Michael Finocchiaro: Engineering Software Startups and the Money Behind Them

Michael "Fino" Finocchiaro unpacks a massive market map of engineering and manufacturing software startups, some of them real products. We debate where AI is truly moving the needle in CAD, PLM, and factory software and what an “OpenAI moment” in engineering could actually look like. 

• Building a 200-300 startup database from founder interviews and repeated data scans 
• Defining categories like design intelligence and cognitive thread to fit modern engineering software 
• Turning research into Threadmoat with interactive filters and advanced startup analytics 
• Using financial heat maps to reason about burn rate and capital efficiency 
• Mapping customers and investors to expose adoption patterns and hidden clusters 
• Tracking funding spikes since 2022 and asking what drives investment in engineering AI 
• Learning from “Prove It” model where vendors must run on virtual factories 
• Debating text-to-CAD versus AI as a new CAD engine that preserves parametric history 
• Arguing for leaner PLM plus a semantic layer and agent-ready integration over customization 
• Sketching a startup-first conference model with hyperscaler and VC sponsorship 

Transcript

Happy Hour Kickoff And Goal

Roopinder 0:08

Hello everyone and welcome to the Masters of Technology Happy Hour, where once a week I have a drink with someone I meet in the course of business, but someone I'd like to get to know better as a person.

Michael 0:18

Good to see you, man. It's been a while.

Roopinder 0:20

Good to see you. It has been a while. I think a lot has happened. It's like dog ears, isn't it? These days.

Michael 0:32

I

The 118-Page Market Map

Michael 0:33

wanted to show you one I've been working on in terms of analytics on top of the startup data I've gathered because uh I'm just trying to figure out how to do a press release. I've never done one before, but it would be good to make a splash. And I think you could also, you might be interested in the data itself. I have created a 118-page uh market map, including all the startups. So I've got six, I've actually got about 80 more queued up, but for Q1, I have 600 startups, 26 incumbents, 15.6 billion in VC. I've mapped 2,400 investors, 1,500 customers across nine categories, and a total val of 56 billion. And I'm I'm pretty sure I'm close. I've calculated three or four different ways and come up with between 53 and 56. Um, I've got all kinds of chapters on methodology and scope, the signals, you know, the big stuff that happened last year, the incumbent landscape, structural trends, um, customer adoption. I just realized it's missing part four. I have to tell Cloud Co to fix that. Customer adoption, consolidation, red ocean, blue ocean, vertical deep dives, pricing disruption, um you know, all kinds of stuff. About 81 charts, another 50 tables. I'm trying to kill SEM data. They don't know anything about all these startups, right? They certainly don't aren't tracking 600 of them, right?

Roopinder 1:53

They they are not, and neither is the industry media.

Michael 1:58

They are not paying any attention to startups and uh and so this is me saying I'm doing it and I've done the analysis and I've had this database, right?

Roopinder 2:07

Yeah, it looks like you're the main person doing it.

Michael 2:10

So how do I push it? How do I push it? So I've I just even mentioned Velotic, right? All kinds of charts. You've got all the lifecycle phase where they're sitting, the startups, you have their agentic stuff down here, in terms of the the also the AI surrogate ratings. I mean, I've got all kinds of data here, right? I mean, just tons and tons. So that's a report version.

Roopinder 2:30

So how did you how did you get that data?

How The Startup Data Gets Built

Roopinder 2:33

Is this just to specifically internet?

Michael 2:35

So no, not just the internet. I've talked to 150 founders roughly in the last couple of months. I've created this database. This is with a core of the database here. You can see all the columns I've got. Well, this is just some of the columns. It's about 60 of the columns. I've also had about 180 columns total. And I divided the market, removing hardware on purpose because I wanted to focus on software. I know hardware FII, who you've already talked to Connor and Benji, I think. They've got their own gig, right? And I'll let them do the hardware stuff. If they want to pay me, I can do this kind of analysis on hardware. In the meantime, since nobody else is doing it, I'll own the software piece. And I'll say that I've renamed the domains much better than some data, and I'm calling it design intelligence, extreme analysis, adaptive manufacturing, cognitive thread, factory futures, augmented operations, streamlined supply chain. Again, this is only supply chain when it's relating to sourcing, pricing, you know, stuff that affects building stuff. Once it gets goes to a warehouse, I don't care anymore. Once it's on a FedEx truck, I don't care. I just want to basically it's all the software to build stuff, right? Whether I'm managing knowledge, and then I found so many cool vendors up here that were just doing construction. I created a BIM category, and you can see I found 96 startups there. And then you can see over here the funding, and that's where I got that number I gave you of uh well, actually the total funding is looks like it's 26.

Roopinder 4:03

Well, you in each case did you guess at it or did you actually guess? No, no, no, no.

Michael 4:08

I I scanned and re-scanned and re-scanned and rescanned. I'm a little surprised. This number was 18. I don't know how it jumped to 26. I'm gonna have to look at the calculations here and make sure I didn't uh there wasn't an error anywhere. No, this is the current funding, and this is the best available. See, I I had four or five calculations of evaluation that I went f backwards and forwards to try to get the right number. Yeah, these guys are raising like crazy. They've already raised 335 million, which is just an insane amount of money. Yep. Yeah, of course.

Roopinder 4:43

Are you treating them as a startup?

