Latest / a16z AI Policy Brief / Turning America’s AI Progress Into a Manufacturing Advantage

Transcript

Turning America’s AI Progress Into a Manufacturing Advantage · a16z AI Policy Brief

0:00
SPEAKER_00: Welcome back to the a16z AI Policy Brief.
0:03
SPEAKER_00: I'm Matt Peralt.
0:05
SPEAKER_00: How can America turn its strength in software and AI into a competitive advantage
0:09
SPEAKER_00: in manufacturing?
0:10
SPEAKER_00: Today, I'm joined by Davide Asnaghi, co-founder and CEO of Diode, to tackle that question.
0:16
SPEAKER_00: This conversation is about bringing better products to people faster
0:20
SPEAKER_00: and what it takes to build them here in America.
0:22
SPEAKER_00: Davide's company builds printed circuit boards,
0:25
SPEAKER_00: the components inside everything from phones to medical devices to robots.
0:29
SPEAKER_00: Today, these critical parts are often manufactured in China.
0:33
SPEAKER_00: Diode brings manufacturing back to the United States with an approach
0:36
SPEAKER_00: that combines design and build.
0:38
SPEAKER_00: They use AI to design the boards, and then they use their own factories to build them.
0:43
SPEAKER_00: At Diode's Brooklyn factory, these elements are literally side by side.
0:47
SPEAKER_00: Engineers might sit at a desk writing code and then stand up,
0:51
SPEAKER_00: move over to the production line, and run it.
0:53
SPEAKER_00: Davide explains how connecting those two worlds can speed up hardware development,
0:57
SPEAKER_00: help workers do more, and make American manufacturing more competitive.
1:01
SPEAKER_00: We also discuss what hardware founders need from policymakers.
1:04
SPEAKER_00: As Davide explains,
1:05
SPEAKER_00: when he commits to spend $10 million on manufacturing equipment,
1:08
SPEAKER_00: he needs to have confidence that policy won't reverse course in a year or two.
1:12
SPEAKER_00: As a small company, having a policy environment that gives him this confidence is critical.
1:17
SPEAKER_00: Enjoy our conversation.
1:19
SPEAKER_02: American software is unparalleled.
1:22
SPEAKER_02: And this is even more true today than it was 10 years ago because
1:26
SPEAKER_02: the models have inherited the American spirit of ingenuity.
1:31
SPEAKER_02: The supply chains are very centralized to very few places in Asia.
1:35
SPEAKER_02: And so if anything were to happen, the US would basically be left without a supply chain.
1:40
SPEAKER_02: How do you think American companies can be competitive?
1:43
SPEAKER_02: Like, I don't think it's a question, like, this has to happen, right?
1:48
SPEAKER_02: If there are no policies that incentivize American manufacturing,
1:52
SPEAKER_02: the United States will be left behind.
1:53
SPEAKER_02: The robotics wave that's coming in the next few years and
1:57
SPEAKER_02: the AI compute requirements are so staggeringly high that if
2:02
SPEAKER_02: the US does not bring manufacturing capacity online, it will be a disaster.
2:07
[music]
2:13
SPEAKER_01: Davide, co-founder and CEO of Diode, thanks so much for joining us today.
2:17
SPEAKER_02: My pleasure.
2:18
SPEAKER_02: Thank you for having me.
2:19
SPEAKER_01: So why don't we start with you talking a little bit about what Diode does.
2:23
SPEAKER_01: There's a lot of focus in AI conversations about the software side of AI and kind
2:29
SPEAKER_01: of what that part of the ecosystem is about.
2:32
SPEAKER_01: Then there's a separate conversation.
2:34
SPEAKER_01: It's often separate about the manufacturing side.
2:36
SPEAKER_01: Diode combines the two.
2:38
SPEAKER_01: So can you tell us a little bit about what your company does?
2:40
SPEAKER_02: Absolutely.
