#699 – CircuitHub, 12 Years Later with Andrew Seddon

Chris Gammell: This is The Amp Hour Podcast. Released July 31st, 2025. Episode 699. Circuit Hub 12 Years Later with Andrew Seddon. Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics.

Andrew Seddon: And I'm Andrew Seddon of Circuit Hub.

Chris Gammell: Welcome back 12 years later. How are you doing?

Andrew Seddon: I'm good, Chris. Yeah, really good. I can't believe it's been so long.

Chris Gammell: You know, time marches on and you have been busy. We were talking a little bit before the show, just like, if you go back and listen to episode, actually, Andy reminded me to look at this, 131. So that's a couple episodes ago. Yeah, you kind of had a different product and different thing back then. I've obviously watched you kind of keep going and watch Circuit Hub keep developing over time. But can you tell us first what it was back then and then what it is now?

Andrew Seddon: Yeah, I mean, we've been through a few different iterations at Circuit Hub. I mean, when I think last time we spoke, our idea was to be a crowdsourced library for EDA data. So, you know, this comes from my background as a design engineer. I kind of always thought it sucked to have to generate the symbols and footprints for any design that you're working on. I kind of just wish that there was this centralised cloud library that you could pull from and you can avoid all of that additional effort. And so the first iteration of the idea was that, you know, we'd have this website. Everybody could contribute their symbols and their footprints and we could sort of crowdsource validations for those and people could download them to the local EDA tool. And I think about around the time that we spoke last, we had probably just launched that product and it did okay. We the big learning there was people want the work done for them, but they don't particularly want to contribute to to to the shared library. And so we had a bunch of people saying, hey, this is great. Like, you know, we'd love love to use this product once the library is fleshed out. But we actually don't necessarily want to spend the time contributing to the library. And so so we kind of we'd always imagined really Circuit Hub as the EDA library was the starting point and that we were going to go and fix manufacturing. Right. So we had this this idea that we would would acquire a bunch of users by the EDA library and then we'd go off and make it much easier to manufacture circuit boards. And so we kind of half pivoted, but I would say it was just a natural extension of what we already did into into electronics, electronics manufacturing. And so we've been through a few different few different iterations on that model. And, you know, the way that we started really was with a sort of network of other people's factories. So we made it super easy to upload the design files for your circuit board. We would give you an instant quote from one to ten thousand units. So we give you a lead time, give you a price. You can click a button, pay by credit card, and we then get those boards made for you and sort of ship them out. And so we operated that model for a few years with this network of factories. And we actually ended up getting super close to a company on the east coast of the US, Worthington Assembly, which funnily enough, to kind of close the loop on our first conversation, Chris, I think I was introduced to the CTO of that company, Chris Denny, by you after our first podcast all those years ago. So I think after that.

Chris Gammell: Yes, I've come to collect. I've come to collect.

Andrew Seddon: Indeed, indeed. Yeah, definitely. You know, that's that's that's been an awesome relationship in sort of many different ways. So we've we've really got you to thank for thank for sort of setting up that setting up that relationship in the in the first place.

Chris Gammell: And so we also mentioned Chris Chris Chris's you can hear Chris every every week, every month. I don't know how often is but the pick place podcast. That's what him and I forget his co host name. So that's weird. Melissa as well. Yeah. Yeah. There you go.

Andrew Seddon: He works on our marketing team.

Chris Gammell: Yeah. Great. And yeah. And that's a great podcast as well. Like deep. I just remember like board cleaning was like one of the episodes I listened to. I was like, oh, board. No, it's multiple episodes. I think it's like board clean. Oh, OK. Yes, I do need to hear this, actually. So we'll link that down below, too.

Andrew Seddon: Yeah, it's kind of Chris gets a bee in his bonnet about, you know, some some particular niche aspect of manufacturing and like, yeah, and then goes on a rant about it. It's pretty fun to listen to.

Chris Gammell: But that's what I want. I want I want to like hire experts, you know, who care about circuit boards being seamlessly sent to manufacturing or manufacturers that are seamlessly cleaning board. You know, I don't I don't think about this stuff. I just want it to show up and someone to be thinking about that. It's great. That's really great. So and he and and I will listen to it. You know, listen, I'll be a voyeuristic listener, you know, just be like, yeah, yeah, that sounds like a tough problem. Glad you solved that. So what kind of in the midst of speaking of tough problems, I mean, like all of these things you've mentioned that circuit has been doing over the years. None of them have been easy. I would say the the library thing still not really solved sort of kind. I mean, like there's other solutions that are out there, but they're all kind of still, you know, like we've had guests in the past that are doing it, but it's not like it's just there's so many parts out there. So manufacturing is not getting any easier. So what have you been doing to solve all these problems, man? Like it doesn't seem easy.

Andrew Seddon: Yeah. I mean, I kind of jump, jump forward a little bit to where we are today, maybe, which is we have essentially had a ground up rethink of how electronics prototyping and small batch manufacturing should be done. So, you know, we started off with this, this network model on top of other people's factories. We actually set up our first whole home factory back in 2017. We did it using, you know, pretty common off the shelf equipment. We wrapped that up in a bunch of software and we've made some process improvements doing that. But, you know, it gave us a bunch of ideas about how we could, you know, really just go back to first principles and think about how can we make a super efficient prototype and small batch electronics facility. And so for really about the last six or seven years, that's what we've been doing is kind of just had our heads down doing this ground up rethink of how to do manufacturing. And so where we are today is we have this fully vertically integrated system on the east coast of the US. And a good way to think about it is a little bit like a semiconductor fab. Like have you ever seen kind of videos of like, say, like the Intel fab or like TSMC? We've taken a lot of the techniques that originate in semiconductor fabrication, which is like one of the most advanced manufacturing processes on the planet. And then we've applied them to the assembly side of the fabrication side of the operation.

Chris Gammell: And does that mean you have foops? Because that's one of my favorite words.

Andrew Seddon: Foops. Foops. Yeah, we don't actually do it directly with foops. So, you know, conceptually it's very similar to a semiconductor fab. The actual problem is quite different, but they are cool, right? Those things dropping out the ceiling and kind of throwing things.

Chris Gammell: Front opening unified pod. Foop.

Andrew Seddon: So, yeah, it's an interesting problem. I mean, you know, semiconductor manufacturing has some pretty different requirements to board level assembly, right? I mean, one thing is that the products that you're handling are like very uniform because you're dealing with a semiconductor wafer, right? So it kind of looks the same every time. Whereas in board level assembly, you know, there's the theme that everything's on a circuit board. But in practice, those circuit boards can all look pretty different, different sizes, shapes. They can have, you know, large components on them, big connectors and things that cause them to warp in all sorts of interesting ways. And so, you know, the mechanics of how you actually put that together are quite different.

Chris Gammell: Well, I think just the input, number of input chemical, I mean, like, so you show up at a, so I used to work at a fab. And, you know, you show up at a dry etch station like I used to work at, and it's got, you know, RF energy and like 20 different chemicals maximum, right? But that's different than like, you know, 20 different types of boards in a day showing up to, you know, assembly station that needs, I mean, I'm guessing you just standardize and have what the, all the caps are on one machine, all the resistors on another machine, that sort of thing. But like, still, you need all of those different machines to have all of those different types of parts. So like that, that's still intense for that sort of thing. Yeah.

Andrew Seddon: Well, yeah, so it's funny, like the original thesis for the company was that we were going to try and drive everybody to 2000 common parts, right? So I kind of had this idea from my own experience of doing electronics design where, you know, I did a work as a design consultant for many years, did a whole bunch of different projects in, did a bunch of camera systems, things for the entertainment industry, medical devices. So pretty, pretty varied end products, but I kind of accumulated this library of parts that I was pretty happy with working with, right? Like I was familiar with their functionality and I was able to combine those individual parts into various different unique products. And so we had this thesis that if we could drive everybody to this single common set of parts, we could make it much easier to manufacture any product that was based on those common library of parts. So we were pretty early to that game. There's a few other companies that have tried to do that since like Octopart years ago came out with their common parts library. I think, you know, a couple of other people have sort of attempted to do this.

Chris Gammell: And Macrofab had in-house parts. I think Seed has their, they use the common parts library as well. That sort of thing. But those were more like standardizing on types of parts, right? And then having like, that was almost like an approved vendor list. And then it was like, well, these are, you know, penny type parts anyways. Who cares what's underneath the hood? You don't have to care about the number underneath that sort of thing. Is that what you, is that what you're playing, playing it there though?

Andrew Seddon: Yeah. I mean, we, you know, we, we were thinking both on the, you know, resistors, capacitors, standardizing those, but then also standardizing on, on the IC side of things. Right? Like you find a lot of commonality between, you know, various like power chips, microcontrollers. And, and so, you know, but the, the, the, the long and the short of it is that it's, it's really difficult slash impossible to get people to standardize on a common part set. Right.

Chris Gammell: So I was going to say, after 700 episodes, I think one through line is that electronics people don't like being told what parts to use. They want to, they want to use the parts they want to use. And yeah, you're exactly right.

