Mark Bonner: 00:09

Okay. Welcome to First Draft Live. It's Friday, September 26. I'm Mark Bonner, editor in chief at BizNow coming to you live from New York. Thank you to so many of you for joining the program today from The United States and overseas.

Mark Bonner: 00:22

We really appreciate it. For those who are new to the program this week, this is a show where we take the mess that is commercial real estate and wrestle it into something you can actually use. And quick reality check, this isn't the main event. The main event is our daily news under the first draft. That's where the real action is.

Mark Bonner: 00:40

Scoops before they hit the wires, charts before they hit the decks, and context before the spin doctors get their hands on it. This is not for tourists. This is for insiders. Written by me and a handful of the newsroom's most obsessive insomniacs, the kind of people who think a good time is parsing cap rate surveys at 06:00 in the morning. So for our first draft live audience this week, consider this your backstage pass.

Mark Bonner: 01:07

Seven days on us. So sign up now. Simply go to biznow.com/firstdraft. That's biznow.com/firstdraft. Okay.

Mark Bonner: 01:19

Today, we've got a story that feels uncomfortably familiar. For years, have warned of an AI bubble. Now the numbers are hard to ignore. Bain says the industry needs $2,000,000,000,000 a year in revenue by 2030 to fund its compute build out. But it's only on track for about 1 to 1,200,000,000,000.0.

Mark Bonner: 01:41

Let me do some math here. That's an $800,000,000,000 shortfall. What are we gonna do about it? Meanwhile, OpenAI, Oracle, SoftBank, they just rolled out plans for five more hyperscale campuses pushing towards seven gigawatts of new load. That's a lot.

Mark Bonner: 01:59

And NVIDIA's market cap has swollen bigger than the GDP of most g seven economies. And on the ground, Northern Virginia land values are up 60% year over year. Similar spikes can be seen in places like Dallas, Phoenix, even Columbus, Ohio. Billions of dollars are being poured into sites that either may be tomorrow's backbone or tomorrow's stranded stock. What does it all mean?

Mark Bonner: 02:25

Well, it could be that CRE is underwriting entire markets on demand that might not materialize fast enough. If the bubble burst, it won't just be tech valuation at stake. It could be land leases, financing, and the very backbone of American reindustrialization. So the question is, is this the start of a new generational supercycle or is this the setup for commercial real estate's next hangover? That's the question that's on the table today.

Mark Bonner: 02:57

And to help us sort the hype from the fundamentals, we've got Michael Pierce, deputy chief US economist at Oxford Economics. Michael, welcome to First Draft Live.

Michael Pearce: 03:08

Thanks for having me. Pleasure to be here.

Mark Bonner: 03:11

So let's just get into it. Is this AI bubble real? What do you make of it?

Michael Pearce: 03:16

Yeah, I mean, honest data has to be maybe, you know, there's, I can't think of any major, major boom in investment that we've seen going back centuries where there's not been at least some speculative activity that ends up, making a loss out of that. But for me, it comes down to is AI, you know, the core question here is AI truly a revolutionary technology that's gonna make us more productive. And I, you know, my answer so far is this, yeah, this does seem to be the real deal. And everything that I can see from kind of the demand side and kind of how AI adoption is rippling through the economy suggests we're still in the very early stages, of this boom. So if it is a bubble, I think there's still plenty of, plenty of room left to run over the coming years.

Mark Bonner: 03:59

There's been some really scary headlines over the last six months. I assume some of this is media hype, maybe not. I'd love for you to help me figure out and my audience what this really means. But this week in particular, we've got two things on deck here that I wanna get into. First, let's start with Bain.

Mark Bonner: 04:17

AI companies will need $2,000,000,000,000 in combined annual revenue to fund computing power by 02/1930, but the revenue is likely to fall $800,000,000,000 short. This is David Crawford, chairman of Bain's global technology practice. And I quote, if the current scaling laws hold, AI will increasingly strain supply chains globally. Then this morning, Greenlight Capital founder David Einhorn said the trillion dollar build out by companies overall, such as Apple, Meta and OpenAI is so extreme that the eventual returns are highly uncertain. I'm gonna quote him here.