2:42
SPEAKER_02: At a very high level,
2:43
SPEAKER_02: we live in this wonderful future where intelligence is available almost on tap
2:50
SPEAKER_02: through the use of large language models.
2:53
SPEAKER_02: And having great software is a very highly leveraged position to be in.
2:58
SPEAKER_02: And what Diode does,
3:00
SPEAKER_02: it uses this new great technology to design a very core part
3:05
SPEAKER_02: of American industrial ecosystem, which is circuit boards.
3:09
SPEAKER_02: Circuit boards are something that is available in every modern piece of hardware.
3:13
SPEAKER_02: And normally they're quite hard to design and even harder to manufacture.
3:18
SPEAKER_02: And so what Diode does, it designs the circuit board using a large language model.
3:23
SPEAKER_02: And then we manufacture them in-house in our factories.
3:28
SPEAKER_02: Specifically, like the first one is here in Brooklyn.
3:30
SPEAKER_02: I'm calling from New York.
3:32
SPEAKER_01: So I'm curious about how you identified this as a problem that needed to be solved.
3:37
SPEAKER_02: I love hardware.
3:38
SPEAKER_02: I grew up in Italy,
3:40
SPEAKER_02: which is a place with almost no electronics or, you know, a hardware culture.
3:47
SPEAKER_02: But it does have like a very specific spinoff,
3:50
SPEAKER_02: which is called Arduino,
3:51
SPEAKER_02: which is a That is trying to bring hardware to the general public.
3:56
SPEAKER_02: And it was incredibly successful over the last 20 years.
3:59
SPEAKER_02: And I always felt that that was an incredible way of taking software,
4:04
SPEAKER_02: which is something almost mathematical, right?
4:07
SPEAKER_02: And so my entire career has been in hardware.
4:12
SPEAKER_02: Like I graduated from Berkeley, straight like went into this company called Butterfly Network.
4:17
SPEAKER_02: We used to make like medical devices, incredible hardware.
4:20
SPEAKER_02: Then I worked at Apple and then I worked at another company called Fortelt or make hearing aids.
4:25
SPEAKER_02: And throughout the experience that I had,
4:28
SPEAKER_02: like building hardware in the United States,
4:31
SPEAKER_02: it always felt that we were moving way too slow because of the circuit boards,
4:36
SPEAKER_02: which is,
4:37
SPEAKER_02: It's wild because circuit boards are,
4:40
SPEAKER_02: at a glance,
4:41
SPEAKER_02: much simpler than very complicated software or very complicated silicon devices,
4:46
SPEAKER_02: which we used to do at both Fortel and Butterfly Network.
4:51
SPEAKER_02: And so this always felt wrong.
4:53
SPEAKER_02: And so this is why we started a company to fix that specific bottleneck.
4:57
SPEAKER_01: And so what's an example of a situation where
5:01
SPEAKER_01: the circuit board design would be too slow or problematic in some way?
5:05
SPEAKER_01: Like how would that be something that like a company might encounter or
5:08
SPEAKER_01: a consumer might experience in some way?
5:10
SPEAKER_01: Maybe let's start from the
5:11
SPEAKER_02: consumer angle.
5:13
SPEAKER_02: Imagine that you hear like a launch.
5:15
SPEAKER_02: There is this new medical device that will like fix all your problems
5:19
SPEAKER_02: because it uses a new transformational technology and it's going to launch at
5:23
SPEAKER_02: the end of the year.
5:25
SPEAKER_02: Then the launch slips three months, six months, nine months.
5:31
SPEAKER_02: Those nine months are nine months that that medical device is not being used in
5:35
SPEAKER_02: the hands of doctors.
5:36
SPEAKER_02: That is terrible.
5:38
SPEAKER_02: This is not the future.
5:40
SPEAKER_02: And a lot of the times,
5:43
SPEAKER_02: because circuit boards are inherently logistically complex,
5:46
SPEAKER_02: like circuit boards are by design,
5:48
SPEAKER_02: a collection of silicon devices,
5:50
SPEAKER_02: the silicon devices are by themselves very hard to produce.