Andrew Seddon: And so like the problem on the problem on the, uh, the kind of like the passive side of things is that, you know, it's more work for me to qualify a different resistor or capacitor supplier than, than, uh, than it's worth. And so I just don't want to do that. I don't want to like switch to your common parts library. And then on the IC side of things, uh, you know, it turns out that there are these 2000 common part libraries, but everybody has their own unique set of them. Right. So as a design engineer, I don't want to switch to the, to the set of parts that, you know, I, I want to use the set of parts that I know. And so, you know, this whole concept of like driving people to a single, uh, common parts library just turns out really not to be possible. Um, so, you know, where, where we've ended up going with this is, is kind of deal with the, uh, deal with the fact that, that people don't want to do that. They want to use a large variety of parts and instead set the manufacturing process up process up to make it super efficient to deal with a large variety. of parts. Um, so we think there's probably. Um, so we think there's probably about a million different skews that, that people want to use. And, you know, a good intuition around that is if you look at the catalog suppliers, uh, you know, you look at mouse, so you look at DigiKey, uh, you look at Farnell over in the UK. Uh, these companies tend to have anywhere from sort of half a million to a million skews that they're dealing with for, for parts that go onto circuit boards. And so that actually represents the set of parts that, that you people want to use that design engineers want to use. And so, uh, so what we're doing at circuit hub is, is setting the factory up to very efficiently handle, uh, that large number of parts. And, and so today we actually stock, uh, I was just checking the numbers the other day, about 50,000 different unique, uh, skews and we have those available ready to go. So when you order your prototype, uh, circuit boards, we often more often than not these days, we'll actually have those parts in stock, uh, ready to go. So, you know, we're keeping like small quantities of, of high value, uh, ICs, uh, you know, we've got a lot of, uh, resistors, capacitors, pretty much everyone that you can think of. And, and rather than trying to have everybody use the same hundred ohm, uh, resistor, we just like stock all of the hundred ohm resistors you possibly want. And then we don't ask you to like substitute into a, uh, into a different, uh, you know, into our, our approved set of parts. We just kind of meet you where you are instead.

Chris Gammell: So then you're almost a quasi distributor then you're like a, you're like a way station style, like distributor without, uh, yeah, pretending, pretending. Sorry. I just want to make a joke about Avnet and arrow being a $50 billion grocery store. Cause they don't really even hold any stock. Yeah. Sorry, our own Avnet. I don't know why you're worth so much. You're a grocery store. Actually, they're a bank. That's what they really are. They're a bank that, uh, that helps to, uh, distribute based on cashflow. So that is, that actually is valuable, but I get it.

Andrew Seddon: Yeah. The, the, the different tiers of distribution, I think kind of have different functions. Like definitely once you go to Avnet and arrow and, um, you know, I got, I got a lot of appreciation for those guys. They, they dealt with me when I was, uh, when I was 12, 13 years old and ringing them up and telling them that I was gonna, I had a demand for 10, 10, 10,000 units of some $20 IC just so I could get a couple of samples out of them. And, you know, they'd always deal with me then. So I got a lot of appreciation for them, but, but yeah, they, they, uh, yeah, primarily act as banks, but then you go to like, uh, you know, like a digi key amounts of these catalog distribution companies and they are holding significant amounts of stock. Right.

Chris Gammell: And you'll, you'll do like a warehouse drama, right? It's like that. That's more like kind of more like a Costco style, like just to put it, you know, put it in other terms, right? A Costco or a Sam, Sam's club sort of thing. Yeah.

Andrew Seddon: It's a similar sort of thing. As far as I know, you can't actually go and walk around the warehouse and pick, pick parts, uh, off the, off the shelves. But I mean, that, that would be cool if, uh, if, if you were able to, to do that. Um, yeah, so, so yeah, I mean, you know, we, we don't, we don't sell parts, uh, just on their own as parts, like everything we, um, the only thing we really actually sell is assembled circuit boards. So we don't sell, uh, just, uh, bare circuit boards. We don't sell parts. We, um, sort of insist on, on selling the fully assembled, uh, boards and, you know, so it's a little bit of a distribution model.

Chris Gammell: So the, the, the reason to have the 50,000 skews like in-house though, is that for, uh, speed? Is that for margin margin on those parts? Probably not. I'm guessing that's probably not it, but, or is it because they're preloaded on a machines?

Speaker ?: Yeah.

Chris Gammell: There was, there was, there was, now from a digit gear mouser sort of thing. Sorry.

Andrew Seddon: Yeah. Well, it, it's mostly to get them preloaded on the machine. So. Okay. All right. The, you'll find like, especially in prototyping, the, uh, most of the cost of the parts in, in low volume is actually on the assembly side, right? Like it's the dealing with the logistics of the parts, not necessarily. Yeah. The actual cost of the parts themselves. Um, you know, did you, did you, did you, is, uh, and, and, and some of these other distributors are, uh, they're really fast. They're really consistent. Um, you know, they, they do a great job, but then the parts arrive at the factory and you then have to, you know, go through this pretty laborious process of, of receiving them and, and putting them at some sort of, you know, way station loading them onto machines, taking them off. And the big one actually is you have to, you have to deal with, uh, attrition. So. Right. Like, uh, I was chatting with Chris Denny about this the other day, actually. And we've, we've kind of come to a, uh, kind of good way to, to explain, uh, attrition to people. Cause we often get asked this, like, why do you need to order 5%, 10% extra parts for the assembly process? Right. Cause the, the, the intuition is why, why are you losing, uh, why are you losing these parts during the assembly process? And if you go in and you watch it, you, you typically are not losing parts. It's not like parts are spraying all over the place and you're guaranteed to lose, lose 10%. Oh my eyes. Well, it does sometimes happen. Yeah. Yeah. But it's relatively, relatively rare. Um, so what, what, what, what's actually happening there is that you, if you have an order that you're trying to build and you are missing a single part, you can't ship that order. Right. So you, you, you may be missing a, a part that costs 25 cents and you've now got a three, $4,000 order that you can't ship out the door. You can't close out. And you may have to kind of like keep set up on the machines. You've got to keep it in your database. It's like actively being built. And so the cost of like being short of a part is super, super high. And so really what the attrition is doing for you is lowering the probability that you're going to run out of parts. And so as a result for every order, every prototype order that's placed, you might be able to run out of parts. You might have to order another one, $200 worth of parts to make sure that you've got enough. Should, should you lose one of those. And so one of the advantages of our model is because we've got all the parts in stock anyway. we're able to make up for that attrition just by virtue of the fact that we have parts in stock. So that's a pretty big benefit. And then you're not having to receive those parts on demand for every order. So you can place bulk orders. Despite what I said about there not being this common library of 2,000 parts, there is a lot of commonality amongst different customers and especially within companies as well. If you have a bunch of design engineers operating within a given company, there's a lot of commonality amongst parts. And so you can do some batching on the incoming material side of things. It just makes your overall operation that much more efficient.

Chris Gammell: That's great. Yeah. I think back to the shortage times 22, 23, and there was no better time to point out that it's just like, yeah, if that one part's not there, it sits, it waits. And we had stories on the show too of people that were like, yeah, I have an FPGA on my board. I'm waiting eight months or whatever it is, whatever those ridiculous lead times were. And yeah, there is no replacement. There is no nothing out there. It's interesting too, thinking about like, it's almost like a physical manifestation of cashflow. Like you can literally see like the board is missing that one thing. So that chunk of cash from circuit hub's perspective, right? You can see it as like that cash is sitting on the line, you know, it's just like, it's, it's a stressful, I'm sure. Very stressful.

Andrew Seddon: Yeah. I mean, you definitely want to be clearing these things, you know, through, through as quickly as you, as you can. Um, and that's our goal. So, I mean, we, we actually ship now about, uh, it's about 81% of our orders that are ordered on our fastest lead time ship within about three days. So that's, that's from, you know, the files are uploaded, the buttons clicked to place the order to us shipping out the door. Uh, that happens within three days. So that's the fab, the, uh, the assembly, um, and everything that's required to get those, uh, get those orders out. Plus you get to see, uh, you get to see the progress of the order as it's happening. We give you a bunch of like cool images now, like AOI images of, uh, as, as the production process is happening. And so you can kind of see what's going on. So we found that that's like one of the biggest, uh, things is like, as a design engineer, you're, you're often working on, you're not generally just working on that single circuit board, right? Like you have a whole system that you're working on around it might be a bunch of different boards. There's like inevitably going to be a ton of software involved. And so just having that visibility in when you can expect those boards to be back, uh, is like, is enormously useful. Um, and so we've recently deployed a bunch of features around, you know, just having that, that visibility, uh, available to, to engineers. So, you know, just to your point, the, um, for us, it's less of a problem because we have this like three day turnaround. It's so fast. We, we don't really have that, you know, that, that cash sort of sitting on the books for, for too long, but yeah, certainly for a conventional electronics manufacturer, you, you might have two, three month lead times on, uh, you know, on, on, on orders, uh, sometimes even longer, you know, I've seen six, 12 months delivery times on, you know, if you've got a really complex FPGA and ordering a thousand of those things, and it's the three, $400,000 order, often they're actually just made on demand, right? Like, so you, you place your order with the distributor and then the distributor places the order with the semiconductor fab, and then it takes six months to make them because it's some really complex process tech, and then they get shipped out to you. Right. And so in that intervening time that all of the rest of the components have been accumulating at your EMS provider. And so that is, you know, that's cash sitting on the, on the balance sheet. And so, you know, these, these companies, the, the distribution companies like Arrow and Avnet really help you to, to manage that, that cashflow situation. But yeah, fortunately the vast majority of what we do is, uh, yeah, it's kind of, it's really quick turn, small batch. And so we're able to get those orders in and ship them out, uh, kind of as quickly as we can. We actually did a two, two day, uh, we had our first two day last year and we're starting to do those reasonably regularly now. So, you know, the order comes in on a, on a Monday, um, and then we'll get them, we'll get those shipped out on a Wednesday. Hmm. That's great. Yeah. Uh, before, I, I think