Mark Bonner: 04:53

The numbers that are being thrown around are so extreme that it's really, really hard to understand them. I'm sure it's not zero, but there's a reasonable chance that a tremendous amount of capital destruction is going to come through the cycle. What's your response to this, Michael?

Michael Pearce: 05:10

Yeah, I mean, they sound like sensible hedging statements, know, there's a lot of language there, you know, this reason, you know, reasonable chance, not zero chance. Of course, there's always a chance of, of all this ending up being worth nothing. But, you know, I always come back to actually, you know, this time around, what we're seeing is a lot of these companies generating very significant revenues. You know, and that feels different from kind of the late stage bubble hype that, kind of you might associate with say the .com bubble where, you know, you had a lot of firms that were burning through cash, they have no revenues. You know, they were taking on a lot of debt.

Michael Pearce: 05:47

I think what's striking so far about this boom that we've seen is how much of it is built on retained earnings on private equity investments. You know, there's not a whole lot of debt going into this boom relatively speaking and the revenues are there, right? So it's, you know, so far the money is, you know, the investments have been backed up by real money. And that suggests to me that, you know, yeah, we, we may run into problems further down the line, but, but right now this, this looks like, you know, this has further to run.

Mark Bonner: 06:22

Every boom looks permanent until it doesn't. We've seen this movie before, Michael. Japan's asset bubble in the nineteen eighties, the .com bubble in the nineties, housing in the February. Each of those moments in time for global economies had real innovation and lasting winners. Right?

Mark Bonner: 06:39

Coming out of the .com bubble, Amazon, they survived. Many did not. What your your your specialty is macroeconomics. What can the Japan asset bubble of the eighties or the .com bubble of the nineties or the housing bubble in the February tell us about the moment we're in right now?

Michael Pearce: 06:57

Yeah. So I think first off, you know, is you're right kind of looking at where the actual old value is and and how sustainable this this is. And clearly, you know, you go back to those those times, you saw this huge unsustainable run up in valuations, you know, valuations running way ahead of actual earnings. In in the case of Japan, you know, those land values running well, well above, incomes. So what's the actual investment case here?

Michael Pearce: 07:25

I think the investment case is this is potentially transformative technology for the global economy. And so I always come back to, you know, demand and supply. The demand for this AI technology is kind of this adoption and rollout across the broader economy. Now, listen, we did a big study on this a few years ago, kind of predicting out the net impact of AI on the economy. And, you know, you have to make all sorts of assumptions kind of around adoption rates, how much it will do for productivity.

Michael Pearce: 07:54

We came up with these three kind of S curves for how demand for AI will roll out. And so far, what we're seeing is those actual adoption rates compared to what we were saying a few years ago are actually running, you know, much closer to kind of the optimistic or the high scenario that we laid out rather than a median expectation. So, you know, I think that's a sign that we're still on the early upswing of the kind of this S curve in adoption across the economy. And one other thing I will just point out on that, you know, great census data that asks firms are you using AI to produce goods and services, but also what's your expectation in six months from now? When we have this persistent optimism gap for the last few years, you know, people always saying, oh, in six months, we're going to be using a lot more AI than we are today.

Michael Pearce: 08:43

Well, that's actually caught up to reality now. You know, there's the actual adoption is running now in line with where people thought it was going to be. That, you know, we've had that huge catch up. So I think that's kind of the key driver of the demand of this. The, you know, from my perspective, doesn't look like it's going to end in the next twelve, eighteen months.

Mark Bonner: 09:03

So look, I'm going to get into the real estate aspect aspects of this in a few minutes, but let's just stay on the bubble for a second. All these all these headlines that are being made in the finance world about these warnings, do you think it's cynical? Do you think it's in self interest? Is the media making a mistake by going all in on this headline knowing that it's gonna click hard? Like, I guess what I'm trying to figure out here is for all of our viewers and our listeners, how do you separate the hype from from the reality of of the moment that we're in when it comes to that narrative that the bubble is gonna burst and it's gonna be scary a lot of people, commercial real estate in particular are going be left holding the bag.

Mark Bonner: 09:44

How do you look at this?

Michael Pearce: 09:46

Yeah. So for me, again, it comes down to looking at those fundamentals. Is the demand there? My view is yes, it is there. And then looking at the investment side, the supply side of where is all this money coming from?