5:54
SPEAKER_02: When you have to collect hundreds of them into one single product,
5:58
SPEAKER_02: any one can become a schedule problem.
6:01
SPEAKER_02: And so what ends up happening as a company is that when you design,
6:05
SPEAKER_02: you basically have these logistical problems during manufacturing,
6:08
SPEAKER_02: but even before getting to A16z.com
6:24
SPEAKER_02: We started by tackling the mistakes first in software.
6:40
SPEAKER_02: We built a wonderful software that prevents as many mistakes
6:43
SPEAKER_02: as we possibly can using simulation, using everything else.
6:46
SPEAKER_02: But then once you want to scale to production and bring that medical devices
6:51
SPEAKER_02: to hundreds or hundreds of thousands of people,
6:53
SPEAKER_02: you also need to go solve real world manufacturing problems, which are quite hard.
6:59
SPEAKER_02: And so Diode does both of these things to prevent hardware from being slow.
7:04
SPEAKER_01: Our audience is primarily on the policy side and my background is on the policy side.
7:07
SPEAKER_01: So this is,
7:08
SPEAKER_01: I think,
7:08
SPEAKER_01: obvious to you,
7:09
SPEAKER_01: but not necessarily obvious to me and maybe to some people who are listening.
7:12
SPEAKER_01: How prevalent are printed circuit boards?
7:15
SPEAKER_01: Like what are the kinds of technologies where you need a printed circuit board
7:19
SPEAKER_01: in order to make it operate?
7:20
SPEAKER_02: My brain is kind of wired in this way, but this microphone, circuit board.
7:25
SPEAKER_02: The computer I'm looking into, circuit board.
7:28
SPEAKER_02: The cameras that are recording this podcast, circuit board.
7:32
SPEAKER_02: If you look around,
7:33
SPEAKER_02: it's almost impossible that you will And this is true for anybody who's listening.
7:41
SPEAKER_02: You will have your phone, obviously, a circuit board.
7:43
SPEAKER_02: There's obviously varying degrees of complexity in circuit boards,
7:46
SPEAKER_02: but they go from very simple to very,
7:49
SPEAKER_02: very arcane and complex, like data center class circuit boards.
7:54
SPEAKER_02: So not all of them are created equal,
7:56
SPEAKER_02: but they're almost as prevalent as silicon in our modern world.
7:59
SPEAKER_02: Anything that does compute will have a circuit board on it.
8:02
SPEAKER_02: And almost anything that we interact with daily has compute in it.
8:06
SPEAKER_01: And is your technology able to address all those different types of circuit board scenarios?
8:11
SPEAKER_02: We are trying very hard.
8:13
SPEAKER_02: Obviously, the models are improving exponentially.
8:16
SPEAKER_02: And we started with like simpler circuit boards, you know, like smaller, lower power.
8:21
SPEAKER_02: And now we are walking up the stack.
8:24
SPEAKER_02: And now we work with aerospace companies,
8:27
SPEAKER_02: robotics companies, companies that need to do data center level compute for AI.
8:31
SPEAKER_02: And like
8:33
SPEAKER_02: In principle, there's no limit to the complexity the software can handle.
8:37
SPEAKER_02: But because I am the one running the line sometimes,
8:40
SPEAKER_02: it is very hard to manufacture some of those most complicated ones.
8:44
SPEAKER_02: But we will get there.
8:45
SPEAKER_02: Right now, we are in the medium tier of complexity.
8:48
SPEAKER_01: I'm curious to understand a little bit more about the founding story of the company.
8:52
SPEAKER_01: So you had designed a sophisticated custom chip that was functioning,
8:57
SPEAKER_01: but then the circuit board that it was running on failed.
9:00
SPEAKER_01: Can you tell a little bit about that experience?
9:02
SPEAKER_02: Of course.
9:03
SPEAKER_02: So I didn't design the chip.