Michael 4:45

Because they've been a lot of well so uh you'll see that I've I've I'm chasing everybody. As long as they're still raising capital, they haven't exited and they haven't IPO'd and they're not owned by PE. I just call them scale of expanders. You know, anything beyond series A, um, I call it an expander. Because they've proven their basically what they've proven their um their ICP, they've proven their product market fit, and now they're expanding and they're sort of in the end of between becoming an upper another open AI or being acquired by ANSIS Siemens PTC.

Roopinder 5:21

Did you see that they've probably are looking for a marketing person?

Michael 5:26

No, I did not see that.

Roopinder 5:28

Their marketing person just left.

Michael 5:30

Well, my um well, maybe Ron will join again because Ron's back in uh Budapest. Maybe Ron is.

Roopinder 5:35

Ron's back in Budapest, exactly.

Michael 5:37

Yeah, he's back, so maybe they'll hire Ron. Yeah, we'll actually schedule a talk to him later today. They wanted me, I did a little work for them to help them understand what PLM was, and then they wanted an article, and I just slapped one together super quick just to as uh with Ape and AI just to give them an idea what I wanted to say, and then they got all offended that I had used AI, like everybody else was using AI. I'm like, but that's not what I was gonna write. It was just a basis of discussion to see what you wanted me to write, and I don't know, whatever. People are weird, but yeah, so the numbers uh kind of speak for themselves. Although, again, I'm concerned with this 26 is I had a lower number than that in the report. But nonetheless, I've got all this metadata. And so what I did was I built Threadmoat.

unknown 6:21

Right.

Threadmoat Launch And Pricing Ideas

Michael 6:22

I'll show you the website. So that's what I want to announce eventually, is this website, Threadmoat, which you can put in night mode if you like. They're part of the Insight Partners group of Joshua Spreadberg, right? With NTOP and Colab and Monolith who exited and um SimScale. I mean, they're all the same group, right? And I'm I'm talking to Adam and I've talked to Brad and I know David, and yet ISFAN is still this is just the teaser, right? And then I I'm thinking of pricing it like this like give you three graphs for free. I'll give you my one report and like 10 graphs for uh five grand. So sort of like the pricing of uh close to the pricing of uh of sim data here. And then um for 19 grand, I'll give you uh quarterly reports and more charts, um, which I'll show you what the charts look like in a minute. Because and I chose 19 because it's sort of four times five, and but it's still less than the pitch book, right? Pitchbook's 35 grand, I think. And of course they'll brag that they have billions and billions of startups. I could say, well, that's fine, except I've got more metadata per startup than you, and I also understand what engineering and manufacturing are, and I can understand the market better. But I think starting out at 19, it would be nice if I could sell some 19 grand ones, right? It'd be nice if I sell lots of them, in fact. So let me show you the the graphs are pretty awesome.

Filters And Financial Heat Maps

Michael 7:46

And I'm working there's uh Cloud's code has been a pain in the ass last two weeks. It's been doing all these outages, but anyhow, so here's the um here's the that same map you saw before, but now where you can actually apply filters to it. So those are all the what I call an investment list.

Roopinder 8:05

Is it because it's uh getting so much use, the Claude Probably.

Michael 8:10

And I uh I downgraded my uh I didn't want to pay 200 bucks a month anymore. So you can like choose you know which uh which kinds of thing you want and the VC range and you know you can do all kinds of uh slicing and dicing and stuff. And here you can also just if you would rather the clustering be around the manufacturing type, you can do this, and then you've got discrete process construction and natural resources in here. You can do it in 3D, which looks really cool too. Um where the has my 3D button? You can do it in 3D and see the whole universe of startups that way. I've got advanced analytics so I can do a report on any startup, which is nice. I have a financial heat map. Uh uh so I've been looking at their deployment model, the funding, um, the headcount, and then based on the headcount, the ARR, um the AR efficiency, the capital efficiency, and then the burn rate based on ARR, but also on how intentions they are in the cloud. And then I did, like I said, three or four ways of calculating valuation to get really close, and then um some stuff on the confidence, how good on how confident I am on the score that I created over here. So that's pretty cool. Nobody else has anything like that, I don't think.

Roopinder 9:23

Um tell me about the heat. What does a high heat correspond to in that heat lab?

Michael 9:31

A high burn, it means that they're burning through cash. Oh, I see, I see running out of cash because they're spending a lot on the cloud, spending a lot on AI, high headcount, and they haven't raised money and they and they're not necessarily making a lot. You can see the multipliers are relatively high, but that's not all the multipliers. This is just the ones in cognitive thread. But if I go to adaptive manufacturing instead, that should change the list quite a bit. Oh, what is it doing? It should be ah, maybe it's because I want to do up to here clear. Let's see, is it gonna show me now everything? No, okay. Well, they may they might be a bug or two. Let's see, that's because design intelligence should always have in top at the top, right? Yeah, ah crap, it's not showing in top. I don't know why. I don't think the pivot to uh industrial co-pilot and to copilots was such a great idea.

Roopinder 10:26

Uh okay.

