Latest / Construction Technology Podcast / Construction Technology Podcast - Episode 017 - MaintainX, Palantir & McCarthy and AI Predictions
Transcript
- Luigi La Corte: Hello everyone, welcome to Construction Technology Podcast. ⁓ I'm Luigi, CEO of Provision. I'm here with Andrew from Join and Mike from Build Vision. gentlemen, how's it going today? Andrew, how are you? So good. Mike, all's good? Baby's good? Andrew Zukoski: Good man, how are you? Mike Powers: All is good. Baby's good sleeping like a champ. ⁓ ten thirty to two thirty. Yeah, night nice is working out great. Luigi La Corte: Amazing. The the nightmares is working up. Yeah, that's a high ROI decision. I would recommend. Mike Powers: I know it's expensive and people it but like it's all it's a one time expense if you amortize it over the life cycle of your sleep schedule it's actually very it's very worth it. Because the the the outcome, the success outcome of a night nurse is that they are sleep trained to sleep twelve hours ⁓ the night nurse sleeps. Andrew Zukoski: ⁓ 'cause the night nurse is a professional who has the incentive to sleep themselves. Like Mike Powers: Correct. Correct. Correct. Correct. So if you think about that, then people like, ⁓ you look good over a three year period, that three year period, a lot of families I know, even at three years the child is not sleeping through the night. So what is the cost for sleep? The cost for sleep is very, very ROI positive. Luigi La Corte: God's forsake. Yeah, that's on your investor update. Your ⁓ your sleep based. amazing. There's a lot to talk about today. a few things this is the first like non-guest podcast we've done in a little bit. but there's a handful of things we want to go through in agenda. the first thing is maintain X purchased by Autodesk for 3.6 billion. Mike, I don't know if you have a LinkedIn post you want to pull up on screen. I'm happy to look for one in the background if you don't. but gut reactions. I mean quite honestly, maintain wasn't on my my radar. I was like slightly aware of them. I've seen their brand here and there. But a lot of the discourse, a lot of the discourse on LinkedIn was Autodesk moving away and diversifying from design to go full lifecycle. They've already been involved in some elements of operations, but this is a size bet. Last time I checked their market cap, Autodesk is around 60 billion. So to bet, you know, 5% of market cap. That's a ⁓ it's actually quite a sizable bet. Mike Powers: So I think I I think so since I Autodesk in twenty eighteen, twenty nineteen, if they have building building connect acquisition, that was completion of the construction stack of Autodesk. They had they've rebranded to make design build. This is the third pillar of that. So they have design, they have construction, and then they have operate, which completes the full life cycle of a building. So wise, it seems heavy. What was it? Twenty what was it, twenty six X on Rev? ⁓ Luigi La Corte: twenty six is some forward ARR. Mike Powers: They're growing fifty percent year over year. So I don't I shared with you guys a little table in our chat, but you know, for assuming things go well and that growth continues to ex accelerate, but you fifty percent this year, fifty percent next year, forty percent in twenty-eight, we got thirty-five percent in twenty-nine. They're sitting three hundred and eighty million dollars of annual revenue. that's only a nine X multiple on enterprise value over which not That far for a company that's growing that quickly with the strategic initiative of them being able do that. the CFO talked a lot about one, they're growing quickly, at maintained X so it makes sense. Two, ⁓ this they're gonna be able to resell into existing customers. I don't know how much overlap their existing cover customers are or ⁓ what they look like. ⁓ I sold in the ADC space and I don't like I've heard rumors of like a JL. ⁓ Like a some of like the distribution prologus companies. ⁓ so there's a huge opportunity to do that. Cash flow-wise for Autodesk, they're like throwing off like I think eight hundred million dollars a quarter in free cash flow. So this pays for like one thing that was interesting was it was two, I think it was two billion dollars cash, one six in debt, or the other way around one six ⁓ so yeah, so like their per they could have done a hundred percent debt. Andrew Zukoski: Yeah. Luigi La Corte: Yeah, two billion debt is what it says here. Mike Powers: But their perspective is like, hey, it's gonna pay for itself a two quarter like that that one six will pay for itself in two quarters of cash flow. And they think that the implied earnings will outpace what they were getting on T bills right now, where their cash was sitting. ⁓ so I think it's very, very interesting. ⁓ again, the bet being that, hey, we now have a full life cycle of an issue tracking process. again, it's strategic and I think it could work out, assuming things go well. It it it closes the loop for design make operate. ⁓ Andrew Zukoski: Yeah. Luigi La Corte: But where's the old like where's the synergy with their existing customer base? Right? The they're not they're not selling this to construction companies, they're selling this to owners. Andrew Zukoski: They're all buildings. No, but like consumes you consume you consume the artifact generated. Like they started with design, which is five percent, ten percent of construction cost. They move to construction. Operation costs of a building are 10x over the lifespan of the building of the first construction cost, right? Like and it consumes all the artifacts that come out of that construction process. ⁓ so I I it might not be the same. Mike Powers: So I ⁓ Yeah. Andrew Zukoski: It might not be the same customer base, but Autodesk outside of Revit didn't have a large construction customer base when they set out to build ACC, but they said we have a Fload of look at me, I censored myself, Fload of ⁓ cash