Physical AI and the Future of Commercial Real Estate | S5E7
In this episode of the A.CRE Audio Series, Season 5, Spencer and Michael explore the game-changing concept of Physical AI in commercial real estate. They break down how robotics and automation, powered by advancements like NVIDIA’s Cosmos platform, are poised to revolutionize property operations.
From cashierless stores to drones that conduct property inspections, the potential for Physical AI to enhance efficiency and reduce operational costs is significant. Using their RV@Olympic project as a case study, the duo discusses the challenges and opportunities of integrating automation into remote property management.
Tune in to learn how these developments could transform real estate operations and pave the way for more innovative strategies in the industry!
Physical AI and the Future of Commercial Real Estate
Or Listen to this Episode
Resources from this Episode
Episode Transcript
Spencer Burton (00:08):
Hello, and welcome back to the A.CRE Audio Series, season five. Let’s talk physical AI. Okay, so this season we’ve talked about AI agents, we’ve talked about generative AI just in general, we’ve talked about the Accelerator 3.0 and how we’re using AI to make that learning experience better. Physical AI is interesting, because it’s potential impact in commercial real estate is almost as big as the potential impact of AI agents. Thing about physical AI is it’s expensive. So, when we say physical AI, we’re really talking robotics. And first off, neither Michael or I are experts on either physical AI or robotics. I consider myself a expert on the practical application of AI in real estate, that generally refers to agentic AI, generative AI, the ways that AI can help the knowledge worker in commercial real estate. When you think about robotics, physical AI, we’re talking about the physical work that goes into real estate, but its impact can be equally or maybe even greater than the generative AI and AI agents. How are you thinking, Michael, about physical AI, either as it relates to RV@ or as it relates to the industry more broadly?
Michael Belasco (01:39):
Yeah, let’s define physical AI a little bit more too.
Spencer Burton (01:42):
Okay, sure.
Michael Belasco (01:42):
Because I want to talk about it, and I want to get clarity too even. I mean, it’s not a fungible term, but it is in a way when you think of the physical bodies and needs of an operation of any real estate project, and what can be duplicated. And if you have, for example, in an RV park, what are the processes by which you can basically minimize costs and increase opportunities, and from a real estate mindset, opportunities for revenue or your customer experience? So, that’s one thing that we’re really focused on. A lot of the challenges in these types of projects, and RV parks in general, is they’re very remote and they’re seasonal. So, think of something in the Sawtooth, so like Northern Idaho where it’s an amazingly beautiful location. How do you get people there?
(02:39):
It’s always a challenge, especially if we’re going to be all over the country. So, how do we implement solutions that can help us operate remote? Otherwise, A, it’s hard to buy that project because a lot of people don’t want to move. So, when people want to sell it, there’s a lot of opportunity here. And so, we’re using RV@Olympic as sort of a playground. I want to say you’re equally as involved in the ideation of what happens here.
Spencer Burton (03:06):
Sure.
Michael Belasco (03:06):
One of the lowest hanging fruit that I thought, and I’d mentioned it in another episode, is our ability to create a 24-hour environment to be able to, cashierless store, which isn’t novel. It’s novel in this space. It’s being implemented, go to any airport in the country and you’ll see it. But what that does, and that’s my question, is this physical AI? I think it qualifies-
Spencer Burton (03:28):
I think it’s a generation of it. Yeah, no doubt. It absolutely qualifies. I think today we expect these robots to be better because we have experienced generative AI, and now that these things can be equipped with some reasoning capability that they didn’t prior have because of large language models, they could be smarter. An example, I remember earlier in my career, in the earlier days of automation, the beginnings of the trend to make self-storage more efficient. And there was this company that was offering the ability to manage self-storage without a on-site manager. And that was novel at the time, it was incredibly novel. And they were using a combination of some automated gates, and access components, and some virtual, so people sitting in a room somewhere else watching and then calling the police that they need. I mean, now we look at it and go, “Well, I mean, that’s obvious.”
(04:36):
But at the time, that was quite the novel innovation, where previous self-storage was there was always someone on-site, checking people in or what have you. And so, as that’s evolved, we’re now to the point where you’re saying, “Oh, the cashierless store isn’t an innovation.” I mean, 10 years ago it absolutely would’ve been-
Michael Belasco (05:00):
Oh, yeah. Not in ours, but now it’s common. So, okay, there’s no robots coming to RV parks anytime soon in that-
Spencer Burton (05:08):
I don’t know about anytime soon. I think they’re coming.
Michael Belasco (05:10):
You think so?
Spencer Burton (05:11):
I don’t know. Well, so for instance, if you watched NVIDIA’s annual conference or whatever they call that thing, they hosted it in the last week-
Michael Belasco (05:20):
Yeah, I did see a little. The robots-
Spencer Burton (05:21):
Yeah, the humanoids?