Michael 10:27

And then you have uh investor stats and all these other kinds of theatre, you know, here's investing things. If you want a hidden gym or pre-seed alpha, or you know, you want to look in the UK. So I've got all these cool things. Um, all the employee. So

Customers Investors And Geography

Michael 10:40

the cool uh stuff I have over here is I have these um that same map I showed you of the ecosystem, but I also did that with customers. So here are all the customers that have three startups or more. And here, okay, there's a problem with the data because there should be 1500, but still, that's um 209 customers that are all using at least uh one startup. There was 1500 before, maybe it was a lot of um redundant data. I'll have to check what happened, but that's anyway. You can see the the NASA and HP and Adidas and Kohler, all using lots of startups, and then I did this one for investors, so you can look and see all the ones that Y Combinator is uh supporting. And actually it should be 88, not 14, so I must have a bug. And I had 1400 investors, so yeah, there must be I I think I know where the problem could be. It still gives you an idea what I what I can what I've analyzed, uh, the tree maps, I've got timelines, and you know, I got basically everything, geographic maps, and then the metro areas, which I was surprised, you know, that okay. Now obviously the Bay Area is gonna be the biggest, but London comes in number two, which is sort of surprising.

Roopinder 11:48

But it's quite a surprise.

Michael 11:51

Yeah.

Roopinder 11:52

Which of the companies have you talked to that are are British?

Michael 11:56

Lots. Physics X is British, right? Rescale is is not is European. Neural is Swiss, so there's quite a few. Yeah, so basically, um I've got this. Oh, and of course you can do your own uh magic quadrant. You can do your own bubble charts, and there's a landscape diagram, so you can dive in to see what are the subcategories of design intelligence. Well, you've got Genitive Cat and Copilots and 3D concepting. So I've done all that work.

Roopinder 12:23

Back back to the uh the countries. Is China at all? Do you see China? I found one.

Michael 12:29

I found I found one. Actually, zero. Yeah, they they don't they're not transparent about it, right? They don't publish.

Roopinder 12:35

Right.

Michael 12:36

I mean, I know they have ZW3D and a couple other ones. Um, South Korea, I just have uh Clo. You know, those clothing guys are used by Rishmo and LVMH. Japan, I have Prism and one other one, just two. Not real big. India I had like nine, a couple in Australia, like XM Pro, a lot of things around mining, right? Um, and then one South African one, and one train training company in Nigeria, and then one French one over here, but that's that's kind of silly, right? 36, 58 in Germany. Pretty well distributed between the east, the west, and the south, in fact. And then 300 in in the US, of course. Yeah, so by far the biggest.

Roopinder 13:19

Yeah. So anyway, I I hear a lot about Chinese investment in AI, but that's maybe for just for internal use or internal companies, and nothing goes outside the walls, the borders.

Michael 13:30

Yeah, plus I'm not sure how much they're investing in these kinds of maybe AI in general, but AI specifically for engineering. Yeah, I'm not really sure, right? This is all for making stuff that I like to call this one. So basically, that's what I want to show you. I want to show you that I've got this killer database with some killer analytics on top and a killer report. And um not sure how to go to market with that, not sure how to how how best to do that piece.

Startup Conferences And Founder Network

Michael 13:58

But also, yeah, definitely want to be on the podcast with you and and talk about uh the state of the market. I I created this startup conference. So I did the first one last week in Warwick. I had 20 startups there. It was really cool. Unfortunately, there was a train strike, so the attendance was down compared to previous years, but uh my 15 startups was were super happy. I had a couple of customers there, including NASA, who was very happy with what they heard. And next week I'll be at CD FAM in Barca. So I'm gonna meet Physics X finally. CDS will be there, a lot of my guys will be there.

Roopinder 14:30

And the week after that, of course, it's a uh did you latch on to what Martin is doing over there? Martin Day? Martin Day has this uh he has a whole convention going on, conference in the Warwick area.

Michael 14:45

Oh, really? Yeah, it's oh you're talking about Developed 3D Live, or you're talking about something else? Yeah, D3D, yeah, yeah. That's where I had my Threaded conference. It was a sideshow to D3D. So yeah, I met Martin. Of course, I worked with the two Steves there, the main guys that I was working with. Nice group of people. Really cool meeting Rushik of CDS. I didn't know that he's the one who opened Japan to uh NTOP, and he gave me all this background on on Brad and nTop. They did their series D back in 2021, and it's really hard to raise Series E after that, right?

Roopinder 15:16

You've got to prove he's discovering himself to be the aerodynamics designer that he always wanted to be since as a kid, but I didn't I didn't know he was doing it for the military, but it makes sense though.

Michael 15:28

No, that's a lot of money over there, right? Yeah, I don't think it bothers XT.

Roopinder 15:32

Have you got your been able to get a hold of uh Paul Murlucki at all? He's the uh CEO and founder of Andorral. They're big and huge. I know they're big.

Michael 15:43

I know, but they're hardware, right? They're both hardware and software.

Roopinder 15:46

They're hardware, yeah.

Michael 15:47

I haven't no, I haven't. Have you talked to them? No, no. I probably will never mention politics on the call with them. I would like to meet Astari. I know Astari is doing a lot of stuff, and they're very present. Um, but I have not been able to contact them. I have I do know um Steve of Siskit, uh, he's a fantastic guy. I really love Steve, he's amazing. I'll meet at seven in 40 minutes. I'll be interviewing Chris Barton of Drafter and uh Giannis Faver of uh Tracepace.

Roopinder 16:15

Okay.