and we can plow it in and we can build the distribution over time. so I don't I don't I need it. Like they can go build that. Mike Powers: And I think it's also a first-party data stream to like actual real-world asset performance, which historically isn't anywhere else. So obviously you're designing a building. Hey, what is the design quality? What is the actual output quality of this piece of equipment that I'm putting into a job site? ⁓ and this tells them. ⁓ and again, if you're creating motes around proprietary data, ⁓ this hardens the moat around proprietary data and workflows that Autodesk will have. Andrew Zukoski: It makes me wonder what happens to the media and entertainment sides of Autodesk. at the New Orleans, whenever they did Autodesk University in New Orleans, like the most hilarious talk was when they combined the construction and media and entertainment keynotes. And so people showed up and so you had, you know, superintendents and media and entertainment people like sitting side by side trying to consume like some coherent message and to add Mike Powers: Mm-hmm. Andrew Zukoski: Like, you know, the guy who maintains the asset, which is this like totally different audience. it's interesting. Like Mike Powers: Yeah. Luigi La Corte: I'm sure they're gonna break it out. It doesn't I don't know if it makes sense to even combine them anymore. It's such a large target audience in construction now. And I don't know, I wonder if any of you have the numbers, but what is the construction business growing at relative to the remainder of the business? Andrew Zukoski: It's most pro. Mike Powers: They don't extend they they don't break it out though. They it's AEC. It's all under AEC. I don't think they break it out, right? Luigi La Corte: Sorry, I mean the ABC business. Mike Powers: ⁓ let's see what is ⁓ ⁓ you can ⁓ Andrew Zukoski: Well what was like Plan Grid okay, so seven years ago PlanGrid Building Connected and those are the only two real revenue sources were at like a hundred in total. Mike Powers: Yeah. Plan grid was at ⁓ eight eighty five to a hundred, building connector was at thirteen to twenty, ⁓ pipe was ⁓ sub three, so they bought a hundred and they they bought the sa which was interesting, they bought the same ish amount of revenue ⁓ for three times as much with maintainers. Luigi La Corte: It looks like the AEC business AEC O is about fifty three percent of revenue. Andrew Zukoski: ⁓ information. Yeah, I would and I would I would bet that's growing faster than the other Luigi La Corte: AutoCAD then I yeah, I'm not sure if that I think these are ⁓ these might be estimates, but AutoCAD is twenty seven percent. Andrew Zukoski: I think AutoCAD is a growth market. Actually no, that's wrong. They did they did a bunch of like licensing cleanup ⁓ and that actually brought a bunch of extra cash in. Like that was probably needed around the world, but it's giving giving them the cash to make these investments, it's it's impressive. Mike Powers: Yes. Yeah. Yeah. Luigi La Corte: I I'll catch my easy one. Yeah, it is. Mike Powers: There's an interesting data point where Andrew Bay, the CEO of Autodesk, back when we were ⁓ acquired that like ⁓ they had a billion dollars of revenue sitting in non compliant software seats. ⁓ and that's external of any of the companies they can't sell ⁓ countries they can't sell into, which is like Revit or Russia, China and Iran. it's pretty wild. Like the like the footprint that they have is is massive. Luigi La Corte: Yeah. And I didn't appreciate this, but AutoCAD is not even included. Well, it kinda makes sense, but it's not included in AEC, even though it has AEC applications. So between AECO and AutoCAD, you're looking at around eighty percent of revenue. Media entertainment's about five. I'm surprised they never really made ⁓ I mean I don't know what's happening behind the scenes, but I'm surprised they they have not acquired a ⁓ reality capture company like ⁓ like open space, yes, but there were a few that were for sale. one was bought by Drone Deploy, I'm to remember the name of it. Struction site. Mike Powers: Open space. Andrew Zukoski: I don't know how complete it is relative to like a struction site or an open space, but Trimble just launched that a few days ago. ⁓ Mike Powers: ⁓ structure site. Luigi La Corte: Hmm. It's a pretty magical product from what people describe. I mean specifically for ⁓ for open space and it does obviously dovetail into ⁓ digital twins. Like that data is very valuable and ⁓ only a only a couple of companies have it. Mike Powers: So I I I sol I sold ⁓ data capture for a year between my auto test in sin starting a company, a company called OnSite IQ. ⁓ the challenge is you're disincentivized as a superintendent to capture the entire project. ⁓ and that was open ⁓ on site IQ's ⁓ position was that they would send somebody out as like an Uber, almost a walk the job site to make sure everything was captured on a consistent basis. interesting data point like Andrew Zukoski: Yeah. Mike Powers: The data's valuable, but it's might not be structured in the same way because of all the different pieces that you have to like come because it's a real world asset. Like making sure the person walks the same exact path every time so you actually can see progression. Then lining up that progression with what's actually on the project. ⁓ and it's also reactive. Like if you misplace a pipe ⁓ in a building, like you're still having to do the rework to take the pipe down. ⁓ you can like semi-catch it if you have like a stub up that's six inches off before you run it across the floor, but it's still like not as proactive as you need to be. Andrew Zukoski: It's interesting. The I sat down with a product leader at Procore. This was like eight years ago, six years ago, four years ago. I don't know, this is while back, and I was like, ⁓ why isn't why isn't Procore doing anything here? And I don't know this is right, and this was