Chris Gammell: that that's a good lead into customers and types of boards coming through and stuff like that. But one thing that was popping up in my head is, um, you know, you mentioned kind of like how the machines are set up and how you have these kind of preloaded and stuff like that. I remember the first time I saw, I think a Neoden, which obviously you're not using, right? You're using big fancy equipment. But the first time I saw Neoden, it's like one thing that clicked in my head is, you know, you have the, the, the little conveyor belt loader thing and just a lower, lower cost, right? It doesn't have to be that low cost, but if you had enough of that and you had a long enough factory, you could literally load every part in the universe onto that. And you just have this thing shooting through. And like, that would be a best case scenario. It's very unrealistic, but how, how closely does the circuit hub, you know, in-house model match that? Like how, how many machines are loaded up

Andrew Seddon: with stuff? Yeah. So, so that was our original thesis actually was exactly that, right? Like we were thinking, okay, we're going to go out, you know, we'll drive everyone to these 2000 common parts and we'll just buy like 25, 30 pick and place machines. We'll load them up with everything that you need. You know, and, and it'll be great. And it'll be, this thing will be like super efficient. But because the, the number of parts is like so vast yeah. It's just incredibly difficult to do that. And so the way our equipment works today is sort of quite, quite different to that really. And it's more about kind of like mass storage of, of individual parts, but kind of disconnected from the pick and place process itself. Right. So, you know, you need, you need the ability to do kind of freeform movement of materials and work pieces. Because pick and place is a part of the process, but it's not, you know, there's, there's a lot else. Material handling is a huge thing. And also like

Chris Gammell: lining up orders and just, yeah. It's a, and, and, you know, a production system nightmare, right?

Andrew Seddon: Yeah. It's, it's kind of, uh, yeah, surprisingly complex, like to, to, to get balls out the door. Like the way I sort of think about it is, you know, for, even for a prototype order, you're typically going to have about one or 2000, uh, fits and for the order to be a success, every one of those has to go perfectly. Right. Like it's gotta be in exactly the right place. It's gotta be in exactly the right orientation. It's gotta be exactly the right part. And if you make a single mistake in that, uh, process, you know, you get one dry solder joint, then, you know, the customer is not going to be happy. And so the, the, the, the level of perfection that's required is really high. And that's, that's what makes the, the, the process difficult. And so SMT is one aspect of that. Uh, but you know, you have, you have through hole parts, you have reflow is just in itself to, to do that reliably for, uh, for a design that you've never seen before is, is, is a pretty, pretty complex thing to do. And you have selective soldering. You often have some hand assembly that's going to be required as well. And then inspections like a whole, uh, a whole sort of big topic on, uh, by itself. And so, you know, you, the, the, the other interesting thing about manufacturing is you, you, you can't generally just solve like one individual process step. You have to solve for the whole system. And so we, we've seen like a lot of point solutions in the electronics industry, uh, before, right? Like somebody will come up with, you know, some way of like storing parts, or they'll come up with a slightly better pick and place machine or a slightly cheaper pick and place machine, or like some piece of VRP software or something like that. Um, and the problem is if you make one step slightly better, it tends to not actually affect the overall process that much. And we, we've had this ourselves where, you know, we, we've saved made SMT like 10 X more efficient. And then you look at the whole overall system and it's not actually that much better. Right. And so you, you have to, you have to be constantly iterating on like every aspect of the process to, uh, to actually make it, uh, make it better. And so, you know, I guess to, to kind of answer, answer the, answer the question, you, um, you, you could buy a bunch of Neodens and put, put them in a line, but, uh, yeah, I think, I think

Chris Gammell: you'll find that it doesn't actually solve, uh, it doesn't solve too much. I have no plans to do that. Just so it's just, just where my head went, you know, I don't, I don't actually act on these things. Um, but it, it is interesting. It was one thing you were saying there and just like this whole solving for the whole equation. Um, one thing that I've noticed, uh, domestically here in the U S at least is like, uh, board people owning board houses ain't getting any younger. And so are you looking at deploying this model on more places as kind of these board houses start to. Yeah. I mean,

Andrew Seddon: you know, it's, it's a huge problem in electronics. It's a huge problem in.

Chris Gammell: CNC it's a huge problem in injection molding. It's like the whole video about, uh, making a wire brush

Andrew Seddon: in the U S. Did you see that one? I haven't seen that. I've seen the toaster one. Was, was, was kind of interesting. And I've, I've got a, uh, I did a tweet about this the other day, actually. There's a, um, cause it was quite, I felt, I don't, did you see the toaster one at all?

Chris Gammell: I don't think I did. Yeah. He was trying to make a domestic toaster. That was the idea.

Andrew Seddon: Yeah. I was trying to make a toaster in the, in the U S and, um, I felt the video was kind of a bit defeatist and there's, there's example over in the, in the UK, there's, uh, as a company that makes these, these Henry Hoovers, these kind of like quite cute, cute looking hoovers with a face on them. Like if you're, if you, if you grow up in the UK, like you see, you've seen Henry Hoovers like all your life. And so it's like, it's pretty familiar thing, but they, they, they call vacuums, Hoovers, right? Uh, vacuums. Yes. Yeah. Yeah. Right. So, um, and so about half of the market is these, is these Henry Hoovers and they're actually made in the UK. So, uh, you know, it's, it's like, you, you can do these things. You, you can, you can make locally, but you kind of have to, you know, really put your mind to it and apply, um, apply automation, specialized, optimized. Yeah. That's it. I mean, the, you know, these guys have been at it for, for many, uh, for many years now. And so they, they, they really sort of optimized the process and it employs a bunch of people too, but you know, super successful company. So, so it can be done, but you know, the problem is a lot of that, that manufacturing base has been, uh, hollowed out over the last, um, you know, 30, 40 years. And, and so these haven't necessarily been particularly attractive businesses for people, you know, for the new generations to get into. And so, yeah, you do find like a lot of these, uh, a lot of these, these shops now, both in electronics and, um, mechanical parts manufacturing, you know, they are owned by people that are kind of looking to retire pretty soon. And, you know, they might consider some sales to, to private equity. There's been a, uh, there was a quite a lot of like roll up, uh, roll ups going on over the last couple of years in, in electronics manufacturing and PCB fab for instance. Um, but yeah, it's a huge problem, right? Like we need this, we need this capacity, uh, locally, because if you, if you lose the ability to prototype in your home country, you lose the ability to do rapid product iteration. Right. And a lot of the, the, the innovation on the product side actually comes from, from the manufacturing side of things. And I think it's this, um, you know, it's, this is, it's kind of false reasoning that you can, you can chuck the product over the wall and get it manufactured by somebody else and, uh, and, and still be as good on the product design side of things. Like you, you have to be, have, you have to have a deep understanding of the manufacturing process, because often as you're making these, these products, you'll spot something that happens in manufacturing that you then feed back into the, um, into, into the product development. And so, you know, I, I'm, I'm a big proponent of just seeing more manufacturing next to design. Um, you know, there's been some, some really good examples of this recently, actually the, uh, especially in electronics manufacturing for the, uh, start the SpaceX Starlink, uh, manufacturing site down in, in Texas. Um, they've actually spun up a fully vertically integrated facility that's producing the Starlink base stations. And so they, they do everything from, uh, PCB fabrications. So they're making the bare PCBs. They're doing all of the surface mount, they're doing functional tests that they're doing the injection molding, CNC box build assembly. And then they're, they're shipping about a thousand finished Starlinks out of, out of that factory every, uh, every day, as I understand it. And so

Chris Gammell: that's the consumer side or the actual ones going up over. That's the consumer side. Yeah. I don't

Andrew Seddon: know so much about the space side of things, but I would assume that that is also done in the U S

Chris Gammell: but it's obviously much lower, um, lower volume, low volume and high, high value, high test. Yeah.

Andrew Seddon: Yeah. Yeah. Yeah. I mean, a lot, a lot of that space stuff has been, uh, you know, that has been continued to be manufactured in the U S but the, the, these base stations are exactly the type of thing that you would, you would have a few years ago sent over to, uh, over to China. And so that's awesome to see those being made in the U S. Yeah. Yeah, definitely. I think, you know, uh, you know,

Chris Gammell: 800 pound gorilla in the room is tariffs, U S domestic manufacturing. I think a lot of it is tied to space and defense and things like it. I think there's the prototyping aspect, no doubt, but like some of it is just, again, I'm very U S centric. You work in the U S a lot. And like, there's just a mandate right now. And, and there is not a lot of options too, right? Like, so my, one of my buddy, one of my good buddies is a defense manufacturing person. He's been in that for a long time. And he's just like, do you know how few PCB houses there are remaining? And she's just like, they're all buying each other. And I have to bid out to two or three of them in any job. And it's just bidding out to the same three, you know, it's just like, Oh, that doesn't sound great. And then the lead times for that and that

Andrew Seddon: sort of thing too. It's just, it's tough. It's insane. I think there's, there's about 50 individual PCB fabs left in the U S right now. And, and quite a few of those are getting rolled up under, uh, you know, and under, uh, sort of umbrella companies. And, and so, um, I don't have the precise numbers, but I feel like that was like a couple of thousand. If you, if you roll the clock back 30 years, right. Like there was this whole cottage industry of PCB fabs and some of them scaled up and

