for that now. Yeah, yeah ah you know you're not doing what they did for sure.
00:23:53
Speaker
i Not even close. Having that hard end day, I wonder how that actually helps. But um yeah there were other aspects. So they did a lot of investment into cycle lanes.
00:24:03
Speaker
The main motorways coming into the city, they had these priority lanes so they can get athletes in during rushdown and things like that. So ah that 6.5 billion on the surface looks like a great investment.
00:24:14
Speaker
Yeah, 100%. And yeah, the return on that would be unbelievable. It's always good. So like probably the best time to do something was 10 years ago and the second best time is is now. um and One day we last nice well was super cheap. We should have done more of them back then. but The idea of sequencing is interesting. And like, I guess the die is cast probably for something like Brisbane.
00:24:39
Speaker
where it almost feels like the opposite thing has happened, which is not only do you have, there's ah there's some amount of infrastructure that's pre-front
Sequencing Infrastructure in Brisbane
00:24:49
Speaker
loaded. So the crossover a rail is a new kind of metro line going underneath Brisbane.
00:24:54
Speaker
That started ah years ago and we'll you know the date recently got announced as moving, but like it'll be done well in advance of that kind of crunch period, which- They weren't originally tied. This wasn't a an Olympic- i Yeah, I don't think so. I think now it's like conveniently from a marketing perspective tied together. But the fact that that would be done before the wave, if that was at the same time, it would be hectic.
00:25:19
Speaker
But on top of that, there is a whole bunch of other things. ah government infrastructure spending around rail that is planned. There's a Logan Gold Coast Faster Rail. There's a road project going out to the Gold Coast because i think a bunch of the stuff is what it will happen on the Gold Coast and the Sunshine Coast.
00:25:37
Speaker
um So these are two coasts kind of next to Brisbane, an hour or so away from there. There's a rail project planned to go to from Brisbane to the Sunshine Coast, so a Sunshine Coast rail ah project. What sort of distance is Brisbane to Sunshine Coast? What's an hour, like 100 kilometers, something like that in that ballpark depends what part of the Sunshine Coast and what part of Brisbane.
00:25:59
Speaker
And then on top of that, Queensland is currently, which is the state that Brisbane is in, is undergoing its biggest ever investment in ah health infrastructure. And they're spending about $20 billion dollars on hospitals, which currently is like forecast to be 30 or something. Yeah.
00:26:14
Speaker
in overspend. And yeah, as I noted, like 14,000 new homes being built. So it definitely feels like that kind of like, you know, the benefit that London had of like being right in the middle of the GFC when it was a great, like almost like stimulus during the lull and then being able to flow onto that to something like a cross rail and other infrastructure feels like it was like perfectly timed.
00:26:41
Speaker
The other benefit that London has we're so slow to build things that we're always over capacity on everything, whether it's homes, transport, whatever. so When we build something, it is used immediately because we're struggling to keep up.
00:26:57
Speaker
Do you get you get this similar sense in Brisbane or do you think it's they're betting on growth in the future and building on the assumption that it will grow or is it already Yeah, brisbane Brisbane is like definitely the fastest growing city in Australia and um kind of spurred through COVID, there was a big shift and that's like just continued and continued. It's actually, I think the most expensive or second most expensive now to live.
00:27:24
Speaker
Definite shortage of homes. And it's got a big problem of urban sprawl just getting wider and wider and wider. um which makes transport ah big old problem. So transport infrastructure would be a really meaningful legacy to leave, but it's the talk at the moment is just like they keep chopping and changing what they're building. They canceled something, decide they can't decide on where a stadium's going.
00:27:48
Speaker
There's no, I think, like I said, the dyes probably cast for Queensland that in terms of like being able to sequence these things, but certainly it doesn't help to keep waiting. um because it just pushes the, it condenses the peak because like you said, it's a hard end date.
00:28:02
Speaker
Unless you want it to open after the Olympics, the end date is hard. So you just do this squishing thing and the peak keeps getting higher. Yeah, yeah, absolutely. So yeah, no meaningful and useful source suggestions, but it's going to get it's going to get crunchy in there for projects being delivered.
00:28:20
Speaker
Presumably, most of the major works haven't started yet. So you've got to probably a couple of years until it really ramps up. Yeah, so 2027 to 2029 is when they're projecting the peak.
00:28:32
Speaker
They're still arguing about like stadium locations and stuff like that. There's a big tender out that a bunch of people are tendering for with regards to ah like a managing contractor role for a bunch of the projects, but that keeps getting pushed back.