a long time ago, but the reaction was like, it's just a lot of photos. We already have a lot of photos running around in our system. Like, we don't we we we're good on photos, basically. I thought that was interesting because it does like on the one hand, like, okay, that's fair, and they're attached to a bunch of other structured information about what's happening. On other hand, boy, this really s does seem magical ⁓ and the only like source of truth for a construction project is the actual site. Like instead of deriving information from people with digital clipboards, you could derive information from like the physical reality and have a much stronger for like whatever insight that you that you pull out. ⁓ Luigi La Corte: Right. And like when I speak to general contractors about it. the like the the killer use case for them is dispute resolution. Anytime that there's an issue with a subcontractor, it's trivial to trace back the steps, back the layers of the yanion and say, at this point your work was done, clearly was available. At this date, you didn't start work until this date. ⁓ It's going through a time machine. And ⁓ end dealing with a lot of these subcontractor issues. So you know it streamlines the whole thing. So it yeah it's so it seems so complimentary to the whole process. I'm so I'm surprised it hasn't happened because neither ProCore Andrew Zukoski: Yeah. Mike Powers: Yeah. Luigi La Corte: nor Trimble nor Autos I purchased someone here and there have been a few companies for sale like struction site was one of them ⁓ then Avenir was one that I think was sold and there's still Mike Powers: Or out of that. Andrew Zukoski: Nimmer check but the one out of Illinois. ⁓ yeah. Mike Powers: And then there is built dots. there's a couple more still out floating around. Luigi La Corte: I think build dots is is gonna crush. was talking to head of innovation at a Canadian general contractor, and they they told the use case, and it was it was pretty brilliant. So builds will keep track. I don't know how the solution works incredibly well, but it'll keep track of production rates and it takes customer ⁓ aggregates it, de-anonym or It and will tell you, hey, your concrete pouring rates are 3x less than industry standard. Something's wrong. So there's like this automatic triage system that allows them or general contractors to go in and like understand the root cause of why they have schedule issues. It's it's really brilliant. it's kind of like schedule intelligence, from what I understand, but it it uses very a variety of sources of information. The data mode seems so straightforward when you have all that. Mike Powers: Yeah. Yeah. Luigi La Corte: It seems like you need to buy it. Andrew Zukoski: I don't know. So the like you get the the challenge I've heard from the reality capture companies that I've talked to is that it ends up being commoditized. That at the project like yeah, we're gonna have reality capture on every project and we're gonna get the cheap one, whoever's cheaper, like the month that it goes out. Mike Powers: Yeah. Andrew Zukoski: And so you you don't you don't ag like you ha you you have to continuously refill the pool of activity to keep the data going. It's not like you sign you don't s the ones that I've spoken to, build outs may be very different, like it's hard to build enterprise relationships, like because it turns out it's really hard to derive anything that's sort of enterprise value from a of messy reality capture data across projects, but I don't know. Luigi La Corte: But how do you how do you that between like any enterprise solution, even like like a project management system? Can't you use the same logic there? Andrew Zukoski: as CFO want clean financials or how much cash is coming in and going out across my entire enterprise, right? And PMI project management systems in part feed that, right? I might also have a policy that says we're gonna do reality capture on every project because it's better for the project, it's better for our outcome. But I just want to know that you're doing it. I don't I'm never gonna look at anything that is derived from like that data. Like whereas the project management system Mike Powers: Mm-hmm. Andrew Zukoski: like there's some sell in the thing that I might potentially go out to the public markets with that's derived from like that trace all the way back to use the system on the projects. Like that's and that that maybe that's changing and I'm I assume these companies are innovating and finding ways around this, but Luigi La Corte: Yeah. curious. I I I would I would bet that they would make an acquisition here, but actually just reminds me something. Mike, didn't you make a prediction that Autodesk would be announcing an acquisition where they'd have to make a public disclosure? I just don't remember if you said it around it as AI or operations. Mike Powers: Can't confirm or denied. Any any any for any for any forward looking Andrew Zukoski: It's a year that starts in two. It's a year whose number starts in two, so Autodesk will announce an acquisition like Luigi La Corte: Yeah. you need to update your LinkedIn to AEC Oracle. I'm pretty you made that. Andrew Zukoski: Yeah. Mike Powers: Base your goal. Andrew Zukoski: Yeah, okay, here, look this. Trimble just announced days ago, three days ago, Project three sixty capture, AI powered capture tool now available. Did they build this? Or did they someone and not announce it? Mike Powers: Trimal bought somebody. No? Andrew Zukoski: I don't remember, but maybe. It's cool. Luigi La Corte: I think the other thing that would be interesting to talk about, I see it being related to construction eventually. the budgetary that large technologies are facing with ⁓ AI spend. So you guys Uber, and I think there was other company announced a cap on generative AI token spend on a monthly basis. You know, people were pretty alarmed at Uber. I think they went through their annual