Michael Belasco (05:21):
… that came out? Yeah, yeah.
Spencer Burton (05:22):
Yeah, the humanoids? When LLMs hit the scene, all of a sudden robots, one of the biggest limiters was they didn’t have a brain, and now they do. One of the biggest limiters was in order to train one it needed to have simulations. And NVIDIA’s now developed this-
Michael Belasco (05:40):
Cosmos? Yeah, the-
Spencer Burton (05:41):
… full-world model of the Cosmos that they can simulate scenarios-
Michael Belasco (05:46):
Which is fascinating.
Spencer Burton (05:47):
… to train it.
Michael Belasco (05:47):
Amazing.
Spencer Burton (05:48):
And therefore it’s training data, it goes beyond the actual circumstances, and they create millions of scenarios. So take an RV park, and let’s say that you want to have a robot that goes around and make sure that people-
Michael Belasco (06:07):
The sites are all clean. Yeah, yeah.
Spencer Burton (06:08):
Make sure all the sites are all clean, yeah. And with technology like NVIDIA has with the Cosmos, you could then simulate all these things and then it knows what to look for and-
Michael Belasco (06:17):
Maybe that is coming.
Spencer Burton (06:19):
But now you’re talking vertical physical AI, and that’s more complex and expensive, but-
Michael Belasco (06:23):
Yeah, but we are, and you have actually alluded to a lot of these things, now we are not going … Again, we have our flagships RV@Olympic. We’re not going full automation here. It is like what I would call our playground to be experimenting, because every park we buy we’re going to push a little bit further. So, the first thing is the cashierless store. We’re going to have this app obviously, that we had talked about. We are going to explore the touchless entry, where before you show up you get a code, or it knows your … The dream would be it knows your car, you let it know what RV is coming in, and we can pretty much tell you where to go when you get there-
Spencer Burton (07:00):
Reads the license plate and lets that license plate in, yeah.
Michael Belasco (07:03):
I mean, that’s stuff that we’re looking at. The other thing is integrating protocols for when … Each site could be metered individually. Now, we didn’t do that at RV@Olympic, but having the meters with shutoffs when people show up, you’re-
Spencer Burton (07:18):
Turns on and-
Michael Belasco (07:19):
Turn on, turn off remotely. So, there’s all these things that we’re trying to do to figure out, because I’d love to get access to some of these remote parks where you’re not relying on someone who may show up. Nobody’s going to make a life, and live out there and raise kids, but it’ll be amazing to have these capabilities. And like you said, every day you go up … One of the big things that a person has to do when they’re on site every morning, go out and check the sites twice. You got to go at 8:00 in the morning, make sure people are leaving, you go out at 11:00, there’s these protocols that happen. Well, there’s no reason camera technology, and I mean, even having a camera in an app that just allows you to handle that stuff remote where you had a property manager on site, well, now you got these remote locations and you have one property manager in a central hub that’s managing all of these things.
(08:14):
So, there’s all of these things that we’re trying to bring to the table. And again, RV@Olympic is going to be, a playground’s kind of a, actually it’s a fun word, we will say it’s sort of like a playground where we explore this while we have that on site manager to be able to test these capabilities.
Spencer Burton (08:33):
Well, and you have the advantage with RV@Olympic where this is not a remote park.
Michael Belasco (08:38):
No.
Spencer Burton (08:39):
It’s within an hour or something of Seattle, it’s right in Port Angeles.
Michael Belasco (08:44):
Couple hours, yeah.
Spencer Burton (08:45):
Oh, okay.
Michael Belasco (08:45):
Well still, yeah, Port Angeles, it’s not a small city. It’s not a big-
Spencer Burton (08:48):
From the Metro Seattle you’re an hour-
Michael Belasco (08:50):
Yeah, there are people that want to live in this area. It’s easy to find a workforce, but it is a safety location in terms of what we’re trying to do for the next-
Spencer Burton (09:00):
But you could imagine, so you’re seeing this actually come. Again, not being an expert on all of what’s coming in this space, but I know that property condition inspectors are starting to use drones, and then technology with the footage to, for instance, forecast the remaining life of the roof. Or you could do a similar thing with structure, with parking lot. And so, you could quickly, before making an offer, you could quickly get a sense, and without having a full property condition report, at least have a preliminary sense of what you’re dealing with from a condition standpoint. And so, maybe you spend $100 on that pre-PCR, that is drone footage together with AI, and you get that within 24 hours. And that allows you then to put in a more credible offer upfront. That’s an example-
Michael Belasco (10:02):
I mean, yeah, you go on. I mean, you can even go to the more extreme ends of the really complicated stuff, like on the construction side. I mean, there’s stuff coming down the road here. You have a friend I know, that’s working closely in that space with remote factories and things like that.