Michael Pearce: 09:58

Is it being fueled by, you know, there's a lot of indicators that would suggest this is highly speculative. There were no revenues there, if a lot of it's financed by debt. And there will be some of that, and there will be losers. That's why you always hear this kind of forest of warnings because they, you know, eventually, you know, can't think of any big boom where there hasn't been accompanied by some speculative activity that's gone wrong. But, you know, by and large, I think what marks out this boom is how much of it is backed by backed by real money.

Michael Pearce: 10:31

This isn't, we're not seeing a huge rise in business debt. You know, if we are going to eventually see some of these more speculative elements that kind of would be classic indicators of a kind of a late stage boom, we're not seeing them yet. Okay.

Mark Bonner: 10:48

If, big if here, if there was going to be a bubble that burst, where would we feel it first? Now let's start macro and then let's dive into commercial real estate.

Michael Pearce: 11:01

Well, I think the big thing, you know, if you saw this, the puncturing of the optimism around AI right now, I think that because so much of it is being driven by these hyperscalers, and that's been backed by, you know, this tremendous run up in equity values. For me, the key vulnerability to the economy today comes through that equity market channel. And I think, know, you'd see that directly, those firms reducing their spending, maybe growing their CapEx a lot more slowly, if we were to see a major stock price correction. But listen, I think the broader economy would be at risk here as well. You know, one consistent theme I've heard from clients is the economy appears to be defying gravity.

Michael Pearce: 11:42

And I do think that AI stories a lot of that, you know, strip out AI related investment, private investments actually declining so far this year. Strip out the big gains in the equity market, chances are The US consumer probably doesn't look so good either. So really, I think, you know, this huge rebound that we've seen in the financial market since kind of April, a lot of which has been driven by AI, not just, you know, feeding the actual investment itself, but supporting this broader consumption outlook, is lifting all boats and making people more optimistic.

Mark Bonner: 12:15

You're bring it up a good point. Mean, we're talking about a trillion dollar build out across the country right now. Right? You can't build it fast enough. There are headwinds on the ground everywhere, which we can get into in a second.

Mark Bonner: 12:24

But AI CapEx is largely carrying The US economy, papering over a lot of the pain that's below the surface. And the physical embodiment of that CapEx is physical spaces. And in this case, we're talking about data centers, right? Those are the structures and the equipment that need to be built to make AI happen for all that innovation. You take that away and The U.

Mark Bonner: 12:52

S. Economy looks pretty weak, or at least it's flatlining to some degree. Supply and demand, right? Tell me about where you're viewing that right now. Demand is stronger than garlic, But if supply cools, right, or if AI CapEx slows down, that does seem like it could be a potential house of cards.

Mark Bonner: 13:17

What do you make of that dynamic?

Michael Pearce: 13:20

Yeah, I think that's where I see kind of all the risks to AI at the minute, you know, it seems like there's still this huge latent pool of demand for AI services and products. But really, you know, there's a lot of challenges that I see on the supply side that could trip this trip this bubble up. You know, just on the physical side, you know, power availability. And if it's not our availability, it's not on impact on power prices. Not just, you know, then you've got you're running into issues of, well, you're going to get local opposition.

Michael Pearce: 13:52

People are suffering pain because their household bills, utility bills are going up. And also they're building this big slab of concrete that doesn't create any jobs in your area, you could see a lot more local opposition to these constructions. So I think there's some physical risks on the supply side. But there's probably a bunch of policy risks here as well. You know, I think one that comes to mind is, you know, given the tariff news we had overnight, you know, on pharma sector, well, you know, one big investigation that's outstanding is semiconductor tariffs.

Michael Pearce: 14:27

We don't know what's going happen to that. That has the potential to create a lot of bottlenecks, and also push up, push up the price of a lot of these investments as well.

Mark Bonner: 14:35

And that's to say nothing of the other tariffs, right? Because you know, copper prices already super high with a 40% tariff going up even higher in the last forty eight hours because of a disaster in Indonesia, which is the number two copper producer in the world. That is gonna screw up pro formas on the construction side even more. Right? So all of this coming together all at once, it makes for some trepidation if you're making these big decisions, Michael.

Mark Bonner: 15:01

I mean, what is the next shoe to drop you think? Or what is the thing that you look at in the data that you fear could be the tipping point?