budget in like one month. The Facebook incentive machine around token maxing had a single developer burn like billions of tokens. and I think you're seeing a reaction to like the questionable ROI. Where do you think how do think this is gonna impact construction companies? and you can look at it two ways. One, like the general contractor, where they're all exploring, hey, like we can now build a lot of things. And then two, the technology companies like Autodesk and Procore. Mike Powers: So I know McCarthy made another announcement yesterday, but during that announcement, the that they launched, there one the pieces that they said was it helped them reduce token spend by 80% by leveraging Pla Palatyr as their platform instead of going to Cloud or OpenAI because of how they orchestrate and manage their chain of thought. So ⁓ my position it's gonna be a continual race to zero. And I would short but I I would short the upside of the foundational models because like open source will take over. Luigi La Corte: Open source will take over. You know what? We use a lot of open source now, and that is true. Mike Powers: Yeah. There's a lot there's there's so you're you're sitting here I'm a contract, I'm a general contractor. say I've say I'm a mid mid tier general contractor and I had to spend three hundred thousand dollars a year on Procore. I have a hundred employees, am I gonna spend another three hundred thousand dollars a year on Claude or three hundred thousand dollars or or thirty thousand dollars a year on an open source model that's ninety percent there? Andrew Zukoski: Wait, but you're not there's not ten X between two. ⁓ Like was the smaller models from Claude, like let's just like but you still have to have GPUs and electricity, which is the pr ⁓ Mike Powers: I'm I'm saying going like op straight open source, like The whether they're ten percent the cost. I so even if you Andrew Zukoski: No no no, I think there are more. I think that the r I think even with the pricing token, I bet and it's gonna be very interesting when claw when Anthropic's S1 numbers come out in coming weeks. Like I think maybe you're converging to cost with the token prices and the pricing models like now pricing to cost. I think you're Mike Powers: Yeah. But they've been they've been dra so I I think we'll get past this point that everyone w constantly wants to use the leading edge model. Because we're already looking at this internally where and we've talked so we use we use Gemini Flash for ⁓ Gemini for a lot of our processing and we've talked to the G C P team and the G C P team's like or the Gemini team's like, Yeah, you guys can go buy a million dollars worth of Gemini pro tokens because we're gonna have Gemini four. ⁓ Luigi La Corte: ⁓ right. Yeah, I see. Mike Powers: In the next six months. And if Gemini 3 does exactly what we need to do it benchmark wise, we're gonna be like a 50 to 75% discount on the price of Gemini, which is already half the price of Claude. So that whole conversation and workflow of like it if it's good enough already, I don't need a foundational model piece. Maybe people that are on the leading edge of need it. But construction is not on the leading edge of like drug disc I don't I don't even know what the leading edge is anymore of like who actually needs a full foundational model because like I my Uber app has not gotten ten times better because they're using Claude four eight, Opus four eight. Luigi La Corte: Yes. That's Uber. Andrew Zukoski: The our the argument there is most people need something beyond the current foundational models to actually deliver ROI. Mike Powers: What's a but what's a 10x better I d I I don't think there's 10x val like what is better what like making you connect better with a driver? Like there are there the the ⁓ so the large the large contracts that Anthropic has, what is an actual ten X ROI for them for these costs? I don't think it's there for any of these customers. Andrew Zukoski: Ship ship the app faster. I don't know. More ads. Luigi La Corte: Well, where I'll disagree with you is Uber is a technology company and are a very mature product. Construction companies are not technology companies. And if you look at the output, like the code that they that they will build invariably, it's not like it's it's not at Uber quality, right? So I think the bar is lower for Mike Powers: Doesn't need to be anywhere quality though. And I Luigi La Corte: Well, it doesn't need to be Uber quality, but now they have the ability to build themselves. So okay, I'll disagree with you in two more places. the harness is still very valuable, irrespective the model. And the Claude harness and the open AI harness, like Codex, for example, I don't need to go mess around with open source. Like for for us, okay, like ⁓ for very specific process, we're switching between Gemini 3.1 Flash or Pro to ⁓ Deep Seek ho through open router. Okay, ⁓ it's an eight eight X cost reduction. But don't have ⁓ I'm not fitting that into a harness. I'm fitting that into a very specific pipeline. And that's easy for me. That's hard to do for a construction company because they're they're probably dealing with the harness and they're probably dealing with the agent, not Mike Powers: But I don't think I don't think they care. I don't think anyone's gonna care if you're using Opus, if you're using Andrew AI. Like you as a construction user, if you're using Provision, you're gonna use you're gonna use whatever Provision offers you. And if Provision's like, Hey, we're gonna reduce the cost by ten X because we're gonna use we're gonna spit up an open source model that we're running in a GPU server in Toronto. So that reduces your token you're not you're you as a technology provider is not gonna say I'm not paying ten dollars for clot claw tokens, I'm paying one dollar for Andrew Zukoski: It's not You spin up a server because you still you still have the infrastructure cost. Mike Powers: Why