Spencer Burton (10:22):
No, you’re absolutely right there. There’s a lot coming here, and the reason why we don’t touch it, we don’t hear about it much is it’s expensive to develop. And until things like these Cosmos, and large language models and the others, it was hard to train, it was outside of very linear things.
Michael Belasco (10:44):
Yeah. The amazing thing that I did see that thing for NVIDIA with the Cosmo app, is that the big revolution there is that they mentioned Tesla in their driverless cars. There’s two things. They mentioned Tesla in the driverless cars. Now, Tesla’s far and away ahead of the primary data for all the driverless cars, all the imagery, everything they’ve taken. But with Cosmo, you can generate all of that stuff that they’ve done in real time in the simulated world. And it has billions, and I’m way far out over my skis when I talk about it but the capabilities, and one of the things they said, which I found fascinating, what’s his name? Jensen Wong, I think is-
Spencer Burton (11:31):
Yeah.
Michael Belasco (11:31):
He’s like, “Before when you had robotics, it needed to be a green field,” meaning things needed to be just flat. You couldn’t introduce any variables.
Spencer Burton (11:40):
That’s right.
Michael Belasco (11:40):
But now, and my understanding was, because of the capability of this Cosmo, this whole new world of, I’m trying to think of the name of it. It’s not even virtual reality.
Spencer Burton (11:53):
A real world model.
Michael Belasco (11:55):
Yeah. You can have robots interact on what he dubbed the term, we know brownfields in real estate is something different, but he called them brownfields where they can operate, they don’t need wheels, they don’t need tracks. They can walk, and they can operate in terrain that’s less predictable or less standardized. And that to me, I mean, the implications there, probably RV parks are way down the road for when the use cases come and the robot is going to show up to an RV Park, but you can see. So, just like watching that, and I’d encourage anybody to watch that because that, again, Spencer’s seen this and you’ve been seeing it forever, but for you to share with me all this stuff and it’s like, Jesus, this is really coming.
Spencer Burton (12:44):
Well, and so if I’m in real, well, I’m in real estate. If I’m a real estate investor, how I think about it or how I look at this is the more complex my operation, the more opportunity there is. And so, if I’m working in hospitality, which high OpEx relative to revenue, you have a greater opportunity to lower OpEx thanks to this.
Michael Belasco (13:10):
That’s right.
Spencer Burton (13:12):
If I’m running a very simple single tenant net lease asset, less so from a physical AI standpoint, although there’s absolutely applications across all. But if you’re listening to this and it’s like you’re running full service hotels, and I’m sure all the big hotel operators-
Michael Belasco (13:29):
They’ve been thinking, yeah.
Spencer Burton (13:29):
… are thinking about that, and it’s exciting. And there’s the whole other scary side of this. This isn’t the venue to talk about what humanoids mean for broader society, broader humankind, but in that keynote where he had the robots lift up out of the ground, that was a little spooky.
Michael Belasco (13:50):
Yeah, it was.
Spencer Burton (13:52):
And you read these articles about how all of us are going to have our own humanoid robot in our house in the next 10 years, I don’t know if that’s real or not in the next 10 years, but it’s coming. Now that they have a brain, now that they have the ability to train on synthetic data, and there are some well-funded organizations that are working on this. The day is coming where we’ll all have our own in-house butler, a humanoid assistant that will fold our laundry. And in the same way that we all have washers and dryers now, and 100 years ago or 150 years ago, that would’ve been this massive luxury.
Michael Belasco (14:35):
Yep. No, they’re all-
Spencer Burton (14:37):
That’s physical AI.
Michael Belasco (14:38):
I mean, well, if you pay attention to the space you see that they’re focused on it. Let’s just hope we don’t have the day where-
Spencer Burton (14:44):
They take over the world?
Michael Belasco (14:44):
No. When does Arnold Schwarzenegger comes back in-
Spencer Burton (14:47):
Terminator?
Michael Belasco (14:48):
Yeah. What’s his name? I forget the Terminator’s name.
Spencer Burton (14:50):
Oh, I don’t know.
Michael Belasco (14:52):
Anyway.
Spencer Burton (14:52):
But well, let’s wrap it up there.
Michael Belasco (14:55):
Yeah, sounds good.
Spencer Burton (14:56):
Physical AI, we didn’t want to have a season where we’re talking in-depth on AI without talking physical AI, it’s absolutely an impact to real estate. With that, let’s move to our next episode.
Michael Belasco (15:07):
All right.
Announcer (15:09):
Thanks for tuning into this episode of the Adventures in CRE Audio Series. For show notes and additional resources, head over to www.adventuresinCRE.com/audioseries.