Michael 16:16

Amazing startups as well. Although admittedly, Chris's software a year ago isn't very mature. I think he's done a good job uh getting it much more mature now. I'm happy in my in Miami. I'm gonna have it hopefully I'm gonna have at least two. No, I'm gonna have definitely three, maybe four female co-founders there. I'll have I talked to Lucy from Violet Lab, she's like amazing. I'm gonna definitely have Masha of Nil Space, who's also awesome. And I'm gonna have Addy of Quarter 20. And hopefully I'll get Valentine. I don't think Valentina's gonna make it, but if maybe Valentina or somebody from Allspice, you know, the ECAT got the folks doing PCB stuff. So that'd be awesome if I could get three, four co-found female co-founders, because there really is not that many, right? There's also Elisa Ross up in Canada, but she's otherwise occupied. But she's also, and by the way, Elisa is the is implicit F rep number three. Like F implicit's number one is N Top. Implicit's number two is CDS, number three is is Alyssa up at Metafold. I had not realized that until I talked to Rushik.

Roopinder 17:20

There's one more. I don't know if you have. Have you heard of Dusty Robotics?

Michael 17:25

Yep, I've got them on my list too. I've not talked to them, I have not met them.

Roopinder 17:29

But there's a female founder for you. Tessa. Yeah. Oh, okay. Yeah. They have her. Dr. Tessa Laud, founder and CEO of Dusty Robotics Construction Technology. But uh yeah, very interesting woman. Very uh, she's been quite successful as far as I can tell, though you could probably dig up more financial information than I tried to do.

Michael 17:51

But yeah, I do have information on them in my database. I I did look them up.

Roopinder 17:56

Okay. Yeah. Yeah, but you're right. It's so hard to come by anybody female in this business. I'm not sure why. It's even less represented than it than in the general CAD PLM space.

Michael 18:09

I agree. And like for Europe, I had zero. I have zero. All these female co-founders, either American or Canadian, and in Europe, I don't know of any. Despite knowing over a hundred startups here, I know zero with female co-founders, unfortunately. I mean they have officers but not founders.

Roopinder 18:26

Yeah. Well, it says it's been a while since we talked.

Funding Wave And What It Means

Roopinder 18:29

Um let's just say roughly nominally a year. It hasn't been that long. Let's just say it's been a year. A lot of would you say there's in the beginning, I think off that year, we were like amazed at how much funding was coming in to this business because, like we said, these VC people didn't even know how to spell CAD or PLM before, and now they're putting money in. Have you tracked the rate at which funding is coming into our space, our industry?

Michael 19:00

Yeah, it's in the report actually. It's it started coming in really heavy in 2022, 2023.

Roopinder 19:06

Uh-huh.

Michael 19:06

Um, two years after COVID, it just spikes like this. The number of startups and the mud the money got comes in at 2022, 2023.

Roopinder 19:15

So how does this uh how does our industry compare to AI money? Is it is it uh part of the same rate as like other industries in general, or is it like uh is it becoming a how should I say a jewel of investment?

Michael 19:33

That's a good question, man. I have not investigated that, but it's uh one I should probably investigate.

Roopinder 19:38

It's because in the public in the public press, so you hear about uh popular press, I should say you hear about it coming into medicine or war, right?

Michael 19:48

Web development with lovable, right?

Roopinder 19:50

Exactly, exactly. Yeah.

Michael 19:51

But into engineering, you don't hear it as much. Well, I think unfortunately, again, a a lot of the uptick, at least the last two years, has been the Department of War, tons of money being thrown at it because of further drone development. But before that, I think there was a lot of there's still a lot of excitement around generative design and um stuff like that. And the what I'm finding it's super interesting also is looking at the uh manufacturing space.