not? But But but that they ⁓ the whole like those like the the infrastructure cost is not what's driving the top end cost. Like those are the two inflection points that are happening. You're gonna see reduction in infrastructure cost, and then you're gonna see increase of capabilities of of the tools to where is the point where where does those lines cross where it makes sense for the functionality that you're trying to do? And it hits and it ev and everything continues to d it's this is Moore's Law. Andrew Zukoski: Yeah, there's a lot of electricity that seems to go into these models. Like even even the smaller ones, even the open source ones. Like I guess it's Luigi La Corte: So is are you Mike Powers: I think I think I think but the math on that is we're running towards edge compute where you're gonna be able to run it on your phone. You're gonna be able to run the quality of models that we have today on your phone in the next twelve to eighteen months. So then your electricity cost is zero, it doesn't matter. You're charging your phone a knife. That's a technology bet you're bet eat making. Luigi La Corte: So so so you think these comp you think construction companies will implement budgets? Like that's what your that's what your bet is? Mike Powers: I don't think they're gonna even I don't think they're gonna notice. They're gonna use Andrew Zukoski: I mean a corollar a corollary of the so like none of the none of the cloud providers agree with you because they're they're building out huge data centers that ⁓ are gonna be amortized over five to seven years and you're saying eighteen months from now like we're not running new Mike Powers: They have to. They have to. That's Apple's bad. Luigi La Corte: Well, that's everyone's bag. Google Google just raised equity. Just sold some equity for this. Andrew Zukoski: Well they license it. They have big contracts. Mike Powers: If I'm if I I my d think the incentives ⁓ very misaligned for hyperscalers because hyperscalers have now invested in all these foundation models with idea that they're the like that's the the three legged stool of Nvidia chips to hyperscaler back to foundation model. Nvidia invests in open AI, open AI pays Amazon. Andrew Zukoski: There's like multiple providers now. Like Goo like what fraction of Google's what what fraction of Gemini's work runs on TV? Mike Powers: Doesn't bet it doesn't move them it doesn't matter. They're they they have the best they have the best business model in the world. Google can Google can Google's one raising eighty billion dollars of debt right now. Luigi La Corte: No, equity. Equity. Andrew Zukoski: No, th no, that was equity, but they did a big debt raise. They're they're raising a fuckload of capital. It's very, very interesting. Mike Powers: They they're doing a b okay, they're doing a big debt raise. While open AI and Thropic are also gonna raise hundreds of billions of dollars in the public markets, while SpaceX is gonna go up. Andrew Zukoski: Yeah, so the price is gonna go way up for all the foundation models because like that has to be paid back. Did you see where ⁓ Anthropic was like, guys, we gotta pause like leading edge model development. We gotta pause. Like Mike Powers: Yeah, because they have the Sam Sam did the same thing during the Biden administration when he was trying to say, like, ⁓ we have to go through all these regulations. We're gonna pull up the ladder for everybody else. Because they know, I would bet internally they know that the open models are gonna get there faster. They don't need all the capital expenditures that they need to actually execute on what we need to see. And they're seeing they're seeing a line in the sand. There's they're saying like what what if we we if we spent a trillion dollars in capital expenditures in the next three years? And don't return ten trillion dollars of value? This is a miss. Ten trillion dollars of value is a shit ton of value. Andrew Zukoski: Yeah. Luigi La Corte: Let me give the opposite perspective here, because I think I think what we're potentially forgetting is that both Anthropic and OpenAI signed these gargantron agreement with private equity firms for AI implementation. And as we know, enterprise implementations are incredibly sticky. So they are effectively replicating some component of the Palantir model where we have these foundation models. Now we're just gonna go implement everything in enterprise. Anthropic is a little bit further ahead, but OpenAI will do the same thing. And then I think you won't even be thinking about token spend. Like maybe you will a bit, you'll have a budget, but those are easily set up. The harnesses. though will do all the work. And I really have a hard time believing that a CIO is gonna say, you know what, let's go open source this and we'll figure out everything else. Because the last mile is a lot of work. The last mile being the harness, how it integrates, the process. ⁓ Andrew Zukoski: But that's that's that value is not gonna go to the that value will go to the whoever writes the harness. That'll go to the firm if they do it internally or to software vendors, done externally. Like but the Yeah. Luigi La Corte: Yes, they could. Mike Powers: I think the challenge the challenge with it is if you're not seeing a ROI, what you what you're seeing frac the the the initial point of this was like you're starting to see fractures with large enterprises like Uber coming out and saying, hey, we're not seeing the ROI on the best theoretical ROI return, which is coding at scale. I don't think it's I for them it doesn't matter. I don't I I this goes back to the point they don't need to write we're not measuring like we don't need they don't need to write more code. There's nothing to change. Like there like there's Luigi La Corte: What do you guys think about that though? Like you guys Andrew Zukoski: Yeah yeah. Luigi La Corte: Well then they should fire their eng they should fire their