00:28:46
Speaker
There's like a $14 billion dollars dam project right next door that they're planning to do that keeps getting pushed back. The government's also like, how the hell are we going to pay for a bunch of these things? So this as they scramble to make a bunch of these decisions, it is just chewing time.
00:29:01
Speaker
yeah And so, yeah you get to a spot where it's like, we got to start making some decisions and commit to some things because otherwise this will get, it gets incrementally more expensive.
00:29:12
Speaker
Yeah. And already strange resource. If the programs are shorter, you need more and more. So more. Exactly. Yeah. You just have to crash it. Yeah. Yeah. Be interesting to see. We will definitely revisit this topic ah ah in the future.
France's AI Infrastructure Investment
00:29:25
Speaker
So jumping to, I guess, a final topic, we talked last week about the $500 billion dollars Stargate announcement, real or not, that came out of the US, which was this combined investment in AI data center and infrastructure with Donald Trump, OpenAI, i and a bunch of other parties and partners in the project.
00:29:46
Speaker
Somewhat in response, French President Emmanuel Macron announced a ah just in the last handful of days or in the last week, a French-led 109 billion euro investment into French AI infrastructure.
00:30:05
Speaker
The breakdown of this is about 50 billion ah from the UAE. We'll circle back to investment partners in this because there's some interesting overlap with the U.S.,
00:30:16
Speaker
um from the UAE for a one gigawatt data center outside of Paris in in ah in a nearby town or city, which would make it the largest data center in Europe.
00:30:27
Speaker
Another 20 billion from Canada's Brookfield for AI infrastructure and 10 billion for from BPI France. They're targeting to build 35 new data centers, including this one gigawatt one ah in France.
00:30:42
Speaker
france France is definitely positioning itself as kind of like the... AI hub of Europe. It is definitely trying to make this not a you know China, US only story and that Europe is in the mix.
00:30:56
Speaker
For those that don't know, the major AI model players are largely US based. So there's an open AI um which is largely funded by Microsoft, rumored to be almost raising another 40 billion from Japan's SoftBank at the moment at a 300 billion US valuation.
00:31:15
Speaker
Then there's something like a Claude made by Anthropic, which is largely funded by Amazon. and kind of in there And then ah Elon Musk has a XAI project that is also building one of the, it was built the world's largest data center um or training cluster.
00:31:32
Speaker
Slightly further into the tier two category is a French company, Mistral, which was one of the early companies to get funding in this space. And so I'd say they're probably them plus this investment is probably like the bet that Europe might be making.
00:31:48
Speaker
So yeah, that's a long preamble. But Carlos, is this a positive sign for infrastructure? And it's it's interesting that it's France leading the way.
00:32:00
Speaker
Yeah, I think it seems positive. um I thought comparing, for example, the US s with Stargate, we had the the announcement from France, and then we also had an announcement in the UK.
00:32:13
Speaker
They've all gone with three different approaches from centralized to decentralized. You've got the US, which is kind of an R&D, grants and partnerships kind of approach, but basically let big tech deal with it.
00:32:24
Speaker
You've got France, which sounds like this very much like state driven orchestrated from the top type investment. And then you've got the UK that are doing sort of their growth zones, partnerships, universities, and trying to improve areas of the country. put it Put another way, though, is like the U.S. like we know everyone wants to come here. So we'll we'll kind of like open our doors and all this money will come in, like a lot of it from the Middle East or Japan and even the Japan stuff is like largely funded by the Middle East. So they kind of like know they're playing from this position of strength.
00:32:58
Speaker
Then you've got ah you've got France who has like definitely a horse in the race with their local model company. And like, it's like a, you have to now ramp up your spend or, or basically you're left behind. Like this is the time to put the chips on the table and have a crack.
00:33:13
Speaker
And there's they're stumping up some and then they're obviously partnering with actually the same one of the same partners that put money into the US, the MGX capital out of the UAE.
00:33:24
Speaker
So that's kind of like they're basically, yeah, s stumping up to position themselves as kind of the leader in Europe and to give Europe a like fighting chance. And then you've got the cd UK who's gone, we've got some spots that you could put data centers.
00:33:43
Speaker
a Nice. they The thing the UK does have is ah the universities. So that is quite a significant resource. Some of the rest of the world that will partner with these. So I think that's the thing they're betting their chips on.
00:33:57
Speaker
but that's data they Do they not know Zoom exists? And like, again, it's just you can drive across. Those people can move elsewhere to build the thing. It's like, we've got them. They're in the island.