Chris Gammell: but that's just tech used to be simpler too. Right. I mean, it was like, you know, we're going to do a photo shot and we're going to do a two sided board and like, Oh, look, we have through holes now. Like, okay. You know, like we can backplate things like, Oh wow. Yeah. Yeah. I mean, that's

Andrew Seddon: some of my fondest memories of kind of make making, uh, making, making circuit boards in, uh, in my parents' parents' home. So you can definitely do it. You can still do it today. It's just, yeah, the technology has shrunk and it's a lot harder to do now than, uh, it used to be with kind of through hole, uh, through whole boards. And so, especially a lot of the cutting edge things, like if you're, if you're looking to do a board with a complex FPGA on it, I mean, it's not quite semiconductor levels of precision, but you know, you are approaching kind of clean rooms and you know, some pretty, pretty, pretty high, high spec. I mean, you'll, you'll know this having, having worked in a fab. So, um, yeah, you know, it's sad to see that, that go away. I think there are, there's definite tailwinds for, for, for bringing a little bit back. I think people are, you know, particularly on the prototyping side, just really recognizing that we have to maintain, uh, the ability to, to prototype all sorts of, of different products. Right. And, and, uh, you know, so I think we, we have seen certainly on the circuit hub side, essentially unlimited demand. So the challenge for us has always been space. Yeah. The challenge for us has always been scaling capacity. Right. So we, as a company have, um, we've had brief, I think the only time we were ever really, uh, limited by demand was the early days of COVID back in 2020. Right. You know, there's the whole world. Yeah. Everyone's like clamming up and yeah, I'm just, I'm not doing anything. Let's see what happens. And then it went, and then the back end of the year, it went absolutely crazy. And we go back to being kind of like, uh, capacity limited again. And, and, pretty much, you know, it's been that way throughout the life of the company. And, you know, there's, there's this almost insatiable demand for a high quality manufacturing service, right? Like you're a design engineer and we, we, we work with a bunch of, um, you know, we design our own equipment, equipment internally. And so we have a pretty varied supplier base of people making, you know, CNC machine parts and, um, bits and bits and pieces for us. And so, you know, we tend to, uh, just finding great vendors is, is hard. And then when you do find a great vendor, you tend to find they run out of capacity because everybody flocks to that vendor, right? Like if everybody else kind of like finds, finds this, this good service and then, and then they want to do it.

Chris Gammell: Do you want to tell us your best vendors right now? Live on AMS? Do you know, I've learned that that's a bit of a, yeah. You know, you gotta be careful, right? You

Andrew Seddon: gotta be like, Ooh, this is careful. Yeah. It's not even a competitive advantage thing. Cause I, I generally, I would share, but it's like, you need your work doing right. And so you, you, you,

Chris Gammell: I know it's weird. It's like weird duality, you know, it's like, Oh, I want to like promote their work and whatever, but I also, you know, stay out of my hair, please. You know, I need the,

Andrew Seddon: Oh, I'm first in line. I recommended them. Yeah. That's, that's it. And, uh, so yeah, it's, it's an interesting dynamic, you know, and if you're able to, to scale one of these manufacturers, then the demands there, it's just incredibly hard to do, right? Like there's, there's so many moving parts in the manufacturing business. And, um, you know, we, we originally thought with circuit hub that we would be able to do it as pure software. I was very aware of how hard it is to do hardware having made a ton of devices over the years. And so I thought, okay, let's, let's do something easy in the, in the software world. But, um, yeah, as soon as you get into manufacturing, like it becomes very apparent why things are difficult and, uh, but it's fun.

Chris Gammell: I do think like you said, that is competitive advantage now though, too, like, because people are like, you know, like, you know, VC for a long time would only fund software sort of things. But I think gradually they're like, Oh, actually there's like a moat here. It's like, no, it's not just like someone's going to spin it up with an LLM and be like, Oh, I've created a new circuit hub with my, you know, Claude code sort of thing. It's like, no, no, no. You need to have some deep hooks and make it all in it and actually have hands-on equipment, that sort of thing.

Andrew Seddon: Yeah. I think there's, there's a lot of debate in the VC industry right now, uh, about how defensible some of these new AI companies are, right? Like you're, you're seeing some just incredible, uh, growth rates in, in some of these AI companies. You're like, there's this massive, ridiculous

Chris Gammell: valuations and like stupid money. And maybe I guess I'll say, you have to deal with VCs more.

Andrew Seddon: Yeah. They're making, it can look that way. So some of them are, yeah, there's no denying some, some of them are just are crazy bets. Right. But like in, in general, like most VCs that I've met are, are, are, you know, they're, they're very smart, but they're, they're playing a very specific game. Right. And that, and that game is making these highly symmetric bet asymmetric bets on things that have a small chance of being very big. And, and that is not the conventional business world where you're trying to scale a business, you're trying to make it profitable. And so, you know, they, they look at these AI companies and, and, and, you know, the, the whole VC industry is like very consumed with AI companies right now. And like, some of them are just like objectively like growing incredibly quickly. And the, you know, there's this, there's this massive build out that's happening of, uh, of AI infrastructure right now, right? Like it's being deployed into, to all of these, these applications. So, you know, the, the, the production of tokens from these LLMs is kind of becoming a commodity and those tokens are being funneled into all of these various, uh, different, different industries. And, and because there's such a big, uh, sort of disconnect between the capabilities of the underlying models and the applications that they've been deployed into. And I get reminded of this on a, on a daily, daily basis. I, you know, I, I chat to, you know, friends outside the technology industry, outside the electronics industry. And, you know, I, I sort of, uh, I, I show them things that are like, that I was doing a year ago with, with LLMs and, you know, what's possible. And like, they're surprised by it. Right. And it reminds me like how father, how you, you and I kind of live on, on the cutting edge of technology because we're interested in it and it's our profession, but the vast majority of people don't do that. And so there's this massive build out that's happening right now. And if you can, if, uh, you know, and some of these companies are directly in the path of that. And so if you're, you're kind of a company that's working on infrastructure for scaling LLMs or, you know, some, uh, uh, training data for LLMs right now, there's been some big, big acquisitions in that space. Um, you just kind of benefit from this, this, this incredibly rapid expansion of the, the whole industry. But then the question is like, how defensible are some of these businesses and where, where will they be in five to 10 years from now? And, and the fear there is, well, if these underlying models are improving at such a rapid rate, will they overtake the products that are built on top of them? Right. And so this is, this is where the whole chat GPT rapper, you know, meme comes from. And, and so that's the debate right now. And that, and that's the risk. Like you look at a company that's growing very rapidly and they've been able to take advantage of a, an area of the economy where, uh, LLMs haven't yet been widely deployed, they can rapidly grow, but then will the models themselves come and, come and catch up behind them. And so debatable how defensible some of those businesses are, but if you're able to layer on hardware on top of that, and, and people, you know, think about this as embodied AI robotics, uh, AI into manufacturing, you know, anything where AI, AI starts to touch the real world, you know, those businesses start to become much more defensible because they have a component to them that takes time to build out, right? Like there's a, there's a natural limit to how quickly you can, you can build out physical, uh, physical infrastructure. So, you know, we had this back in the, in the pandemic with, uh, semiconductor fabs, you know, this was a, that was a big story around the time, like, Hey, we need, we need more semiconductor fabs. And everyone in the industry was like, yeah, we know, uh, come, come back to us in six years when we, when we built it. Right. And it's just like a, a natural limit to how quickly you can build these. And, and so, yeah, you know, expensive these things are, these are, they're expensive, but like, interestingly, you can't necessarily chuck more capital at them to get it done faster. Right. Like you can to an extent, but there's just like a natural limit to you, how quickly humans can coordinate to build something in the physical world. Now, if you ever had any construction done, just, just in your,

Chris Gammell: yeah. Right. Right. Right. It's like, you think it's gonna 10 X. He's like, well, I'll get do it when the board, then the, you know, the wood shows up, I guess. Right. Yeah. That's just like,

Andrew Seddon: there's 500 dependencies. Like the right people have to be in the right place at the right time. The materials have to show up. And it's like, you know, you, you start these projects and it, it's, it goes really quickly. At least this is my experience of kind of construction in my home is the, the first two or three weeks go really quickly. And then everything cries to a halt. And you know, that, that dynamic plays out, uh, in the large scale as well, when you're trying to sort of spin up these, these, uh, these, uh, these big facilities. And so, so yeah, you know, my, my thesis is certainly that there is this much more defensible dynamic to applying, uh, AI into, into hardware. And my hope with that is that, um, yeah, it's gonna lead to a bit of a renaissance, not just in manufacturing, but just in, in kind of electronics in general, and the ability to make electronics, uh, product, uh, companies, um, that, uh, you know, are, are sort of venture,

Chris Gammell: uh, venture backable. What I just heard you say is, uh, circuit hub is going to have an MCP soon. Is that, is that correct? Uh, what I'm hearing. So there's a model contract protocol. My LLM just says, Hey, just buy this from circuit hub. It'll be fine. You know, make me a board and order from circuit.