Prove It Conference And OT DevOps

Michael 20:19

I went to prove it, you know, the Walker Reynolds thing. I don't know if you know that whole world and how there's this weird cult of personality. No, that doesn't ring a bell. Okay, so if you go to the manufacturing side of the world where you've got the Rockwells and inductive automation and guys like that, right? That turns out I really like the Ignite guys. Rockwell, not so much, but Ignite, I really like those guys. So over in that world, this guy Walker Reynolds came out of nowhere a couple five, eight years ago. Um, had this idea of unifying the namespace on top of the PLCs and scatters everywhere, right? So we came to this EONS, unified namespace. Vendors started adopting it. Big customers started saying, holy shit, that could really help fix a lot of problems. He became moderately wealthy as a result, mostly because this kid named uh Zach Scriven knew Discord really well, knew really well how to promote things on YouTube, and basically created a phenomenon out of this Walker Reynolds guy. And Walker is he is brilliant, he's a great engineer, and he's very passionate about engineering, but he's also pretty MAGA in a lot of ways. He did have this idea approve it, and his Prove It Conference is awesome. Um, it's only been two years. The first year there were 25 companies there, and he claims that the ones that did not prove it, four went out of business, and two were reduced to two or three people from being a hundred people at the conference. And why do I say that? Because if you're gonna uh present at Prove It, you have to run your software on his virtual factory. So he basically creates a virtual factory just spitting out data from all the in all these formats, from Zemens and you know, FanUC and all these other. Controllers. You're just spitting data out of a real, you know, from which looks very realistic. And you have to run your software on top of that stack. You have to prove that your thing works. And this year, I wasn't there in 25. I didn't even know it existed until after the fact. I went this year. There were three factories. There was a like a drinks factory, a labeling factory, and a side life sciences factory. And so each of the vendors had to run their stuff on top of this stack. They only had like nine, eight or nine weeks to get it to work. I love this idea because I would love to see Zima's DSPTC try that. I don't think they could do it. Maybe they could, maybe I'm off. But I I think it would be an interesting experience, right? This year he had 51 companies, 51 ISVs running on this virtual factory, which is like pretty amazing, right? Only the second year of a concert he had 50 people. I saw these demos that were just absolutely mind-blowing. So, of course, the big problem in the industry now is the death, the resurrection now, but Kepware and ThingWorks. That whole PTC sells them off to TPG, and then Jim and Brian come back with Velotic. But still, Kepware, I don't know if you've ever seen it. It's pretty crappy. I mean, it's awesome because it connects to basically to anything in the factory, but it's kind of crappy because it's an old Windows software. It's just really like Windows 3.1 almost software. PTC just let it die on the vine. That's why they sold it off because they never invested a penny in updating it. Same with ThingWorks, right? ThingWorks, as much as Rick loves it, it was state of the art 2008, right? And it never evolved from that state. We'll see what Jim does in the version three or whatever, you know, whatever we want to call this revolution over there. All that to say that the entire industry absolutely adores Kepware. Everybody adores it because it can read anything. But they've already started raising the price of it. Litmus, I don't know if you've ever heard of Litmus. They're a startup on that side. They basically demonstrated, they said, you know what, we've we're gonna do all three factories. In fact, we'll add a fourth factory. And they they showed that they could do the whole thing within in literally minutes. They could deploy the level one, level two, Isa 95 stuff like instantly. It was pretty awesome. And then induction inductive automation freaked me out because they said, okay, we're gonna take this one factory um and we're gonna use Git and we're gonna use Kubernetes, and we're gonna deploy the whole factory in three minutes. Right. And they did the whole thing all the way up to like that. You can see it was astounding in three minutes. So all this stuff we've been doing in IT for the last 15 years of DevOps has now hit the OT space. So I think there's really is like over there where we don't see very much, there's already like an open AI moment happening. Yeah. Over on our side, on the engineering side, we're not there yet. We're still arguably six months to two years away from the open I moment. It'll come, but we're certainly not there yet.

Roopinder 24:55

Back to AI and CAD for a second. No, that's

The Startup Flood And Consolidation

Roopinder 24:59

okay. I love this. You're gonna listen to you talk all day. But listen to this. They so there was a company, and I think you referred to it, uh, it was uh Parth Meta.

Michael 25:10

Yeah, the guy Maker Street out in Austin.

Roopinder 25:13

And he said he got funding before he could even legally toast it, right? Yeah, he's only 20. I forget the number. You got like millions. Uh when I when I first saw that, by the way, I interviewed Parth, by the way.

Michael 25:29

Very super sweet guy, yeah. I love that guy. Someone's gotta hire him. Someone's gotta hire him because he's too smart to or he'll just become an anyway. Go ahead. So you interviewed Parth.

Roopinder 25:39

He's a super smart guy, very knowledgeable. He's he's a he's an athlete, he's he's just amazing to me. But it made me think, though, uh it's not the if I was a VC in the in the normal world, let's say just five years ago, I wouldn't have given this guy any money because he's just he's still in school. He's not he's not an MIT graduate, right? He doesn't do all these things, he's not an industry that I think is going to take off, like fintech or medicine or anything, right? Here, but it made me think like, how easy is it for somebody just to raise their hand in CAD or PLM, right? And say, hey, look, I do AI, give me some money. Because it seemed like if you give it to a 19-year-old, you know, he's an exceptional 19-year-old, don't get me wrong. There's that is it that easy to get funding still in for AI companies in our industries?

Michael 26:32

I have no idea. I I know it's that easy to throw up software because of cloud code, that's for sure. Right. Um, to get funding, I don't know. It depends on which VCs you're talking to, I guess. I I've tried to map that ecosystem and try to understand um where the money is going, and uh it's still uh I'm still studying it. I I don't know. I there probably are a few that would do it. I some might be a little bit more skeptical.

Roopinder 26:57

Yeah, yeah. Now, what do you think about these companies? Uh uh now charted how many over a hundred of these AI startups for CAD and PLM?

Michael 27:09

Uh well CAD there I have 99. Uh 99, okay. And analysis have 58, and after manufacturing have 30, and then cognitive thread I have about 75 or 80. So there's about two two to three hundred.

Roopinder 27:23

Two to three hundred companies.

Michael 27:25

Yeah, it's a lot.

Roopinder 27:26

It's growing, and uh a few of them you know you noted have have taken off or are in the let's say the high end, they're gonna get traction. Some of them already have customers. It's where do you see this heading? Is there gonna be like a winnowing out of all these companies? Because obviously there can't be that many that are going to succeed and become mainstream, or do you see another market where established players are going to start buying buying them up or buying a few buying at least the talent from these companies?

