- 2 days ago
Category
🗞
NewsTranscript
00:00Data centers, is that still a thing? How big of a thing is that?
00:04I think it's still a thing. I think it's going to continue to be a thing.
00:07I think, you know, it's interesting for me.
00:10I've been building digital infrastructure for 30 years.
00:13And, you know, having had the privilege of building digital PCS networks
00:17and building the backbone of the Internet and now building the foundation for AI,
00:22I find this AI trade on and off thing a bit not funny in the sense that,
00:28you know, we like to sort of buy on the good news and sell on the potential bad news.
00:34And I think we're investing for decades in AI.
00:37This is not something that we're going to be investing in the next few quarters.
00:40The commitments that our customers are making on a global basis are at massive scale.
00:45And so, you know, quarter to quarter, we're looking at obviously the patterns of capex spending.
00:49And we've seen no abatement of that capex spend.
00:53In fact, we've seen an acceleration this year of that capex spend.
00:56And we started the year with about $225 billion of capex across the major hyperscalers around the world.
01:02And today that capex spend for this year is forecasted to be between $360 and $380 billion in capex.
01:08So from an infrastructure perspective, we really see an acceleration
01:12into what's required to build AI for the next 10 years.
01:16How do you even choose where to invest, whether it's geographies or even kind of what data center?
01:23I mean, I'm sure there's a plethora of opportunities out there for you right now.
01:27When I look at Asia, though, there's still a lot of other investment competitive likes of EU and the U.S.
01:33How do you sort out those opportunities now?
01:35Yeah, we're really excited about Asia.
01:36I mean, particularly Southeast Asia.
01:38So we made a commitment to build 300 megawatts here just across the river in Johor.
01:42I'll travel there this afternoon to go look at our data center there.
01:46And, you know, we've committed another $1.6 billion of new capital that we formed with GIC and Adi a couple weeks ago
01:52to invest in Vantage Asia PAC, which we're really excited about.
01:56We've been investing in cables.
01:57We're building over 9,000 kilometers of dark fiber at Zenith IG.
02:01And we're building over, we've built over 19,000 cell towers between Japan and here in Southeast Asia between J Tower and Edgepoint.
02:08So we're very excited.
02:09Digital infrastructure is under-supported here in Asia.
02:13And so it's one of the lagging markets in terms of infrastructure, but it's on the front end of technology, which is an interesting juxtaposition.
02:21So right now we're forming capital.
02:24We're here today to talk about, you know, this $11.2 billion of capital that we've raised in our third flagship fund,
02:31$7.2 billion in the fund, another $4 billion of stapled co-investment.
02:35For us, that's a huge win.
02:37And it adds on to the $80 billion of capital that we've raised in the last 12 months to support AI infrastructure.
02:42So from an infrastructure perspective, the AI trade is definitely on.
02:46I mean, the fact that you raised $80 billion and then raised another $7 billion
02:52just goes to show that you almost have to turn money away.
02:55When you look at the infrastructure, like, there are gaps everywhere.
02:58Yeah.
02:58How do you look at energy infrastructure?
03:00Because people seem to be talking about it recently, but I'm sure that's not a new thing for you.
03:03Yeah, look, we've got a power bank of about 22 gigawatts around the planet today.
03:08We own 11 different data center companies at DigitalBridge.
03:11We own over 420 data centers.
03:13So we've been in the data center business for over a decade.
03:15So this isn't new for us.
03:17We really started accumulating power about two to three years ago.
03:21And in fact, we've announced a new strategy at DigitalBridge called Digital Power.
03:24Yes.
03:25Where we're building grid-independent power.
03:26And this is really the key to where power is going in the future is the ability to work with the grid,
03:32work with the regulators, but at the same time introduce other sources of power,
03:37bringing that into a microgrid strategy.
03:39Yeah.
03:39Where in surplus moments of the time, we're selling power back into the grid.
03:42And then in off-peak, when it's traditionally between 11 p.m. and 6 a.m.,
03:46we're buying power back from the grid at wholesale.
03:49So there's this interesting opportunity to trade in power,
03:52provided you can build microgrids and you have access to solar, to BESS, to wind, CCG, LNG.
04:00There's no single bullet solution to solving the power for AI.
04:04Yeah.
04:04And we're working in Europe.
04:05We're working here in Southeast Asia.
04:06We're working in the United States.
04:07And we're going to build over 12 gigawatts of new power in the next 24 months at DigitalBridge.
04:12We're investing close to $15 billion in CapEx just to go along with the $80 billion that we're putting into data centers.
04:18You mentioned solar.
04:19You mentioned some of these commodities that are very much needed.
04:21I mean, China seems to be that market that has all those things.
04:24How close are you looking at that market right now?
04:27Look, for us, China is a difficult place for us to invest.