Andrew Seddon: Yeah. Yeah. I, I, I think that would be a really interesting thing to do. I mean, everyone, everyone's going to be right. Like, I mean, I, I, I think, uh, I'm seeing some really cool stuff happen in the electronics side on, on the design side of things now, like people are coming up with tool, tools to, uh, create, you know, circuit board designs more easily. I mean, obviously key CAD has, um, you know, been awesome in, in, in, in that, in that space over the last few years, we, we've seen a really big uptake in, in people using open source, uh, design tools. I know you've, you've been a big part of that, that movement, Chris, and, um, that's a much more, but yes, I I'm still a big fan. How about that? Yeah, for sure. Well, you know, I think a lot of that content lives on right. And that, and that enthusiasm. And so, um, and, but I think it's also, it's just a function of like the tool, the tools are getting a lot better, right? Like if you, if you look at EDA tools, that the, the, the, the sort of, um, pace of improvement of the commercial tools has tapered off because there's only, only so many useful directions that you can, that you can go in with any, with an EDA tool. And so, uh, I guess to a certain extent, the open source tools have caught up, but, uh, so cut, yeah, you know, key CAD is, is pretty comparable to, uh, to an Altium or, or cadence or other commercial tools these days, uh, certainly for a particular domain. Yeah. I think low to mid end for sure. Like,

Chris Gammell: no, no, no doubt about that. I think high end, there's still definitely like differentiation and, you know, uh, workflow tools that are out there, but I think most people don't get there. Right. And especially the ones that are going to be the, the kind of the broad base users.

Andrew Seddon: Yeah, that's it. And, and then, yeah, what happens when you chuck AI in the mix, I think it's going to be super interesting, right? Like I've seen a few companies working on these co-pilots for, for design tools. Like I feel like just having used the coding tools that it's like a pretty tractable problem, at least to get to an initial design. Right. Like I, I always felt like in electronics design, the, the bulk of the work that I did was more systems level design. It's kind of coming up with a high level, which parts am I going to use? What am I not going to use and why? And then the

Chris Gammell: software side of things. I'm going to put out into the universe, the thing that I've been saying for, I'm going to do it again. Right. The, so like the AI tool, I feel like a lot of them are like, oh, I want to like do your, you know, your, your traces for you, or I want to do, you know, I'm going to auto route everything, or I'm going to do your controlled impedance, everything. You know, I'll do this perfectly. I'm like, that is not the hard part. That is a hard part. That is a hard part, but it's kind of like what you were talking about with the, the, the system, like making that 10 X better does not make the board 10 X faster. Right. Yeah. To me, the thing, the most impactful thing to me is like, pick all my caps, pick all my resistors and don't make me think about it. Right. I'm going to put in a 10 K 0 4 0 2, figure it out. I don't want to think about it. Right. Like that is the kind of level that I want to be at. I want a tool that's just says like, give me the specs. I'll send you back the stuff. I'm going to give you a report that shows you how many are out there, that sort of thing. Like stuff that I think circuit hub actually does on the sourcing side as well, but I still have to put in that it's a, you know, R 0 4 0 2, you know, whatever the Yagi-O number is. Right. I don't, I don't, I can't just put in 10 K

Andrew Seddon: a day and I want to, I want to, I want to be lazy. I think you might be in luck is my, is my guess. Like that, that feels pretty feasible, even with the, the, the level of where LLMs are right now. Feasible, but nobody's making that tool. That's what I'm saying. It's like,

Chris Gammell: nobody's making that tool. And I want that tool. I want to like buy that tool. Everyone's like, Oh, I want to, you know, I feel like the, the disconnect is, you know, fancy, you know, thing gets funded, boring, important thing is like left to like, Oh, you could figure that out yourself. I don't

Andrew Seddon: want to figure it out myself. Yeah. That's a, that's an interesting, well, you know, he cards all open source circuit hub should do this. Well, I think, you know, where, where we'll, we will probably play in that ecosystem is, um, yeah, you know, we, we, we do want to expose circuit hub to our genetic systems. Exactly. What form that's going to take is a little bit, you know, TBD, I think whether that's MCP or whatever comes next, but you know, the general idea of, you know, right now, all of our customers are, are people, it's people clicking on the website largely. And you know, placing orders themselves, but you know, you, you fast forward that a couple of years, and I think you are going to have these agendic systems doing a lot more, uh, on their own. And so I think that that share of interaction with circuit hub between humans and AI is going to, it's going to increasingly shift to AI. And so, you know, we may do things like provide access to our, to our inventory via MCP, uh, the ability to place orders, query, uh, status of, uh, what's going on with your order, that type of things. I don't want to speak too much for our, for our product team and kind of roadmap on those, but you know, it's all pretty logical. And I think a lot of, a lot of companies are starting to look, look at this. I mean, we actually do it a little bit internally. Now, if you, if you go to our site, we have a kind of support bot that, that pops up, you know, you click help and you can have a chat with the support team. And, and that is, um, the frontline support there now is via an AI, uh, chatbot that is hooked up to our internal systems and has access to information about your particular orders. So you can ask questions like, Hey, what's the status of, uh, of my order. And that is like able to tell you what's going on with that. And so,

Chris Gammell: um, I should go ask for your API key from the, from the chat bot and see how,

Andrew Seddon: see how it, uh, secured. Yeah. Yeah. If you want to go, give us a sort of pen test, then, uh, yeah, yeah, make sure it's all okay. I think, um, you know, there's, there's definite, you know, security, uh, issues that you have to overcome as you start to expose more of this, this, this data publicly. But, um, yeah, I feel, I feel that's the way things are going.

Chris Gammell: That did get to one kind of topic that I want to talk, talk about with, um, you know, just quick turn manufacturing generally. Like one thing that I've always been frustrated about, you know, I'm sending a, say a high, not a high spec, but like an edge of the spec board to like a JLC or something like that. I send it out 12 hours later, you know, I'm going to bed and they're like, Nope, that's wrong. And then I am like, okay, I'll deal with it tomorrow and submit it again. 12 hours later. Nope, that's wrong. I'm falling asleep. You know, like, just like that back and forth and back and forth and back and forth. So, you know, DRC, ERC stuff that I should be doing on my computer. That's all there. I should have been better on that side of things. But then also, like, I feel like intelligent checking and just the quick turn process requires that like direct, fast feedback sort of thing. So what, what is there? Like, so I click submit on a circuit over order. How are you guys checking? How are you telling me I'm an idiot? Because believe me, I am. Chris, you said 10K 0402. That's not enough. You got to tell us more.

Andrew Seddon: You know, so we, so we did, we did quite a bit of like verification before, uh, you know, before the order goes into manufacturing, right? So there's various different aspects to that. There's the, the bare PCB fabrication. So we run that through a whole suite of tools to, to verify their design, check it against the netlist, make sure it's physically can be manufactured. If there's any, uh, complexities on the stack up side of things, make sure that that's feasible to, to manufacture. And so we've automated a bunch of, of that checking process, but there's, there's still humans involved. And I think always will be as you, especially as you start to push the envelope on, uh, what's possible, right? So if it's a simple two four layer board, that stuff generally is not touched by a human, but then as you, you start to push up in complexity, you start to, uh, push what's possible in terms of vias and, and track gap, then, uh, often the, the kind of the, the combinatorial complexity of that, when you also can consider the various different stackups you can have is just enormous. And so right now it does require a human to, uh, to look at that. And we will, uh, often write back to you and kind of give you feedback in, uh, the same way that most other suppliers would. We're just generally able to be quite a bit faster because we've, we've automated, um, more of the, more of the process. My, my sort of feeling long-term, I think most of that should live in the design tools themselves. Like it's not really been an ambition of ours to kind of be a DFM company as such. Like, cause I think you, you want, you want most of those design constraints in the tool ultimately. And you get this in, uh, in, in, in, you know, if you're designing a chip, then you get these, uh, these PDKs are released by the, uh, by the fab. And they, they have a really complex, uh, set of, of design rules, which, uh, you know, isn't a spreadsheet. It's, uh, it's more like code really that allows you to, how do we get that? Does that

Chris Gammell: exist that I'm not, that I'm missing it? Like, are there, so like, I've never seen like a published thing that I could plug into like a key CAD checker. It's always like, oh, here's our specs, put those in your tool. It's not like, yeah, put them into key cat. You know,

Andrew Seddon: I think I I've seen various companies try to do this over the years. I think some of the complexity there is that in the semiconductor world, there's a very small number of semiconductor fabs. And there's a very small number of, of tool vendors. And there's a very small number of people that are designing chips. Right. And so they, they all go to the same conferences and they're very well connected and they have the ability to, to talk, not just through the PDK itself, but they, you know, their teams are very intertwined. And, and when you look at the PCB industry, it's not really like that, right?

Chris Gammell: But you have, um, Well, we have Gerber's. What are you talking about? Come on, man.

Andrew Seddon: Hey, I'm a big fan of Gerber's. I think Gerber's and spreadsheets. What else could you need? And so, so, you know, I can see a world where that that's possible. And I think, especially with, you know, some of the, these AI tools that are coming online, it makes, um, AI is actually really good at doing things like defining rule sets. So a lot of the complexity and defining rule sets in, in, in, in code is to, is to do it well. It's just like a really hard task. And LLMs allow you to do that in English language. Right. So to some extent they act as constraint solvers where you, where you input the constraints in, in, in English language. So I'm quite hopeful that these tools are that, that kind of checking is going to increasingly be, be pushed down into the tool. Um, for, for the time being, we, we certainly do a lot of it. We do, you know, not just the fab side we do on the, you know,

Chris Gammell: checking parts availability. So what I'm hearing here is the amp hour should be defining a new standard that all of these tools adapt. We need a third party, a neutral third party that gets a generous, uh, licensing fee for doing this. Is that what I'm hearing? Well, you know, the electronics

Andrew Seddon: industry is littered with people that have created, created new, new standards and they've never gone in the US. So I, I'm definitely, I'm definitely a fan of like, just make a single thing that works and then, you know, open source it and let other people copy that because like the standards-based approach is like, it's really hard to, to get anybody to kind of integrate that. And it never

Chris Gammell: quite works the way that you think. Let's get like a hundred grumpy old people with a financial vested interest in a room and then have them shout each other for four days and then nothing comes out of it. Yeah. Yeah. Yeah. And then there's been, eventually something comes out of it. Yeah. Yeah.