00:34:10
Speaker
Yeah, the the interesting thing with the UK the growth zones are quite far from the major cities, so that's going to be a difficult resourcing job. um and it's not the most attractive investment.
00:34:21
Speaker
But um the thing that I can't stop thinking about is suddenly we've got these announcements with a huge amount of money being thrown at these things. The amount of duplication even between them seems crazy. But at the same time, there was an article the other day around this treaty and the US and UK saying, we're not part of this treaty because they're trying to sort of pool AI resources and i be really open and open source for everything that we're doing. So i think China, India,
00:34:45
Speaker
ah France are part of this, but the UK and US s said no. right um it' it feels like they're saying no because they don't want to be tied into regulation, pushed on everyone that's part of that community. so the The duplication and the amount of money that's being thrown at, ultimately, there's probably going to be one or a few winners really in the space um to try and monetize Obviously, there's the like defense and things like that that you would keep in internal and it's not about monetization.
00:35:14
Speaker
but the the retail side of it, um that's ah that is a headache. um Yeah, I think the you know a state of play right now is that most of the large model players like OpenAI or Anthropoc are actively resource constrained. So some of them are not releasing certain models because they couldn't support the inference or the people wanting to use the thing with the data set.
00:35:40
Speaker
data centers available. That's part of why this investment has kind of come around from a Stargate perspective. From the country and the investor perspective, building a data center, you don't actually care who's using the data center.
00:35:53
Speaker
So it's not like you're betting on Mistral to win or whatever, because ultimately if OpenAI does you know sweet the sweep the board or whatever, they're going to need data center capacity in Europe to serve European customers.
00:36:10
Speaker
Ultimately, there's a capacity bottleneck right now and the big bet that people the big risk is that you know a large majority of that is being used for training models. And ah the question is, well, do you still need the capacity to keep training the incremental models join for you know over time or do things evolve such that there's more and more efficiency in the way that they train in the application layer and the way that the the thing is structured, that that demand goes forward.
00:36:41
Speaker
And on the other side, there is the like current state of play on inference, which is people asking questions and using the the model. um the current state of play is like a lot of the the current applications are still in the early and baby stage of what this will eventually become.
00:37:00
Speaker
And there are, for those that haven't played with them, things like Google's deep reasoning model, deep research model, OpenAI's deep reasoning model. I think I jumbled up the names there.
00:37:10
Speaker
But these are things that might think for 10 minutes, 20 minutes at a time on a problem. um And that goes from needing 1,000 tokens or 10,000 to thousands and hundreds of thousands of tokens.
00:37:23
Speaker
Basically, units of compute would be one way to think about that. So the the demand for inference is predicted to go like up a million times.
00:37:35
Speaker
Yeah, the majority of the world's population are either non-AIE users or early adopters. So, yeah, when that ramps up. And so, yeah, from ah from a government or from a country perspective or from a private investor like MGX um the out of the UAE, it's a road or it's a, you know, anyone can drive on
Global Data Center Demand
00:37:56
Speaker
the road.
00:37:56
Speaker
It doesn't, you're not really making, ah you're really not making a bet um on, you know, Who's going to win? From an infrastructure like data center point of view, yes. I guess I was thinking there was potentially a lot more money in the actual AI models and everything else that comes with it, which is the thing that you're betting on more than the the infrastructure that enables like data centers to run.
00:38:19
Speaker
Yeah, margin-wise, I think that the the history of this of the technology industry has shown us that the high margin businesses are in the application layer, which may in fact be on top of OpenAI, so not even that frontier model layer.
00:38:33
Speaker
It remains to be seen what margin there is on that frontier model layer. Like OpenAI, I think, has quite good margins, but others, even though they're overall losing money, but on the incremental or the marginal serve of the platform,
00:38:47
Speaker
the actual compute. The underlying infrastructure, we it's comparable to Google data centers and hyperscalers now. it's you know it's It's a more stable and lower margin business, but it's a good business.
00:39:00
Speaker
Yeah, for sure. So yeah, very, very interesting to see how that plays out and drives increased demand for data center infrastructure. Yeah, for sure. Right, we're at the end of our time, so I'm going to have to cut us off. um Thank you very much, everyone, for tuning in into today's show.
00:39:17
Speaker
If you did enjoy the episode, please do think about liking the video or following us on your chosen podcast platform. ah We appreciate your support and we'll catch you next week. Bye-bye.