9:05
SPEAKER_02: Somebody way smarter than I am actually designed the chip.
9:09
SPEAKER_02: And the same person also designed the circuit board.
9:12
SPEAKER_02: And that shit was a work of art.
9:16
SPEAKER_02: My contribution was mostly the software.
9:17
SPEAKER_02: At the time, I was like writing software for hardware.
9:20
SPEAKER_02: And we were so worried that the silicon, which costs millions of dollars...
9:28
SPEAKER_02: It's called a tape-out because originally you used to have this little tape
9:32
SPEAKER_02: that you would put on the masks and you would expose with ultraviolet light.
9:36
SPEAKER_02: Today it's not the same anymore, but anytime you build silicon, you call this a tape-out.
9:41
SPEAKER_02: And it is an incredibly expensive operation.
9:44
SPEAKER_02: So you are very worried as a company or as an individual building silicon.
9:47
SPEAKER_02: And so we spent all of this time simulating it and checking that it would work.
9:52
SPEAKER_02: And when it came back, it worked first try.
9:56
SPEAKER_02: And I was astounded.
9:58
SPEAKER_02: I was like, wow, hey, you designed this silicon and this silicon is amazing and it works.
10:04
SPEAKER_02: But then the same person designed the circuit board,
10:06
SPEAKER_02: which by all accounts, it was a much simpler process.
10:10
SPEAKER_02: But because the tools were not helpful,
10:13
SPEAKER_02: the same way that the tools are helpful for silicon,
10:16
SPEAKER_02: we basically had to savage some of those very precious silicon and transplant it
10:21
SPEAKER_02: onto another circuit board.
10:22
SPEAKER_02: And this is just not right.
10:26
SPEAKER_02: It doesn't feel right.
10:27
SPEAKER_02: This feels so wrong that we are wasting time bringing a product that can help people immensely.
10:34
SPEAKER_02: When you're building medical devices, there are people that really benefit from those things.
10:39
SPEAKER_02: Mm-hmm.
10:39
SPEAKER_02: Having a six months or a six weeks delay just because
10:42
SPEAKER_02: the circuit board is not right is just unacceptable, in my opinion.
10:46
SPEAKER_02: So that was like a radicalization.
10:48
SPEAKER_02: I got radicalized by that specific moment.
10:51
SPEAKER_02: I felt, no, this has to change.
10:53
SPEAKER_02: And there are places in the world...
10:54
SPEAKER_02: So can you talk a little bit more about that?
10:55
SPEAKER_02: Why does it function better in parts of Asia than it does in the United States?
11:13
SPEAKER_02: So I was fortunate enough to spend part of my undergraduate in Hong Kong,
11:18
SPEAKER_02: partially because I was so interested in electronics and I wanted to understand,
11:21
SPEAKER_02: like, why is Southeast Asia so much better?
11:24
SPEAKER_02: You know,
11:24
SPEAKER_02: like Taiwan,
11:26
SPEAKER_02: Vietnam,
11:27
SPEAKER_02: Thailand,
11:28
SPEAKER_02: and especially China can build these circuit boards at astounding speed and a very,
11:34
SPEAKER_02: very competitive cost.
11:36
SPEAKER_02: And the reason is that in the 90s and early 2000s,
11:42
SPEAKER_02: The United States and Europe decided that this was a skill
11:45
SPEAKER_02: that was probably not useful for them to cultivate.
11:48
SPEAKER_02: The United States made a decision that it was easier to outsource some
11:52
SPEAKER_02: of these very core parts of the manufacturing infrastructure.
11:55
SPEAKER_02: And at the time,
11:56
SPEAKER_02: the rationale The idea now was that labor was much more expensive in
11:59
SPEAKER_02: the United States than Asia.
12:02
SPEAKER_02: And it is still true to a degree today, but that is not really true anymore.
12:07
SPEAKER_02: Like if you look at a modern Chinese printed circuit board factory, it is incredibly automated.
12:12
SPEAKER_02: With very sophisticated technology.