Michael 28:00

I think there'll be a bit of consolidation, but at the same time, it's not like when I look at some of the the startups on the on the software development side or the product management side, and they a lot of them look like BS. When I talk to a lot of these startups, I'm like, they're actually actually really serious, and they all have these brilliant ideas. Like take sphering, the idea of let's model the inside of objects rather than the outside. Let's use spheres and isotropic modeling. And I was just talking to Claudio today, and he was telling me how he they've also redesigned golf clubs because they figured out where the sweet spot is. So they design the golf, redesign the rebuild the golf club around where your sp your particular sweet spot is as a golfer. I'm like, wow, that's super cool, right? And plus what a great demo thing. You go to the, you know, you you you do any trade show and you just give away a golf club. That would be you get a lot of people to come by the booth that way. Um I think that a lot of them are doing a the wise choice of focusing on one vertical. Like compute maritime, all he does is boats, or like Axial 3D, Ollie does is medicine, second mind only automotive. Now in top only aerospace. That's probably a good idea because you that way you can really kind of fine-tune the technology that one industry, and then hopefully scale when you get to your series B, then you scale out to other industries. Not what in top nonwithstanding it being not the right example there.

Roopinder 29:20

One company that's I think is kidding that it's tried in marketing is uh Theo Theo AI and Mayor Farid. I think he's at least if you look at LinkedIn, he's like all over it. He's like you know, he's got a lot of followers. I notice you're in conversation with him quite a bit, a dialogue with him on internet. What do you say? Do you say do you see him as one of the his company as one of the forerunners here?

Michael 29:47

Well, he certainly he was very early on that same idea on the engineering co-pilot idea. And of course, he he was at very early, got John excited, and John carries a lot of weight, right? Yeah. Uh but I know five or six other ones. You know, John and I'm not an on-shape user, so I can't say like I've used it or anything like that.

Roopinder 30:08

You know, John Herstick's son is selling for mayor.

Michael 30:13

No, I didn't know, but I'm not surprised.

Roopinder 30:17

Yes, and not a I don't know if he's investing in it, but certainly gave them his firstborn. I don't know.

Michael 30:25

I found it interesting that I tried to do I did a podcast with SolidWorks with um experts that were not employees of for SolidWorks, Solid Edge, and Creo. Um and then I wanted to do a follow-up with the other three, which would have which would be um OnShape, uh KTA and NX, right? If I look at the big three. I've I spent seven months trying to fit in an X, NX. I had people coming out of the woods saying, Yeah, I'll do I'll talk about KTI, I'll talk about this. NX, nobody. Uh it ended up because I had a I was having a conversation with a guy that wants to create a startup in the aerospace industry, and it turns out he's an NX guy. I'm like, oh, would you like to be on my podcast about NX? And he's like, oh, okay, sure. So yeah, he that's the only way I could find a guru. All I'd say, yeah, I think the on-shape is I don't know, I don't know what the problem. I don't know, understand. I mean, it's great technology, but the adoption curve has not been as spectacular as one would have expected.

Roopinder 31:22

They seem to be surviving. The group seems to be on space group seems to be surviving on its own, sort of acting independently. They may not have the uh let's say the blessing of PTC on a on a every level, but uh they seem to be acting independently. I don't know. I don't know what it is either, Michael. I see a lot of young, younger engineers and startups taking on Shape On as their primary CAD tool, but it's not enough to change the game, change the map, as it were.

Michael 31:53

No, and they're not getting the volume, and and a lot of those users are free licenses, they're not necessarily paying users.

Roopinder 31:59

Yeah.

Text To CAD And CAD As Code

Roopinder 32:00

Hey, I read I'm reading your I read your uh article, Text to CAD, OpenAI. By the way, I love the way it starts out. Chat GPT landed and world split into before and after. I just it's just a great great opening sentence. Uh I'm getting two schools of thought on this subject.

Michael 32:20

One is like the guys like Zoo and Adam and guys like that, right?

Roopinder 32:24

Exactly. But what there's one school that says AI is text to CAD is gonna change basic CAD, and there's another school that says AI is going to use CAD or the other established programs and just change the interface, right? It's just gonna make them easier to use. But the tools are still gonna be there, the tools are still necessary, right? So it seems to be like a divergence, right? We're either gonna use AI to make our existing tools work or work better or work easier, or we're just gonna throw them in the trash and now cat this AI is gonna somehow make those shapes and do that simulation or you know, do that organization that we use these tools for. You seem to be on this on the side of yeah, there's gonna be a new engine. AI is gonna be changing the engines.

Michael 33:16

Well, I think that at one point AI will be able to we'll have models, foundational models that'll be far more versant in 3D and in geometry and all these problems. I mean, um already we've I've seen lot uh several well Maur, of course, did it large mechanical model, but it's not really 3D, it's more the mathematical equations behind it, it's not the geometry. I'm already I've I'm talking to another company in Switzerland, AWV Informatik, who's already doing CAD as code. He can already ex he can already express geometry in JSON all the way down to the feature tree and stuff like that, which is super interesting. Because if it's text, then an LM can work on it. And you can do CPQ and simulation and all kinds of stuff with it. I think it's a matter of time. I I think that um if I could get those guys to figure out how to pull the parametric kit the fib feature tree and all that stuff out of SOLIDWORKS, GTA, NX, whatever, then I think the big three would have to be word. Well, they'd they'd sue because they'd say, Oh, it's priority, you're not allowed to look. But once the cat was out of the bag, you could get all that stuff to text, then you can add a limit, then you can basically then go back. See, I think the text to cat is nice if the cat at the end, if the end product has the parametric history, if it's just a dumb step file, yeah. Who cares really, right? I mean, what you want is the rest of the history of my engineering history, my design intent that was captured in the initial model. So I don't know. I don't will we hit I I'm I'm wondering, I think the open AI moment or one open moment will be when we can go from parametric CAD with all the history to some textual thing where I can actually talk to my model and figure things out and then go back to a 3D model that's got that doesn't lose any fidelity and doesn't lose any of the the history. Yeah, that'll be an open AI moment because suddenly now you could basically thin uh we could create stuff just with uh with LLMs or whatever the next generation of LLMs is gonna be called. Yeah, yeah. That said, I I did talk to, I don't remember, it was one of the founders I talked to. He said, Yeah, well, the other day I was talking to my phone and uh it was sending a command over to my 3D printer and it was printing shit as I was speaking. Okay, that's pretty wild. Um, I mean, we're not talking about building uh uh um a cyber truck or uh a 737 or something, but still, you know, it's interesting, right? And then um another thing I I talked to uh I invit I had a my first paid podcast, first time someone paid me to do a podcast was Vince last night. I don't know if Hardick and Sarah over at Vincey.