engineers. Like Mike Powers: But they're having to make 'cause they're having maintain it. They're maintaining code bases. And and and you can't say and you can't say c and you can't say Anthropic is the best maintainer of code when their website has nine is ninety percent up ninety eight percent uptime. GitHub's is below ninety percent now. Luigi La Corte: Yeah, but it doesn't seem good enough to do it at all. Andrew Zukoski: Yeah, I mean Luigi La Corte: That's their that's their Yeah, that's not fair. Andrew Zukoski: I don't yeah, that's not I don't know that that's a fair comparison. Like that's a service That like is very difficult to scale, that involves like big buildings and construction and all these long lead time engineered systems that you gotta configure and price and order, right, in order to deliver that uptime. Same for GitHub, like all the AI generated pull requests led to like it was a business that was priced steadily growing ten to twenty percent a year, and they knew how to grow it and scale at that, and then all the AI generated code started coming in, and just like every system is now trying to handle two to four times the volume essentially overnight, which in a scaled system is Luigi La Corte: Yes. I Yeah, and probably went from f like three billion to like fifty billion in revenue and they had to deliver on that revenue and like ninety percent or eighty seven percent of time is like pretty decent in that regard when you have a lot of hardware. Andrew Zukoski: I mean no no nines is an interesting like reliability standard. Zero nines. Yeah. Luigi La Corte: No nine is the same strategy. Yeah. Mike Powers: They're not no one's talking about H.I. anymore. Luigi La Corte: I don't know, I still see a lot of AGI in my timeline. There's always questions. The the the post the posts do keep moving. that's fine. Okay, let's talk let's talk Palantir because it's very similar in the same conversation. You know, I watched the Okay. Andrew Zukoski: I don't think it is, but go ahead. Mike Powers: Wait, we also point out that Join got a shout out on the Palantir announcement. Look, here, huh? I think. Is McCarthy a Join customer? Right here, look. Andrew Zukoski: Are you fucking kidding? Luigi La Corte: Wait, what? Andrew Zukoski: Yeah, yeah, yeah, they're like one of our oldest impingers. Holy shit. Luigi La Corte: Procurement join. ⁓ my God, you're hearing it you're hearing it here first. Andrew Zukoski: Are you sure that's not I think that do you don't you think that isn't join is in like a sequel join? I mean yeah, it's definitely it's definitely it's definitely true. Luigi La Corte: ⁓ yeah, I think it I think it might I think it might be a C. Mike Powers: No, so all of these are all of the ⁓ I don't know. I I saw that this morning. So that what what this is showing is so McParthy announced the partnership with Palantir. ⁓ there's been a lot of back and forth of like they spent a lot of money to do this, ten million dollars plus a year. What is the ROI? ⁓ the most interesting clip from this video, they did a demo for ⁓ I think it was like four to five minutes at Palantir's A AIP conference or AIP, whatever they call it, conference. ⁓ what's interesting about this was just this like chain of reasoning and chain of thought. and what this is doing is this example that they have is there's an RFI for an AHU model mismatch. ⁓ and it goes through and pulls ⁓ like all the meeting notes from the OEC meetings, the data logs, the drawings, ⁓ then ⁓ then it like reviews the schedule impact, procures or reviews procurement verification. And then McCarthy has all these SOPs, like a database SOP, so it pulls from these SOPs. ⁓ and it actually gives It gives like recommendations on what they can do. So it can they can optimize, they can whatever domain means, they can own it. Like and the output of this is instantaneous scenario planning. So, hey, do we pull a substitute from the supply chain or hold on to the architectures variant? So, this example, they had another air handling unit going to a different project that they could swap out here, which is that's a stretch. But ⁓ Andrew Zukoski: Yeah. Mike Powers: And or they hold to the current process and what the impact is around that. ⁓ Luigi La Corte: Sorry, can you dumb can you dumb down the scenario again? Like what is the scenario? Mike Powers: So this scenario is they the engineering firm which miss spec'd the chiller. So or sorry, the air conditioning unit. ⁓ they spec'd like a train seven fifty and it was supposed to be a train one thousand. ⁓ it wasn't caught. ⁓ the chiller or the air conditioning unit had been released. ⁓ so now they're submitting an R5D engineering firm saying, Hey, ⁓ We order the wrong ch we order the wrong air handling unit because of X, Y, and Z. So it goes through the process of doing an RFI to an addendum or a bulletin update. But because the piece of equipment was already released, they did an analysis on like what would the change cost be if they ship that unit to the job site to have to change the curb size or something. Or do they pull another air handling unit from a different project that fits that's the pad that's already laid down because of all the different pieces? So there's like 9,000 like nodes that have to connect there. ⁓ Because Palantir has an ontology and can connect all these data sources, that's why I thought it said join, because it would connect theoretically join. It's able to give you a scenario of like pull the substitute from your supply chain, so swap out this sh this this Voyager unit ⁓ or hold the architect's variance and do the rework to expand the concrete pad ⁓ to fit the unit. Luigi La Corte: Okay, so the use case here is like I'm a PM, I'm going over my schedule, I just got noticed that the this has been misspec'd, we maybe ordered it, and now I need to correct schedule issue. Mike Powers: Yeah, so it's mechanical subcontractor says, Hey, the pad ⁓ we just got the