Michael Pearce: 15:12

Yeah, I mean, where we have more insight is kind of on the policy side of things. And there, you know, I think the clouds are gathering and they're all inflationary and potentially limiting the supply side here. You know, whether it's the direct impact of tariffs, whether it's tighter immigration policy, which now seems to be focusing also on legal migration, you know, that's kind of raised the challenges of finding skilled workers to come and help build out this infrastructure. We already see that. We already see that firms across the economy report that quality labor is one of the biggest bottlenecks, biggest constraints on expanding out there.

Michael Pearce: 15:57

So those are kind of some of the factors that could really help weigh or slow down this expansion?

Mark Bonner: 16:06

So let's go back to this Bain report that came out this week. Just to reset here, Bain estimates the AI ecosystem needs $2,000,000,000,000 in annual revenue by 2030 to cover its build out, but it's tracking closer to 1,200,000,000,000.0. That's an $800,000,000,000 hole. For CRE, potentially, that's not abstract. It means hyperscalers may lease space today on assumptions that revenues don't deliver tomorrow.

Mark Bonner: 16:31

How serious is that $800,000,000,000 funding gap for real estate, Michael?

Michael Pearce: 16:37

Well, I think, you know, in the near term, it's only gonna grow. Right? Because all the focus is on on building out capacity first and worrying about the the the revenues later. You know, if this, you know, if we're right that this is a transformative technology that we are seeing this adoption pick up across the economy, the demand is still going to be there and we're going to see much stronger, much stronger demand from a much wider base across the economy. And so I think, know, the case is definitely there, but the customers will certainly be there.

Michael Pearce: 17:11

The question is just the price and the price is a balance between the demand and supply. But as everything that I'm seeing so far suggests that all the problems are on this, you know, all the limits are on the supply side, the demand right now at this stage feels limitless, but that won't be the case forever.

Mark Bonner: 17:29

Does it cap valuations for data center assets?

Michael Pearce: 17:33

Well, I'm just talking about the demand and supply of compute and therefore data centers and for these AI products. I think that adoption is running very far and I expect that to continue.

Mark Bonner: 17:47

If you're just joining us, this is First Draft Live. We're unpacking the AI boom and whether it's really a bubble with Michael Pierce, Deputy Chief US Economist at Oxford Economics. Okay. Let's talk about the ecosystem in itself. NVIDIA sells trips into projects it helps fund.

Mark Bonner: 18:04

SoftBank sees operators it also backs. Then Oracle lead leases space it helps develop. It looks like a virtuous loop, but CRE has seen how quickly that kind of reflexive structure can flip when revenues disappoint. We're fresh on our minds as CMBS in the year 02/2007. Michael, is this financing model sustainable, or is it building fragility into the system?

Michael Pearce: 18:29

I mean, I always go back to the fundamentals. And if, you know, if we're right that the demand is going to be there, then I think every all the other details, you know, mostly take care of themselves. And for me, you know, when we're tracking this, it's difficult to get a scale of what we should be watching to track how useful this technology is to the economy. Right now, it's very easy to measure the dollars going in the investment, the costs, The benefits, I think, where people are a lot less than. And that has been true of every technology that we've seen going back centuries.

Michael Pearce: 19:05

And very often, the economic statistics are not the place to see that impact first. You know, if we go back to the 2000.com boom, what we saw there was the GDP statistics, you know, these IT technologies were not adding to the economy, they weren't showing up in the data. And it was only when we changed or, you know, added better data collection that we saw that positive impact build, build labor. You know, I think the best for me leading indicator right now is what's going on with corporate profits. And we see very strong profitability, not just for these firms involved in the build out, but the firms across the economy that are already beginning to start to apply this.

Michael Pearce: 19:45

So I think that's, you know, the place to track this adoption might be building. And I think that's a positive sign for ultimately for future demand there as well.

Mark Bonner: 19:56

So in this ecosystem, right, theoretically, if something falls apart and the economics don't work anymore, who's left holding the bag? Banks, vendors, landlords?