Roopinder 36:01

Oh pretty brilliant, you're actually making money off this. You know, each all this knowledge for free. That's good.

Michael 36:08

Yeah, finally, finally. But Vinci is basically they built their own financial model for heat transfer, and they're doing they can solve a trillion equations in less than 20 in like 26 hours or something, which is just mind-boggling. And he's focused on semicon and stuff. Very cool. You should talk to Hardic. Very cool uh organization over there.

Roopinder 36:26

So not doing a like linear algebra, large matrix solutions or to find some other math for this.

Michael 36:34

No, it's all differential equations. It is okay, yeah, yeah. Just differential equations. It's just he's figured out he's got some fundamental models and and uh foundational models and some ways of solving some techniques that are super fast that uh run on these GPUs natively. Basically, he's written a bunch of code to run natively on GPUs and just go as fast as you can. And he said the minimum is 80 gigs of uh of NVRAM on the GPU, and then he can solve just about anything. And like that's pretty amazing. That that's sort of an open AI moment. We can do trillions of equations in in hours. That's pretty close to an open AI moment.

Roopinder 37:11

Which makes

Rebuilding PLM And PDM With Agents

Roopinder 37:12

me think. What do you think of this? Now, I uh just listening to this, I'm thinking I'm marveling at it, you're marveling at what's going on, how much technology is available. What do you think that person on the street is saying? Let's say the engineer on the street, how are they just like paralyzed by stuff that's going on? There's so much going on that they don't know how to implement it, don't know where to start even.

Michael 37:32

Yeah, I well, I think that they're held back by the bat, the the big three, because the big three are sort of like basically standing on the as someone put it, I won't I can't take credit for it as much as I love the quotes. They're standing on the garden hose of innovation in a lot of ways, right? Because what have they announced in the last three years? DS, PTC and Z nothing, nothing in terms of AI, like barely even kit catching up to the co-pilot train. So I think a lot of some of them are coming to me. A few are coming to me saying, Fino, you seem to know the startups, and and we really need AI because the board is on our butts, and we have to show them something.

Roopinder 38:07

Right.

Michael 38:08

So I think some of the I think there is a bewilderment factor, as you said. I think that people are starting to realize that if I find if I'm able to pinpoint my use case, I can probably find a startup that can do it. And so I don't know if you saw my article about it, but I thought, well, maybe what we need to do, we need to scale back our ideas of PLM to just what PLM was in the beginning, before it sprawled off into all these other things. And just focus on PLM is product data, right? It's got the parts, it has the assemblies, it has the eBOM, it has classification, change management, and that's it. All that other stuff we build is on the outside. So you put the everything in the core, you wrap that in a semantic layer, you wrap that in an ontology, then you put an agentic layer, and then I can plug in a PDM. Because why would PDM be part of that? PDM is something else. It's a file management, right? And why not put the MBSE over here and the work instructions over here? And suddenly you end up with sort of a kaleidoscope of these tools that are best in class that agentically are able to go and talk to and by the way, if there's no agents, I'll still use REST. I don't need, I don't have to have the agency backplane because I can put a REST interface instead, but at least I can eliminate all this customization that these guys have written for 30 years that doesn't work. One of my customers is stuck at Creo 10. They can't move off of 10 because too much customization there. They're not they can't support anything else, which is insane, right? I mean, they're just invested in it.

Roopinder 39:37

It work you said about PLM works for CAD too. I mean, PLM should be good at just doing product data, but there was a time where they thought, oh, PLM should be the overarching software of the company and its business processes, and and the same thing with CAD. CAD companies got really kind of got too conceited and thought, oh, they should also be simulation companies, simulation simulations, eCAD and and software management and PLM, right?

Michael 40:07

And all that stuff. My vision too, Rupinder, is that um we someone needs to build it. Maybe and maybe someone will come with it. Maybe Aeris is already building it because they sort of hinted at it, but I'm not sure they quite understand. Maybe build will do it. Maybe Pradu will build it. I don't know if you've talked to Pradu, absolutely brilliant engineer, also, awesome guy. He came to the show and work. We need a PDM that understands simulation and manufacturing, as well as software and ECAD, right? Because why do I have different platforms to manage all these other things that I'm that are all based on my 3D data? Right? Why doesn't a PDM system understand what a G code is? Why doesn't PDM system already know the input and output formats of all the simulation programs, so I don't have to send out a stupid step file, which gets lost, and then it's out of touch anyway. I'm out of date and I broke my digital thread and the simulation data comes back and it's not even relevant because my design's moved on three three revs since then. How do we get rid of that? And plus, I think the other effect that we're gonna that we're seeing, the other open opening moment is the compression of all these things into the same workflow. I don't think you can be a mechanical engineer anymore that just has mechanical engineering. You have to understand manufacturing, support, service, simulation. You all those things are now available to you through agents. So you should, I mean, as in a mechanical engineer, I think you're gonna have a much wider scope than you had before because you're gonna be able to immediately test an idea and send off to simulation to see, well, then the CFD thing does it work. And you know what I mean? I think that those are the things that are gonna be moving in the next two, three years. And those are the people that are gonna do the best in terms of survival, the ones that solve those problems.