shop drawings for the air handling unit and it's ten feet wide. The pad that's already poured is eight feet wide. What do we do? ⁓ so that's where this workflow is kind of if you go back here, like this is this this chain of thought. So it goes through, ⁓ does all these things, which I think is what I think is interesting. I the the suggestions are whatever. ⁓ what I think is interesting is it automatically ⁓ of course it went to an ad. ⁓ it automatically. ⁓ I know. Luigi La Corte: Hey, we're not getting money from these people. We can't show their eyes here. Andrew Zukoski: ⁓ Jenny cast a hone cut. Mike Powers: ⁓ what is interesting about this is it it automatically c drafts all the emails that send it all to people as well. so it sends emails to everybody notifying them, connecting them back to the source of truth. if the subcontractor is in build ops and McCarthy's in Procore, those nodes connect. I know I think it's an interesting ⁓ use case of an issue through the full life cycle. Which historically ProCorn Autodesk have failed at doing. Luigi La Corte: Okay, but are you blown away by that example? Because it like based on the recommendations, it looks okay. Andrew Zukoski: I think there's a I think there's a I don't think AI has anything to do with this and I think there is a ton of value in it. And I think there's a key thing in assumption that the PM is the user. if you at what like Mike's Mike's like polstering go Turner money on ⁓ zero percent CM fee. If you look at like one of the things we consistently see is large general contractors getting much more sophisticated about their businesses where they have many more potential revenue lines and then those revenue lines are much better integrated and controlled across a national portfolio. And so I'm not I didn't actually watch the live stream. This looks to me like an an actual ERP system, right? Construction companies, when they talk about ERP it's actually accounting. That's that's all that gets done. This is an actual ERP implementation across like nationally centralized equipment yards, nationally centralized procurement, and all the project teams. And so the I don't I'm not impressed by the AI piece either, but the fact that like someone in Louis in McCarthy's case know could could actually know every air handler that's planning to be purchased on every project across the country, that's a big fucking deal. And new, right? Luigi La Corte: Yeah, that is very cool. Yeah, that is. So so who who who feels feels the pressure here? Is this like Jonas? ⁓ this Procor? 'cause I gotta imagine I gotta imagine it's kinda both of them, but I I don't just don't know enough about the Palantir solution to know who feels threatened. Mike Powers: Yeah. Andrew Zukoski: If you look at like I don't know that that's true. Like if you look at these companies, ⁓ d do they if you fast forward a few years, do they look like construction companies of ten years ago where they have half a percent spend on technology, or do they look more like diversified manufacturing companies where margins are definitely going up? Like There are more sophisticated organizations running more like from a central like a central office and with like less default out to projects. Like I don't it it might just be incremental spend. Mike Powers: Yeah. If they can push it back to the owners. Luigi La Corte: At least for the labels, yeah. Andrew Zukoski: Yeah. Well, ⁓ I mean the other the other way to look at it is your like one of the consistent themes in is leveraging aggregated demand for goods and services. So the other place we get pushback is to OEMs of and materials, right? Like if you're being smarter about purchasing across ⁓ multi billion dollar volume as to purchasing project project, you're probably getting a better deal. So maybe it's the OEMs, maybe it's the maybe it's the downstream providers get, you know, beat up. Mike Powers: It was a if you l listen to the old podcast, it was my bet on what if Turner signed a palantier agreement, the same thing. It's like how do I get insight into what equipment rentals are needed on a job site and how do I leverage first equipment to facilitate that and move nodes around? Andrew Zukoski: Yeah. Or even like what's gonna be needed a year from now or two years from now so that I can make good decisions about like the volume we're gonna have purchased versus the rental agreement volume that we're gonna have to cover it. Yeah. Mike Powers: It's actually creating a true supply chain. Correct. And I think the friction point that's gonna happen is if it's a ten million dollar a year cost for a Palantir subscription top of what you're spending at ProCore and Autodesk, they're probably doubling, maybe even tripling their tech spend because of this. Is there ROI in that? Is sav i the them think they said they said again, this is like this going back to this principle, but is rework actually a trade a traded back to the GEC or the owner? So that that that example, they saved. somebody $160,000 by not reworking the concrete pad for the air handling unit. That $160,000 does not go into McCarthy's pocket ever. That unless it's a lump sum project, but they're not doing lump sum projects. So that $160,000 is now going back to the owner. So when you're evaluating the $10 million of spend, do you stack up to enough of a, you know, we always say $0 ROI to $50 million in savings a year on rework. ish, which is fuzzy. It's tough. It's a tough so that but that and that's going back to this initial point around token spend. Andrew Zukoski: I don't I don't I think that's a I think that's a di I think that's a digestible for an for a marketing presentation for their vendor. I don't think that's what's I don't think it's rework. Mike Powers: Correct. Right. So then the other bet, then the other lens is we're having a hard time hiring people because people young people want to work on tech forward solutions. So now we're investing ten