Andrew Seddon: I think, you know, there's, there's some foundational stuff where that, that's really important. If you're, if you're defining specifications for, you know, the real bottom of the stack and the hardware level, like RF layers, uh, you know, USB and things like that, like that probably is what you want to do. But if you're defining a software standard, often it's actually better to just start with the implementation and then, uh, and then let people kind of use the code as the, as the reference instead of trying to come up with this 500 page document. So, you know, I, I think with open source, you, there is the potential now for, uh, you know, for the key CAD team to do something like that. Um, I I'd love to see it. I'd definitely support that. Um, and, and be open to sort of thinking about ways that we could, we could help them to do it. Yeah, that's great. I, I should, I should,

Chris Gammell: uh, I should push on that. I'm sure they're already doing it. What am I saying? They're already doing it. I'm sure. So I'm sure somebody is working on it somewhere. Yeah. Someone's doing it. Yeah. Yeah. What about the, um, so we had talked a little bit about like the types of customers. I mean, you have listing of, you know, your kind of high, high level customers on scrolling on your site here. I see Tesla, blue, blue robotics, foreign labs, Barrett, things like that. But the nature of the, you know, is it big teams that are like putting their, you know, kind of internal libraries in so that they, they can, you know, on a, on a quicker turn basis, they just kind of have access to you as an approved vendor. And it's just like, I need to board fast. I'm going to circuit hub is that sort of thing. Or is it more, is it a lot of small people too?

Andrew Seddon: Yeah, it's, it's a real mix. So we, we work with most of the largest companies in the world right now. If you sort of take the, take the 10 largest, largest companies on the planet, we, we work with, um, certainly a majority of those. Um, but they tend to operate as individual design teams, right? So if you, uh, you know, you might have a company of two, 300,000 people, but ultimately you've got an engineer that's working on a, a particular circuit board to do some sort of task. Right. And so it actually looks a lot like dealing with a, with an individual engineer. They're often still like paying via credit card and kind of operating reasonably independently. And you just push up when you see like a John

Chris Gammell: Deere in the email address or a Tesla in the email address, you're like, Oh, hello.

Andrew Seddon: Yeah, it's, it's cool. I'm often surprised these days, honestly, like I'll, I'll, I'll see, see some more to come through from, you know, from a big company. I'm like, cool. Yeah, that's, that's awesome. But you know, we, we've worked with a lot, a lot of these, these companies for so, so long now it's, um, it's, um, it's kind of second, second nature. And so, um, what, what we tend to see is those, um, you know, the, the, that spreads within companies. So we'll, we'll get an initial in someone will give us a, give us a try, you know, it's pretty, pretty low risk to, uh, to send out a prototype. If you're, if you're making 20, 30 prototypes a year at one of these larger companies and you just say, Hey, let's give circuit hub a go. Um, and as long as we do a good job on that, we tend to kind of see these repeat orders. They tell their friends within the company and it kind of spreads within these design groups. And then we'll, we'll go in there and we'll, we'll, we'll kind of talk to them a bit more at the corporate level and say, Hey, you know, you've got 25 engineers that are using us. Maybe we should, they seem to be having a good experience. We're seeing a bunch of good feedback. Uh, maybe we should, you know, have a conversation about rolling this out to, to the rather two, three, 400 engineers, however many it is. And so we've had a bunch of success on, on that side of things. Then, um, then we have, uh, kind of startups and SMEs. So companies, uh, that typically kind of 20 to a couple of, a couple of hundred people. And for those, we tend to do the prototyping and their production as well. So they'll often be doing, you know, somewhere between 500 to a million dollars a year in, uh, in between prototyping and production. They've got like some sort of long tail, uh, product that, uh, you know, is never going to be a consumer volume product. And, uh, you know, we, we've seen a bunch of success in a few different companies and kind of like drone light shows recently. That's, that seems to have like taken off.

Chris Gammell: That's cool. Yeah. Um, yeah, I was really, I looked around at July 4th this year in the U S and I didn't see anything in my area. I'd love to go to one, you know, I still, yeah, I, I still haven't seen

Andrew Seddon: one, but I'm, I'm, I'm really, really excited to, uh, to, to go and see one. They just, they look very cool. So yeah, we, you know, we will serve companies in that space on, on the production side of things. And, um, and then my, my goal ultimately is to kind of take that all the way down to, uh, the students and the, and the hobbyist level, which is my background and, you know, to, to, to, to a large degree, kind of why we started circuit hub really, you know, I, um, trying to build electronics from, from my bedroom as a kid and had a whole bunch of difficulties doing that. And, and, uh, so I'd like to enable, uh, you know, circuit hub to be viable for, um, for people that are just getting their start in electronics. And, uh, so to do that, you know, we really have to look at cost reductions, price reductions, um, where, where we are right now is our kind of entry point into the market has been we want to be the fastest and we want to have the broadest service offering. So we want to be able to make advanced products and we want to be able to make them incredibly quickly. So typically three days, sometimes two days, and then on top of that is then do the price reduction and get them, get the price down. So, you know, you look at where a lot of the hobbyist market is right now. It's, uh, you know, they're typically using these kinds of quick turn shops over in China. So you might order from like a JLC. Um, you're going to get your boards back in 10 days, two weeks. Uh, they're going to be very cheap. The specification is going to be, um, relatively limited. They're obviously, you know, kind of expanding the envelope there, but it's like a pre preset, uh, you know, list of things that you're, that you're able to do. Um, but you can get boards for 25, $30. Um, and so it's very cost effective for a student or a hobbyist to, to do that. And so, you know, as we get increasingly better at this and we optimize our manufacturing process, we're able to bring those, bring that pricing down and we do sort of see a scenario where it's going to be possible to, to serve that, uh, that sort of hobbyist, um, hobbyist and prosumer, uh, market as we, as we continue to, uh, to drop pricing. But, you know, we really have to start at the point of make sure the quality's there, make sure, uh, the speed is there, make sure the variety of designs that we can produce is there. And then once you have that system in place, you can then kind of walk down the, uh, cost and pricing curve. So, you know, I see, I see a scenario where that, um, you know, over time we, we do increase that, um, you know, that, that, that share of market, but right now it's a lot of kind of enterprise customer startups, um, you know, pretty, pretty much most people making stuff, right.

Chris Gammell: They're actually like producing some things for the world, right?

Andrew Seddon: Yeah. You know, we, we've got touch points with most, most funded startups now, uh, tend to, you know, use circuit hub as kind of a first, uh, first port of call. And, um, so yeah, very, very diverse. I'm always fascinated to see the things people are making, you know, like we, we have a lot of orders on the system now, so I'm not often able to, to see what's going through, but every now and again, I'll take a look and, and, uh, I think, yeah, this is super cool. What, you know, new things people are finding to do with electronics.

Chris Gammell: Yeah. Yeah. What about the, uh, so you publish, you know, I'm just looking at the circuit of page design rules. Obviously that's great. That that's what I need for under capabilities, that sort of thing, PCB stack ups. But what about on the, you know, thinking about that, that prosumer hobbyist crowd, that sort of thing, it's almost like, uh, how do I optimize for the cheapest stuff? So you'd mentioned those, you know, those 2000 parts, you guys don't publish those, but just thinking about like, if you really wanted to optimize a design, it's like, you need to have kind of a preset constraints. One of the problems in the electronics industry, from my perspective, getting started is like that you can do anything. And the answer is like, I don't, if I, if you tell me I can do anything, I'll do nothing. Right. So instead, it's like, give me some constrained area there for the lowest cost. Is there any plans to ever like publish that sort of thing? I mean, I know, I know that was kind of the old way of doing things, but, but also that as a way to optimize.

Andrew Seddon: I think that's a way to optimize. We have, um, so on the, on the part selection side of things, like we really don't have any constraints there now, you can kind of use whatever you, whatever you want, like as a general rule there, you know, the less, uh, the fewer, the fewer, the number of different part types that you have, the lower the price is going to be, uh, the fewer, the total number of placements, the fewer, uh, the lower the price is going to be, if you can make it single sided versus double sided. Uh, that's, that's also going to reduce price as well. Um, and then if you can avoid, you know, particularly fine pitch parts, that's, that's kind of, we're going to reduce the price as well, cause that increases or lowers, lowers the yield. But the problem is like, you know, so I can tell you those things, but you know, you, you, the thing that you're trying to design is the thing that you're trying to design. Right. And so you can, there's only so many degrees of freedom that you can, uh, that you can optimize for. And so what we tried to do on the assembly side is, is design a production system that is very compliant and can, uh, is, is highly optimal for any configuration that you can, you know, you can conceivably want

Chris Gammell: to do. So we don't really, you know, we, we're trying to probably over indexing on that idea of like the 50,000 in-house parts, but there's no like price, there's no price benefit to doing that.