12:14
SPEAKER_02: And there are some things where labor is still like,
12:17
SPEAKER_02: you know,
12:17
SPEAKER_02: like China has very highly skilled, fairly low wage jobs in the electronic industry.
12:24
SPEAKER_02: So a factory in China,
12:26
SPEAKER_02: if you like just transplanted it to the US would not work
12:28
SPEAKER_02: because you wouldn't be able to find the people that like would be able to do
12:32
SPEAKER_02: that work economically.
12:34
SPEAKER_02: But it is very automated.
12:35
SPEAKER_02: So like right now, we just don't have that skill set.
12:38
SPEAKER_02: And it's a skill set that needs to be rebuilt from first principles on American soil.
12:42
SPEAKER_01: That sounds like an incredibly challenging competitive issue
12:45
SPEAKER_01: and I would think would be one for me sitting on the outside not as
12:48
SPEAKER_01: a founder would be just incredibly daunting to figure out how to compete effectively
12:54
SPEAKER_01: in that when there's that dynamic.
12:56
SPEAKER_01: But you are optimistic about it.
12:58
SPEAKER_01: Like how do you think American companies can be competitive?
13:01
SPEAKER_02: Like, I don't think it's a question, like, this has to happen, right?
13:06
SPEAKER_02: Like, it's either you say, okay,
13:07
SPEAKER_02: pack it up, like,
13:08
SPEAKER_02: we're not going to do any more, like,
13:10
SPEAKER_02: cutting-edge technology,
13:11
SPEAKER_02: we're just going to, like, sit down and, like, wait for somebody else to do it.
13:14
SPEAKER_02: Or you figure out a way to do it both cost-effectively and speed, like,
13:19
SPEAKER_02: in a very fast way in the U.S. Like,
13:22
SPEAKER_02: I love the United States because the United States will never sit down and say,
13:27
SPEAKER_02: yes, I'm okay being second at something.
13:29
SPEAKER_02: No.
13:29
SPEAKER_02: Okay.
13:30
SPEAKER_02: The US needs to be the best.
13:32
SPEAKER_02: And so there is just a lot of competitive dynamics that you're right are very hard.
13:37
SPEAKER_02: But at the same time,
13:38
SPEAKER_02: it is such an important part of both critical infrastructure and the ability
13:44
SPEAKER_02: to scale If you want to build new type of products,
13:52
SPEAKER_02: new type of hardware,
13:53
SPEAKER_02: if you don't have this core part of circuit board infrastructure,
13:56
SPEAKER_02: you are going to be left behind.
13:58
SPEAKER_02: And the supply chains are very centralized to very few places in Asia.
14:04
SPEAKER_02: And so if anything were to happen, the U.S. would basically be left without a supply chain.
14:09
SPEAKER_02: The goal here is basically to take a problem that is very strongly felt in the U.S.
14:15
SPEAKER_02: and solve it.
14:16
SPEAKER_02: And I find that this type of challenge very interesting.
14:18
SPEAKER_02: I do find that this used to be almost impossible to like wrap your head around.
14:23
SPEAKER_02: And now it's just like.
14:25
SPEAKER_01: As you tell it,
14:26
SPEAKER_01: like one of the interesting things about how you frame the challenge and
14:30
SPEAKER_01: the solution is there's this competitive dynamic that's an issue.
14:33
SPEAKER_01: But if the United States tried to respond to that competitive challenge
14:36
SPEAKER_01: by just running the playbook that China has, it wouldn't be the right one.
14:39
SPEAKER_01: We wouldn't be able to compete effectively.
14:41
SPEAKER_01: So from your perspective, like how can you set up an American system to compete effectively?
14:46
SPEAKER_01: Like what is the American playbook that can be responsive?
14:49
SPEAKER_02: When I came to Berkeley,
14:52
SPEAKER_02: one of the things that I absolutely loved is that American software is unparalleled.
15:00
SPEAKER_02: It is truly outstanding how innovative the ecosystem in the U.S. is.