Michael Pearce: 20:10

Well, right now, it it there doesn't seem to be much debt backing this. So, you know, I I think mostly it's gonna gonna be the direct equity holders that have invested in these projects and are backing these names. But like I say, you know, they're the big spillover is, you know, a lot of people hold these names, you know, it's hard to see some part of the AI ecosystem going down without, you know, seeing a major friction in the large names, you know, things like Nvidia method, they're just too big for the pain to be, narrowly based. Know, I think the whole economy would feel that slowdown and that reduction in optimism. Okay.

Michael Pearce: 20:48

Let me go to

Mark Bonner: 20:49

a question from the audience. I remember doing bank workout of server hotel projects during the tech wreck of 02/2001. Warehouses that had been converted to server use with large costs spent on raised floors, electrical and mechanical systems, etc. That ultimately had to be demoed out to restore the warehouse use. How will this boom be different for lenders, Michael?

Michael Pearce: 21:15

Yeah. I mean, listen, you know, that brings up a good point of of stranded assets. You know, if if this build out does go too far, and listen, I think that's that's a very real risk eventually. But right now I just don't see I don't see that in the near term. But clearly, you know, there's a lot of factors here.

Michael Pearce: 21:37

You know, the supply for for me, the biggest factor is if the supply pipeline is so strong, that we get to a point, you know, right now demand is easily outpacing that supply, but we get to that point, you know, the flattening of the kind of S curve of adoption across the economy where that supply outruns the demand. And you see the price, you know, you see that reflected in prices. And I think that would be the point at which I'd be very worried about kind of this, fallout and the rise of standard assets. But, you know, my big point is it still feels like that point is at least a few years away.

Mark Bonner: 22:16

Okay. Let me get one more here. As AI infrastructure develops over the lifespan of new buildings, will its size requirements per building decrease in response to new technology and chips that have higher capacity and less space?

Michael Pearce: 22:31

I'm not sure I've got much to add on that from a macro standpoint. Feels like everything's building towards great scale.

Mark Bonner: 22:40

Some of these mega campuses are the size of Manhattan. Mean, at least that's theoretically, they're going to be that size. It's unbelievable. Michael, I know you're a student of economic history, right? I mean, we kind of touched on a few different moments in time where there were bubbles that burst.

Mark Bonner: 22:59

If you're a commercial real estate player today, right, which that is the majority of our audience on the call today and we'll be listening to on Spotify and Apple later, what lessons from past bubbles should commercial real estate players take most seriously right now? What's your greatest insight?

Michael Pearce: 23:16

Yeah, I think it's when, you know, the problem with the bubble is everyone can feel that you're in it. And we're very clearly in the midst of this great transformation, but knowing, you know, trying to identify that point of where you've gone beyond the fundamentals, there's more speculation than fundamental drivers there. You know, for me, the things I come back to are, is demand is the fundamental, you know, is the technology fundamentally useful? And are we going to still see demand? Everything I see on that front is still looking healthy.

Michael Pearce: 23:49

And then I really worry about well, what's what are the other players doing in this space? Are we building overcapacity? Listen, one thing I will point out is, know, that hyperscalers have been sharply raising their CapEx this year. But when you look at the forward projections, the growth rates actually taper off. So I kind of push back on the notion that we're kind of in this unconstrained CapEx boom, you know, is some moderation kind of in the forecast.

Michael Pearce: 24:16

You look at kind of the consensus, what markets expect the max seven to be spending. You know, if you were, if you were seeing those continually being ramped up and the other big speculative development that you tend to see is when you get a lot of investment, that's not backed by real money. So you see kind of measures of investment, or CapEx relative to, relative to revenues relative to income, really beginning to decline. And you can see, if you look back to the early 2000s, you can see that peaked in '98. So kind of those last few years of the it boom, you had this continuing investment in the space, even as the revenues began to start to disappoint.

Michael Pearce: 25:01

So, know, that would be kind of the warning indicator to look at for this boom. And and like I say, that that's still that's still rising. So there's nothing nothing on the immediate radar, but it's challenging. It's a fast moving space, and there's there's you know, there's gonna be some losers.

Mark Bonner: 25:21

So we have to remain vigilant. Speaker Like who could be a loser?