04:30We do own about seven data centers in Hong Kong through Vantage Asia.
04:34Yeah.
04:34We work with the Hong Kong authorities there.
04:36We have one of the data center seat licenses.
04:38It's a good place for us to allow our U.S. hyperscalers to put infrastructure into China safely and do so with full cooperation from the Chinese government.
04:48So, you know, our hope is that over time, we think AI is not a politicalized opportunity.
04:55We shouldn't have China versus U.S.
04:57It's like the Internet.
04:58It is because we're going to have AI everywhere, right?
05:01We accept the fact that AI is very pervasive.
05:03And, yes, there are three predominant hyperscalers in China.
05:06There are six in the U.S.
05:07But my view is this becomes commoditized.
05:09It becomes globalized.
05:11And eventually, over time, there should be AI should be without the politics and without the boundaries.
05:17So we look as an investment firm.
05:18We're investing in data centers all around the world.
05:20We're investing in five different continents today, and we serve all of the Chinese hyperscalers equally, and we'll continue to serve them.
05:28One interesting thing that's come up is, and I think we were just having this conversation yesterday with Kai-Fu Lee, that the model that China has adopted as far as AI goes.
05:40Open source.
05:41Open source.
05:42And I wonder if they pursue that route and the U.S. pursues this route, very general terms, of course, is, I don't know how to phrase this question properly, is the world going to need the same things that are needed to build the U.S. model when you look at what China is doing in that path they're taking?
06:02Well, look at the adaptation of DeepSeek, right?
06:05We use that as sort of a test tube, per se, an adaptation.
06:08And what we saw in DeepSeek was this unique moment in time where you saw massive adaptation, and you saw cost coming down.
06:16And by the way, that has happened in other technological advents, whether it's the 3G radio, 4G radio, or 5G radio, or you look at the adaptation of public cloud.
06:24Same thing happened.
06:25As we have more adaptation, David, cost comes down, right?
06:29And usage goes up.
06:31And we see that more pronounced in AI today.
06:33And whether you're building a small LLM or you're building a large LLM, it really doesn't matter because ultimately this is moving to inferencing.
06:41And as those models move towards the edge of infrastructure and as you turn on those inferencing opportunities, remember, most of AI is going to be machine-to-machine connections.
06:50It's going to be robotics.
06:52And if you think about the future of where AI is going, the use cases are not the same use cases as public cloud.
06:58Public cloud was really active applications, right, in the enterprise and in consumer.
07:03And you think about where AI is going, you have five unique use cases, industrial use cases, enterprise use cases, consumer, which will be less than 10% of AI.
07:12Then you think about machine-to-machine connectivity.
07:15And then the last thing that nobody talks about is data sovereignty.
07:18This is one of the biggest areas of AI, innovation.
07:21And so I think about these five different areas of growth in AI.
07:24It's much bigger than public cloud.
07:26And so whether you're building an LLM in China or you're building an LLM in the United States or in Norway, ultimately those models are going to be harmonized, to a certain degree weaponized, but ultimately they're going to be turned into specific use cases.
07:40Whether it's industrial use cases, enterprise use cases, or consumer use cases, the aperture for AI is widening.
07:46It's not narrowing.
07:48And I think people need to really be pretty sanguine about the whole AI trade.
07:52Look, there's a lot of noise out there.
07:53Focus on the signal, not the noise.
07:55The signal is that this is a multi-decade opportunity, and that's what we're investing in.
08:01Well, I mean, asset allocators have that question in mind.
08:03It's like a very simple one I have is there are clearly opportunities in the public markets, right?
08:09There's the invidious of the world.
08:10You talked about cloud, you know, go along on Microsoft.
08:13Do people need to rejig and just change the time horizon of these investments?
08:19Are we too short-sighted in the public markets?
08:20Does that underscore the case for guys like you?
08:23Well, let's take a history lesson, right?
08:25Let's go back in time 13 years ago when we started building public cloud.
08:29It took the hyperscalers, remember, it took them six years to monetize public cloud before there was real demonstrative earnings in public cloud.
08:37Here we are.
08:37We're in the third year of AI.
08:39If you really think about AI infrastructure, we got started in 2021, and we really started accelerating in 22, 23, and now 24 and 25.
08:48And so we're three years into that journey of building the infrastructure, and we're already monetizing AI.
08:53We're monetizing AI 24 months quicker than we monetize public cloud.
08:57So investors need to exercise patience.
09:00And you need to understand that this is, again, this is a complete tectonic shift in network topology,
09:05and, most importantly, the applications and the use cases.
09:08And so we're going to see this play out over a long period of time.
Recommended
6:51
|
Up next
6:18
5:34
1:31
6:36
4:08
5:19
6:51
4:47
7:31
7:20
1:58
2:05
2:51
5:38
5:03
12:39
Be the first to comment