15:05
SPEAKER_02: And this is very easy to encode in software.
15:08
SPEAKER_02: When you have a fast-moving environment like software where you don't need
15:12
SPEAKER_02: to wait six weeks,
15:13
SPEAKER_02: very smart people will be drawn to that idea and the ability to go
15:17
SPEAKER_02: and do wonderful things.
15:20
SPEAKER_02: And this is even more true today than it was 10 years ago because
15:24
SPEAKER_02: the models have inherited the American spirit of ingenuity.
15:28
SPEAKER_02: They are able to generate software at an unparalleled speed.
15:32
SPEAKER_02: And if you couple that deep understanding of technology I can design systems
15:37
SPEAKER_02: that are very autonomous using software.
15:40
SPEAKER_02: And you also understand how a factory works.
15:43
SPEAKER_02: You can merge these two things and basically have your models design a circuit board.
15:49
SPEAKER_02: That is incredibly efficient to produce with American technology.
15:54
SPEAKER_02: And so this means that there are some things that you cannot do.
15:56
SPEAKER_02: Like if you have something that is very manual and labor intensive,
15:59
SPEAKER_02: like there are some manufacturing operations that require a human to go and solder,
16:03
SPEAKER_02: for example.
16:04
SPEAKER_02: That will be very, very expensive.
16:06
SPEAKER_02: And there are some factories in China that work completely in this way,
16:09
SPEAKER_02: and it is economically viable.
16:21
SPEAKER_02: But in the U.S.,
16:22
SPEAKER_02: we want to be able to empower a lot of the talent that is available,
16:26
SPEAKER_02: not to do manual tasks,
16:28
SPEAKER_02: but to supervise machines that are able to do these tasks at a much larger scale and
16:33
SPEAKER_02: in a much larger scale.
16:36
SPEAKER_02: And so if you design your circuit boards to be produced in the United States,
16:40
SPEAKER_02: you will be able to make these trade-offs before the design is finalized.
16:44
SPEAKER_02: This is what I think will unlock era of American re-industrialization.
16:49
SPEAKER_02: It's not just, hey,
16:50
SPEAKER_02: let me copy a playbook that works because of the labor dynamics in
16:54
SPEAKER_02: a different country,
16:55
SPEAKER_02: but it is,
16:56
SPEAKER_02: let me take what we have, play to our strengths, and then double down on that idea.
17:02
SPEAKER_01: What is it about software right now that makes this a unique moment
17:06
SPEAKER_01: for doing what you're trying to do?
17:08
SPEAKER_02: I think it's not even what the software is capable of doing.
17:12
SPEAKER_02: It's just that there are thousands of paper cuts that compound.
17:17
SPEAKER_02: The way that you program a machine is very time intensive
17:20
SPEAKER_02: and usually you need individual steps that are done by somebody on
17:25
SPEAKER_02: the factory floor.
17:26
SPEAKER_02: And now the ability to automate those steps is just one click away.
17:30
SPEAKER_02: You can ask It would have been a one-year project for a team of engineers
17:34
SPEAKER_02: a year ago, and now it's like...
17:48
SPEAKER_02: A16z.
18:03
SPEAKER_02: But what has not happened is that it's quite hard still to produce high mix circuit boards.
18:11
SPEAKER_02: And now with software,
18:12
SPEAKER_02: you can kind of make these two things look like the same problem
18:15
SPEAKER_02: and do it incredibly efficiently.
18:16
SPEAKER_02: And so uniting design and manufacturing at the same time is a superpower,
18:22
SPEAKER_01: allows you to unlock a lot.
18:23
SPEAKER_01: There's a lot of focus in the policy conversation now about software.
18:27
SPEAKER_01: I'm curious how you think about that dynamic in your work and how do you ensure that
18:31
SPEAKER_01: the AI usage that you do then performs in a reliable manner?
18:46
SPEAKER_02: I really like the idea that you just evoked of like an orchestra and like a conductor.