Michael Pearce: 25:27

One: Well, you know, I just saw this morning, the friend.com measure the little pendant of a friend speaking. There's clearly some, you know, maybe that's evidence of speculation. Maybe I'm just getting old, but an AI companion or friend that you carry around with you that's going to text you doesn't seem like doesn't feel to me like a, you know, a fantastic use of resources. So, you know, I think we will see lots of applications that end up falling well short of, of expectations.

Mark Bonner: 26:03

Part of Bain report was and the reason why for the $800,000,000,000 gap was because they don't believe that AI companies are charging enough for the product, right? And they're basically giving it away. And there's great tension there. Some of them are free and open to the public. Gemini comes to mind.

Mark Bonner: 26:26

OpenAI ChatGPT has a free model versus a premium model. The premium model I believe is around $9 a month. They insinuated that it should be three x that if you want to close the gap, right? And there's great fear that on the consumer level that there is an appetite to pay for it at that level because it hasn't revealed itself to be that useful again at the consumer level. Enterprise, a different matter.

Mark Bonner: 26:54

How do you square that?

Michael Pearce: 26:57

Yeah, I mean, it feels like it's the same in every tech revolution, right? We get these new technologies that come in, they're free or mostly free at the point of service. And then once they've built sufficient scale, that's when you get the monetization piece come in. I feel like, you know, you can see that happening over the next few years, you know, it's going be a range of bottles. Don't think, you know, these companies are wedded to the free premium tier, you know, obviously that's how the consumer space is kind of played out for now, but you know, there will be advertising.

Michael Pearce: 27:27

There will be, I'm sure a range of models that, you know, the key is that other consumers and other firms getting the value that they see. And that's going to be the key to, you know, what they're willing to pay and what they're willing to bear.

Mark Bonner: 27:40

I've got an audience member here who wants to go back to the .com era. Dotcom involved the build out of a brand new infrastructure, fiber, etcetera. AI is demanding expansion of existing public utility I. E. Electricity, which is also a major headwind for data centers right now in The US.

Mark Bonner: 27:58

Does that make things different this time, Michael?

Michael Pearce: 28:01

I mean, I think it means the challenges are a lot. I think the challenges on the supply side are great at this time, you know, because a lot of these things require things that are controversial, right, like new power lines, new power stations, you know, potentially, know, big data centers near population centers. So these are all things that could generate a lot of public backlash and are difficult to deliver. We're already seeing that, you know, this kind of backlog of interconnection requests, it's difficult to add that electricity capacity and we've not done it. You know, the last time we saw electricity production in this country rise sustained basis kind of year over year, you've got to go back decades.

Michael Pearce: 28:44

So it's something we're going to have to learn to do and learn to do very quickly. And that's why, you know, for now, kind of all these risks on the supply side that you're just not going to have the skilled labor base, production in place to support all this supporting infrastructure to build it out. I think that's a lot more challenging.

Mark Bonner: 29:05

Hey, Michael, my producer is starting to give me the hook here, but I do want to land on one question. Looking five years out, what do you think is the biggest risk here, that you see for real estate?

Michael Pearce: 29:18

I mean, for me, think the biggest risk is that people put all their eggs in this one basket. And, you know, we see a collapse in the price that investment goes way down. And you look for the bunch of stranded assets, you know, if this AI thesis is real, you know, I think the big lesson from history is that the effects of these large technological changes is to boost the economy overall. I mean, right now we're focused on all these negative headlines, you know, job losses, it's getting a lot harder for young people to find jobs, particularly recent college grads. And those are real stories, but look back at every wave of technology.

Michael Pearce: 29:59

The result is an economy that's a lot more productive. It's a lot larger. And guess what the people earning those, know, generating those revenues go out and spend them. And that's good for, I think that's good for all asset types, you know, boost spending across the economy, boost productivity and raises living standards. That should be a kind of this broader boost for the entire portfolio rather than just narrowly, you know, narrowly focusing on this one.

Mark Bonner: 30:27

We have to wrap up. Michael, thanks so much for being here, man.

Michael Pearce: 30:31

Thanks. That was just pretty fun.

Mark Bonner: 30:34

If you missed any part of this episode or wanna hear earlier shows, you'll find every conversation on your favorite podcast app. Just search First Live. We'll be back next Friday with another deep dive into where capital technology and CRE collide. I'm Mark Bonner. This is First Draft Live.

Mark Bonner: 30:51

Have a great weekend, y'all.