Andrew Seddon: Cause that's just a market price sort of thing. Yeah. On the, on the assembly side now, because, because of the way we we've got the factory set up, like that we don't have these kinds of pockets of like design space where, where things are much cheaper. Um, so, so assembly is pretty good, but on the fab side, we, we do, uh, still have some pockets in the design space that are cheaper. Right. And so we have this thing right now that we call proto board, which is, uh, it's a, a two or four layer service, uh, with black solder mask. Um, and if you stick within the specs, it's like a 0.1 millimeter track gap, uh, then you're going to get a much more economic, uh, PCB fabrication. So we have a, we have a really good service right now where, yeah, if you speak, if you stick to this two or four layer black solder mask, you can get boards made in three days. So fully assembled, and you can get that done very, very cost effectively for, for, um, uh, sort of a few hundred dollars. And so, um, um, most of what we sell actually now by order volume, I think it's over about 50, 50% actually fits into that, uh, fits into that service. And, um, so there's a bit of a discoverability problem. We do publicize this, but I think, uh, you know, a lot of first time customers to circuit hub don't necessarily know about this, but if you are, um, if you're a hobbyist with an urgency requirement, it's a really good service, right? Like it's not as cheap as JLC, right? Like if you've got two weeks, it's not gonna, uh, you know, you should probably use JLC, but, uh, if you're, if you're on a tight, tight timeline and you've got three days and you've got a few hundred dollars and you need to get a reasonably complex design produced, like that, that's a really great service.

Chris Gammell: That's killer. And is that like the one I'm just looking at the, the, the capabilities page, that's the ones that are like under instant auto quote, like anything that fits in, in that kind

Andrew Seddon: of bucket is going to be that. Well, it's, it's a kind of, it's a, it's an even narrower, uh, sub-sale.

Chris Gammell: Yeah. So one to 10 layers on the instant auto quote. Okay. So like, so, so the incident also

Andrew Seddon: quote is like one to 10 layers, but then the, this proto board service, which is an even more limited subset inside of that, this, this one to four layers, uh, that is just like really economical. And, and so we do actually show now, if you upload your design, we show, uh, we show whether it's eligible for all this proto board service. And if it is, then you're going to find that you get like a, a really good, uh, short lead time, uh, price, but longterm, my hope is that we, we don't have to have those types of constraints and this special pricing for particular configurations. Right. And so the, the manufacturing process, uh, shouldn't, shouldn't have these kind of anomalies where it's particularly cheap to do something. You should just design like a very general purpose process that is kind of good, good at everything. Um, and then have these very, uh, linear, easy to understand, uh, pricing models. And, and so that, that's the direction we're going with things. I think we're, we're kind of there with assembly now, um, we're working on it on the, on the fab side of things.

Chris Gammell: I don't know if I necessarily agree with that though. Cause like there, there are optimizations in any system, right? Like I, I just think about like, okay, so I had a friend and a former coworker who was working on like working on molds and like there was, it was just cutting metal, right? So it was like CNCing a mold, but there were better ways to like retract pins and, you know, do injection points and stuff. There's always that even in the most like uniform environment of like cutting metal, there's still like better ways to do things. So I think it is just a fact of life, but you're saying just to minimize it. So that is starting from that same, everything's metal sort of thing.

Andrew Seddon: Yeah, definitely. And I think, you know, to some extent this is volume dependent, right? If you're going to make 5 million units or something, you probably want to trade your design time for your production cost. You know, you can, if you can shave 10 cents off, then that's probably, probably worth you spending a few days trying to do that. The reality of the situation is most design engineers in the electronics industry are not working on iPhones. You know, they're working on, you know, some product that's going to be produced a thousand times over. Relay controller. Yeah. So it's like a piece of industrial equipment or medical equipment or something that's just never really going to be produced in volume. And so the correct trade in that circumstance is often not to optimize for manufacturing costs, but, you know, things like accessibility and ability to rework and robustness of the design. And like, there's, there tends to be just like much more important things. And so I think of our job as like, let's, let's make it so that you don't have to worry about the manufacturing process as much as possible. And we can just, you, you, you worry about all the constraints that we can't control and we'll just try and adapt to, to, to what you want to do. And, and obviously, you know, that's just like an ongoing, like tug of tug of war there, right? Because you get, you get advances in what's possible on the manufacturing side of things as well. And so any, any cutting edge process, if you're really trying to push the boundaries of what's possible and sort of do a, you know, do a 24 layer board with, with multiple layers of microvias and things, right? There's going to be design constraints in that. There's, there's physics, physics gets involved at some point. And so, you know, it's just trying to be, you know, trying to, trying to push us as much information as to the designer as we possibly can.

Chris Gammell: Got it. Got it. Yeah. Watch out for those Lucas Henkels of the world, right? And the, each watches. Oh, I love that. I love that guy. Yeah. It's like every day.

Andrew Seddon: I feel like every time I see it.

Chris Gammell: How many, how does he make so many designs? I know he's working with the team too, but like, damn, like, wow. Every time. So cool. Right. Like, yeah. Just, just awesome to look at. Yeah. Yeah. That is, it's eye candy for sure. What about like, so people, you know, you were talking about production. So what is the, what's the off ramp? I mean, you said startups often still go with you, which is great. What is the, now I say, Hey, I've got to get to a hundred thousand units. Where do I go from there? Do you guys shepherd that through to a, to a different

Andrew Seddon: manufacturer or over to the Worthington side? How does that work? Yeah. So we, we, we do some higher volume, quite, quite a bit of higher volume work now. The way we kind of think about it is. So we think about it as though we flatten the, the pricing curve, right? So if you look at like typical electronics manufacturing to make a 10,000 units is, is very efficient on a, on the unit cost basis. And to make one unit is really expensive. And so because we flatten that pricing curve, we enable people to produce smaller batches. And so instead of making a single production run every six months of 10,000 units, we actually just enable you to do smaller production runs on a weekly or a monthly basis and then make that, make that cost effective for you. And so it's going to be

Chris Gammell: good for you guys too. Right. Cause then you're kind of feeding the machine and you have this kind of known business. That's interesting idea. Yeah. It's a, it's a much more kind of, you know,

Andrew Seddon: this is what we, when we say on demand manufacturing, that's, that's really what we mean is to kind of move away from this world of I'm going to make 10,000 units. Like, you know, you go through this, this design cycle where you're making three or four prototype runs over the course of a year. Those prototype runs are, you know, really expensive and they, they, they take a bunch of time. And so you're spending a lot of upfront work on the design side, even just to try and get those prototypes right first time. And then when you move into production, if you're going to push the button on 10,000 units, it's going to cost you half a million, a million dollars. Like you've got to make sure that design is right first time. You know, I've, I've had heart stopping moments where I've, I've got, you know, half a million dollars worth of, of, of product in, in production. And I, I, I spot an issue, right? Like you uncover some issue through writing code for your microcontroller and you realize you've mixed up a couple of I, I squared C lines or whatever it is. Right. No, it's never, no, nobody's ever done that. Never. Yeah. It never happened to me. I promise. RXTX whatever. And so, you know, that, that, that can be, that can be like really scary to, to, to, to, to scale up into those, those production volumes straight away. And so really what we want to see as a world where it's much more gradual, you can just produce a few hundred units, ship those out to customers, scale up from there over time. And so that's really what we're starting to see with our customers is whereas previously, you know, they may have thought about going to a, you know, a dedicated production shop, possibly over in Asia. And they're now starting to see us as a much more viable option for their ongoing production. And so there's obviously a point at which it makes sense. Actually what, what, what we're seeing is for the, for the really high volume work that we are not currently capable of doing, we're actually seeing people start to set that up

Chris Gammell: themselves now. Right. So the, you know, that'll be customer dependent, right? The space X's of the world probably have more resources than, you know, I don't know. I like maybe other manufacturers are doing that themselves, but I, I not heard of it personally, you know, it's, it's still, it's still

Andrew Seddon: pretty niche, right? There's, there's very few companies that are capable of both developing a product and developing a manufacturing process at the same time. Now I definitely, you know, we, we speak to, we've spoken to a couple of companies that have considered doing this and we're just like very open and honest with them about like what's involved because, you know, you look at it and you think, oh, you know, I'm just going to go and buy a couple of SMT machines and it'll be fine. And like, there's a lot more complicated than that. There's a reason these, these things are hard to do, but there is an advantage to, to having a facility that is designed to produce a specific product. Right. And so if you look at high mix manufacturing in the U S about 90% of the cost of goods sold there is labor, right? So if you, if you think it's expensive to get stuff made in the USA, it's because it's very labor intensive because all of these factories are set up to do low volume production across many, many customers. And so if you, if you can set up a factory that just specifically makes this one particular product, it can, it can actually be much more, much more efficient.

Chris Gammell: Yeah. Yeah. I mean, Raspberry Pi does a bunch of their, I think they're partnered with Sony, but I think they do UK, all UK based, or maybe they mostly UK based, I think. Right.

Andrew Seddon: Yeah. I think, I think it's a hundred percent. It's over in, over in, in Wales. They, they, they make those. And I think it's a great example, like I've been able to do high volume and, and they've really gone to town on kind of optimizing that production line for like a single product. Yeah. And so, you know, hopefully at circuit hub, we were able to take that level of automation, but then give it to everybody. Uh, even if you're making one unit, 10 units, a hundred units. And, and so, yeah, we're certainly starting to see more and more higher volume, uh, work and, and that's the direction that we want to go in. Um, but yeah, there's, there's a scale at which it either makes sense to make your own factory or, or, or sort of go to a, a dedicated high volume supplier.