Roopinder 41:47

One thing we talked about before, which was like a re revival of something like COFAS.

Michael 41:54

Yes.

Roopinder 41:54

And uh, and boy, wouldn't it be nice if all these companies that you're looking at had a place to show off their own.

Michael 42:02

Okay, so here's my thoughts on that, Ripender. So

A New Engineering Startup Festival Plan

Michael 42:05

I'm already I've I that's why I started the whole the threaded thing, right? I've did my conference of Warwick, I'm gonna do one at AirSace. I'd probably do another one in in maybe in um in maybe Tokyo, but certainly um war Frankfurt later in the year.

Roopinder 42:22

Okay.

Michael 42:23

I have a friend of mine that had that owns, it's his own thing. He owns a conference center in Denver. Denver is like right in the middle. You've got Boston on one side, LA, San Francisco on the other. Denver's in the middle.

Roopinder 42:36

Yeah, you know, everybody could get it.

Michael 42:38

I'm thinking we should do we should do it in Denver. We should do it maybe closer to the winter so people we do it late in the week, people can fuck off to Brackenridge and do some skiing. We still have to figure out a format and we have to figure out sponsors. The sponsors will not be DS Eemus PTC. No way. No way. No, they're not allowed. I don't want to hear 3DX or Windshill or Team Center at all. I don't want to hear those words. I want to go for AWS, Azure, Nvidia, GCP. They're the ones that benefit from a healthy ecosystem. Let's try to drag Apple into this debate. Yeah, because Apple with unified memory is the only people that can really compete against NVIDIA in terms of LLMs and stuff for an engineer. Of course, CAD is another thing. Unfortunately, they don't run on Mac, right? SOLIDWORKS and all it doesn't run on the Mac. But if I'm doing engineering and I'm doing manufacturing, I can run all that on a Mac and I can run it faster probably than any Windows machine on the planet. So why isn't Apple a bit more aggressive trying to sell to engineers? Anyway, it's probably a pipe stream because they've never sponsored anything outside of WDC and they probably never will.

Roopinder 43:41

But they seem to have no interest in engineering, with the exception of Shapr, which they once gave an award to, right? Exactly. Yeah, that's been their loan foray into our world. Uh and I can't get through them. They're like a walled garden, which is like sealed off from I'm trying a couple of avenues, but I haven't been in a It's frustrating because their computers are really good computers.

Michael 44:03

They're awesome. I love them. That's what I'm using now. So think about Colorado late in the year.

Roopinder 44:08

Yeah.

Michael 44:09

Think about how we bring together because I think we could easily get 100-150 startups to show up. If the venue is free, all we need is the catering. The catering is going to be less than 200,000. So we're not talking a huge budget either, right? It's not a very big budget. So if we could get again the hyperscalers to kick in, if we did uh if we could also get in VCs to kick in, maybe uh plug and play or YC or Bain or A16C.

Roopinder 44:36

That's a good idea because the VCs have already got a lot of skin in the game, investing.

Michael 44:40

And I can say which ones are putting into which startups. I already have that information.

Roopinder 44:45

By winter, they ought to be, I think they ought to be like, what's the next step? Where do we go here? How do we get it into the public's hands?

Michael 44:51

So work with me on that. Let's let's think about that. So I'm thinking there's three there's three axes, right? There's the axis of a day just for the startups to talk among themselves without the distraction of customers or VCs, right? Maybe there's a day or two days where VCs are there, and maybe we could do speed dating, or maybe we do like Web Summit where we have areas for different maturity levels, and maybe there's a pitch contest which ends in a 200k endowment. And you know, you get to present your thing in front of the guy from Bain for 200K or something. But then I think we also need a day somehow where we bring in customers, where we tell Lockheed and NASA and Boeing, do you want to see the 150 best engineering startups on the planet come to Denver on this day and you have access to all of them?

unknown 45:35

Right.

Michael 45:35

How do we do that?

Roopinder 45:37

And then we have the best show on the planet. I think it'd be the best show and the most interesting. I think it'd be fascinating. I think everybody would go there and say what's the what's all the fuss about? Let's see what works.

Michael 45:47

And we would do it would be awesome.

Roopinder 45:49

Yeah, yeah.

Michael 45:50

It's always a pleasure to talk to you, Pender. And I'd really like to talk again in like two, three weeks and brainstorm a little more on this uh new fest thing.

Roopinder 45:58

All right, so much fun. Thanks. Thanks for coming on, Michael. Take care, enjoy. Talk to you later. Okay, bye-bye.

Closing Thoughts And Farewell

Roopinder 46:03

And that concludes this episode of the Masters of Technology Happy Hour. I hope you enjoyed our chat and will join me as we talk to some of the most interesting people in the design and engineering software community.