million dollars a year as a tool so that when we hire a mechanical engineer out of college, they wanna work in McCarthy because we have Palantir and the competitor down the street. There's something there too. Luigi La Corte: I don't know, well ⁓ maybe. I think that's a firm's benefit. Andrew Zukoski: Let's say you hire let's say you have to hire ten a thousand people a year. well that's ba that's easy then. Luigi La Corte: McCarthy, what's total top line? What's total top line at McCarthy? Five billion. Andrew Zukoski: Five. Rough. Maybe eight these days. Single single digit build. Mike Powers: There's more ⁓ yeah, s yeah, five or six nah. Well I gotta Luigi La Corte: Okay, got it. And so spending ten million a year on Palantir. Okay, that makes sense. It's probably comparable to what they're spending on Procore. You have to imagine that based on what Procore is offering, that Palantir is to go into that space because I mean Procore's a very impressive system, but you can rebuild these things ⁓ because they really are like database wrappers with like workflows and look I've spoken to people who have purchased Palantir and the way Palantir works is like you have like a you have a first initial period where you're literally building proofs of ⁓ proofs of value or or proof of concept and they ask you what do you want to build? You can tell them like I want to build an expense management solution. I want to build a PMIS and then they go build it with with Palantir. And then they come you know they aggregate those things and they offer those like car ⁓ think over time it I I do think it's going to be very competitive with with Procore and Autodesk. I think Autodesk is really well situated in that sense because you know they're going up and down the stack but I think Procore probably should be a little bit concerned. Andrew Zukoski: like what fraction of Procore's revenue comes from NR four hundred even? ⁓ okay, never mind then yeah. Mike Powers: Majority. They're not small business. They don't sell they they do sell small business, they've tried. Luigi La Corte: I thought that was I thought was like a a tongue in cheek question. Andrew Zukoski: I couldn't remember. I've only had one cup of coffee. Mike Powers: Yeah. They're w one point three seven five billion revenue. revenue is seven billion dollars because someone's pushback was like, ⁓ the yeah. Andrew Zukoski: That's it? ⁓ they're gonna struggle to to displace Procore. Because like all the tools we just looked at, you don't have to drive distribution across all the offices. ⁓ I need to plan, I need to use it. It's like an IT team needs it and you need integrations into the applications that people are already using, right? Like pro like the grind in construction technology is implementation. And like that's the that's the distribution that Procore has. Mike Powers: I so. Let me turn from Palantir Rev. Luigi La Corte: But isn't that then bullish for Palantir because they have a purely consulting model and they literally ship engineers on site to make it work? It seems like it's bullish, not bearish. Mike Powers: Yeah. Andrew Zukoski: Yes, I would be if if McCarthy is starting to do this, I would be bullish on SAP, I would be bullish on Oracle, I would be bullish on McCarthy's ability to expand into the technology needs of these increasingly sophisticated general contractor businesses. Luigi La Corte: When we say bullish on SAP, you think it's bearish for SAP, but bullish for Palantir? What? Andrew Zukoski: No no no, I I I would be bullish on SAP. I I I look I mean I could be wrong, but I look at this and I see like this looks like a big ERP. With a really good demo on top of it. Like that's not bad. Luigi La Corte: Yeah. It's date data un it seems like it's data unification and then whatever custom workflow you want. Andrew Zukoski: Yeah. And there's a ton of new workflows and a ton of new data as these businesses aggressively stand up like new business lines and then as they look to integrate those business lines nationally as opposed to letting it run at the regional or project level. Alright. Mike Powers: ⁓ so I think Luigi La Corte: Hmm. Okay, so predictions. Andrew and Mike, in three years, you think Palantir will take market share aware away from Procore or Autodesk or it be additive? What do you guys think? Mike. Mike Powers: I it's of construction growth. It will it will not beat ⁓ inflation. Growth. What I mean by that is I think the both revenues will grow because of of like Mar Hey, we're growing twenty percent year over year. I don't think you see a pro core in three years that's at one point three seven ⁓ billion in revenue or less and Palantir at construction absorbing the growth on that on top of that. Luigi La Corte: I think it'll trade off. I th I can see them becoming competitive. because once you unify all the data, I don't know, it just it it feels like it feels like ⁓ of their model where they are going to be taking individual workflows from various companies and just selling to the next company, like over time it feels like they'll just get there. Mike Powers: ⁓ yeah, yeah. So here's this is crazy. So autodest is a year ahead because they do the Salesforce numbers. So they're in fiscal year twenty they're already in fiscal year twenty-seven. ⁓ That's the that's what it looks like right now. Yeah, yeah. Luigi La Corte: You can use a minute. Can you zoom it? Just control and plus or scroll. Andrew Zukoski: That's all that Palantir does? ⁓ fuck, growing fifty percent year over year. That's nuts. No. Luigi La Corte: Is that is that construction revenue or is that full revenue? That's the price of earnings is ridiculous. Mike Powers: No, this is all palantary revenue. They're projecting ⁓ they're projecting seven billion dollars. Yeah. Yeah. Andrew Zukoski: They lent free in everyone's head. Luigi La Corte: This is ridiculous. ⁓ I know it's 10 o'clock and I think we soon. Andrew Zukoski: That's guessing. See man.