Chris Gammell: Yeah. It is interesting thinking about like, so if, if I did like a, you know, low to mid volume and then I'm like, okay, I need to go high volume now. And I go and hire a Jable or something like that. Right. That's like a mixed domestic overseas, whatever. Right. They don't have just like extra capacity. Like they're going to spin up a new factory too. It's like, they're applying process, but then they're like buying machines and hiring people, but they have the process there as well. So it's like a, kind of like a, you're like buying process from them in a, maybe they'd have some capacity depending on size, but like you said, at some point, if you can learn, it's probably some benefits to bringing it in house and learning that process. But if you can kind of ease your way into that higher volume type of thing that, that actually could be very beneficial. Yeah. I mean, those,

Andrew Seddon: those big EMS companies are tough to work with. If you're realistically, if you're doing like sub 50 million a year, like, yeah, it's hard, but they're just, they're just, you know, they don't get up to,

Chris Gammell: to deal with that. Right. Exactly. Yeah. I, I, I don't plan to, but that is an interesting idea of like before you said it, I, I don't think I've ever thought of that before as an option. The idea of like the, this is probably just my stupidness. Not, not that it didn't exist, but like the idea of like kind of that ongoing, just the steady drip of, of boards, weekly boards, monthly, that is a, that is a very intriguing idea I had not thought of before. Yeah. I think kind of the whole, to, to, to, yeah, to some extent the whole, the whole,

Andrew Seddon: trying to optimize the cost curve. Yeah. It's, you know, what, when you think of flattening that price curve, like it's, it's not, it's not that interesting in isolation to reduce the cost on making a hundred units. Right. Like, but it becomes like, cause you, cause you think, okay, well I order a hundred units once as part of my, my, my scale up process, right? Like I order a few prototypes and then I order a hundred units and then I go straight to 10,000. Right. So in isolation, reducing the, the cost of those hundred units, it's not that interesting, but if you, if you switch the mental model a bit and you think, okay, well look, I'm never actually going to order 10,000 units. I'm just going to keep ordering a hundred units, 200 units, 300 units. Yeah. But I'm still over the course of the year going to, be able to achieve similar sort of pricing. So if I had ordered 10,000 units, well then it starts to get really interesting because you get the benefits of like lower working capital. You get the ability to iterate on the design. You don't get those heart stopping moments where, you know, you've, you, you realize you've messed up and you've just sunk the business, uh, because you've got $500,000 worth of boards that you can't use. And so there's a, there's a ton of, ton of benefits to it.

Chris Gammell: Yeah. And it does kind of benefit all parties too. Like you said, cashflow manufacturer likes it. I like it. It, it kind of does, if it, if it was possible, it'd be good. And maybe, maybe that goodness in the short term, short to medium term, it actually makes up for the, the, any, any differential that would be there from the kind of higher volume anyways. So yeah, it's very interesting. I'd like to, I know we're way over time already, but I'd like to just circle back one more time to just kind of tariffs and domestic stuff, because you're kind of on the ground, seeing that stuff in the U S. Speaker 1 What, what, what should I do? What the hell man? Like, what should I do? What are you, what are you seeing? What are you, what are people doing or, you know, how is it impacting things? And then people can turn this off if they're sick of hearing about it too.

Andrew Seddon: Look, I mean, it's, it's very unpredictable, right? You do, who knows what's gonna happen with the terror situation, uh, tomorrow. I, I believe, uh, we're, we're going back to, uh, I think it's 50% tariffs from, uh, from China now as, as of today or tomorrow. We're pre-recording this. So I w I would

Chris Gammell: say we should just, you know, this is, uh, probably two weeks away from publishing date. So I'd say we should really, really hedge our bets on that one. Yeah. Yeah. So, so, okay. Yeah. Let's, I guess I

Andrew Seddon: won't try and predict, predict the future, but I think there's a, there's this general, uh, you know, there's a, there's this general trend to, uh, to increasing tariffs, I would say in the U S. And so we've seen, we've seen certainly, uh, a whole new type of customer approaches that we'd never really seen before. And that is companies that make consumer electronics. So it would have historically been very rare for us to, you know, have a conversation with a company making, uh, you know, like watches or laptops or, uh, you know, various wearables, consumer devices. Um, because generally they will, because they're made in such high, high volume that they're generally going to work like an ODM, uh, over in China, typically who are going to help them both on the design side and on the manufacturing side. And then they'll deal primarily with the software and the marketing, uh, actually increasingly less on the software side, actually, they tend to be more marketing heavy operations. Um, but obviously with the tariff situation, that is becoming less of a sustainable model. And so we, we were having conversations with these consumer electronics companies about, um, not just prototyping, but like, can we feasibly do consumer electronics manufacturing, uh, in, in the U S. And that's, um, I think it is, uh, feasible actually like as much as my longer term view now, is this something that you can spin up overnight? No, but I think it is something that you can, you can work towards over time. Um, and then the, the, the question then becomes how do tariffs relate to the, to the inputs on that? Right. So if you're going to do, uh, electronics assembly, well, that is often the final step in the process. Right. And so you have a bunch of inputs into that. You have the parts, you have the, uh, the bare board fabrication, and then, you know, you go, you go up a step from that and you have the inputs into bare board fabrication, the chemistry, the laminate, you have the inputs into the semiconductor industry in terms of power and, uh, a lot of chemistry and such. So how the tariffs relate to those things, uh, is also increasingly important. So, so you really need a lot of these, uh, the, the supply chain to be built out like simultaneously, uh, so that you can bring in, you know, sort of increasing amounts on, on shore for it, for it to make sense. Um, so, I mean,

Chris Gammell: So you're gonna start a chip hub anytime soon? Is that what I'm a chip hub? Well, I, I think there are,

Andrew Seddon: uh, yeah, it seems to be a lot of people working on it. And obviously the, the U S government has invested quite substantially into the, the chips act. And, uh, there's a lot of fabs being mixed results,

Chris Gammell: I think, but yeah, I think some, some movement. Yeah. Yeah. I mean, there's, there's quite a few

Andrew Seddon: big crabs and I'm aware of at least one, uh, company is working on the, on the chip side of things. And so I think there'll be some, uh, some movement in that space. Um, so yeah, it's a bit TBD. It depends, depends what happens. I mean, there, there was this, that short period of time where the tariffs went up to about 150%. And at that, at that point, like you just, you can't do consumer electronics manufacturing in China, right? Like it's just no longer feasible. And so Apple have rapidly moved a lot of their production over to, to Vietnam. And I think a bunch of it's going over to India. They're obviously making the bet that those regions are going to get favorable, uh, tariff rates. Um, and then I think they, Apple are, uh, are looking at on-shoring, uh, their production.

Chris Gammell: They, they do. I think everybody's giving it lip service at the very least, right? We're going to see news articles first and then five years from now, we'll see where they land. Right.

Andrew Seddon: Yeah. But you know, it's SpaceX with this Starlink facility have certainly proven that it's, it is possible. And so, uh, you know, if the will is there, it's, it, it, it, it can be done whether, you know, it's debatable, whether you want to drive that through, uh, through tariffs there. I think as somebody who runs a business, you just want stability, right? Like you want, you want the game to be understood. You want everybody to know the rules of the game. And then you just let the businesses optimize around that. And so what's really not good for anyone is uncertainty, right? Like it's incredibly difficult to make investments and, uh, and plan ahead when, when things are changing rapidly. And so, so my hope is we get some more certainty on, on what's going on. And then once everybody understands a new game, then we can, you know, we can, we can all play to, uh, all play to that. But, you know, as a general theme, I do hope we, um, you know, we see a lot of manufacturing, uh, returning to, to, to design engineers, right? I'm a design engineer at heart and I, I don't, I, I want to be involved in the manufacturing side of things. Right? Like, I think that's, that's, that's fun. And, um, it provides a lot of inspiration. And so, yeah, I hope to, hope to see that happen one way or another.

Chris Gammell: I got my start at Keithley. I got David walk out to the pick and place line, like, and like, have an engineer be like, this was a bad idea. Like there are benefits, like you mentioned earlier, like there are such benefits from that. I don't know if this is how we get there, but like, I am all about it. Like, I, I think, I think it's great. I think, you know, having quick turn country, like that stuff is awesome. I just, yeah, I, I, I don't think, I don't think the current path is, is it, but maybe there will be some tack ons that help. I don't know. Yeah. So it's interesting to hear the, how the, how the kind of the customer mix for you has changed. Um, that is, that is very interesting. We'll find out when this is published. I don't know if we'll have any updates on it, but

Andrew Seddon: we'll, yeah, we'll all say, I wouldn't want to place a bet on it right now. I think, uh, yeah,

Chris Gammell: we'll just have to see how it pans out. All right. Well, Andrew, thank you so much for being here again, 12 years later. Uh, I'd say let's check in less than 12 years. What do you think?

Andrew Seddon: I think so. Yeah. No, it's been fun, Chris. I, yeah. I appreciate, uh, appreciate you taking the time and, um, yeah, let's, uh, let's not leave it so long next time. Great. Well, uh, people can find

Chris Gammell: you at circuit hub.com and you, anywhere else that people should look to see your latest stuff.

Andrew Seddon: Yeah, that's the, that's the main thing. Um, yeah, circuit hub.com and, uh, you can see to them some links to, uh, some of my content on there. All right. Well, thanks again for being here. And we'll chat soon. Thanks Chris.

Speaker ?: Uh, let's go.