18:52
SPEAKER_01: I wish I could take credit for that idea.
18:54
SPEAKER_01: You just invoked that idea.
18:55
SPEAKER_01: I like it.
18:56
SPEAKER_01: But you give me credit for the orchestra metaphor.
18:58
SPEAKER_01: I'll take it.
18:59
SPEAKER_02: This is very, very apt to the way that we think about humans in our systems.
19:05
SPEAKER_02: So if you look at the number of electrical engineers in the United States compared
19:11
SPEAKER_02: to the number of electrical engineers in Southeast Asia,
19:13
SPEAKER_02: for example, it is a A huge difference.
19:18
SPEAKER_02: You have droves and droves of manufacturing and electrical engineers in,
19:24
SPEAKER_02: for example, China, and a very few, comparatively speaking, in the U.S.
19:30
SPEAKER_02: But the need...
19:30
SPEAKER_02: It's just impossible to produce this amount of talent.
19:45
SPEAKER_02: And so it is much,
19:47
SPEAKER_02: much better to let the relatively few but very smart human conduct a symphony
19:54
SPEAKER_02: of AI agents that will run your design process,
19:57
SPEAKER_02: that will run your simulations, that will run your factories.
20:00
SPEAKER_02: And you are able to empower a very strong talent to do much more with less,
20:08
SPEAKER_02: which is how we think about both manufacturing and design.
20:11
SPEAKER_02: These two things are very important to us.
20:13
SPEAKER_02: One of the tenants of design We need to be able to open a new factory location
20:20
SPEAKER_02: in Arizona,
20:22
SPEAKER_02: New Mexico, Austin, and go to the local high school or college and take anybody.
20:29
SPEAKER_02: It doesn't matter what major they are.
20:30
SPEAKER_02: It doesn't matter if they have a technical background.
20:32
SPEAKER_02: They don't even need to be engineers and let them run the line in a way that is very efficient.
20:37
SPEAKER_02: This is a very important part.
20:39
SPEAKER_02: We're not quite there yet.
20:41
SPEAKER_02: Right now,
20:42
SPEAKER_02: you still need some technical background,
20:44
SPEAKER_02: but the software that we develop on the manufacturing side is incredibly intuitive.
20:49
SPEAKER_02: It will literally talk to you and say, this is what you need to do.
20:54
SPEAKER_02: It's quite impressive in my opinion, but this is how we try to design it.
20:57
SPEAKER_01: It sounds like the software is unlocking something on the human side,
21:01
SPEAKER_01: but then you also need humans there to oversee and direct the software.
21:05
SPEAKER_01: Of course.
21:06
SPEAKER_02: I mean, this is just like the nature of manufacturing.
21:09
SPEAKER_02: I don't know if you've ever seen like any manufacturing facility,
21:12
SPEAKER_02: but like we are moving things, right?
21:13
SPEAKER_02: You know,
21:14
SPEAKER_02: like people have been theorizing that you can use robots in factors,
21:18
SPEAKER_02: but the reality is like,
21:19
SPEAKER_02: My customers need circuit boards today,
21:21
SPEAKER_02: and you need humans to make those circuit boards and ship them out.
21:25
SPEAKER_02: And I need to make so many that fewer humans need to do way more than it's currently possible.
21:31
SPEAKER_01: But I mean,
21:31
SPEAKER_01: it does sound like,
21:32
SPEAKER_01: again, you're talking about a specific American way of doing this.
21:35
SPEAKER_01: It sounds like the comparison to China is in part more robots in factories.
21:39
SPEAKER_01: The way that you're describing it is not a substitution issue
21:42
SPEAKER_01: of substituting machine labor for human labor.
21:45
SPEAKER_01: It seems to me,
21:46
SPEAKER_01: at least the way that I think about it using the policy narrative that's in my head,
21:50
SPEAKER_01: is it sounds like a version of an upskilling or reskilling program
21:54
SPEAKER_01: where you're actually unlocking American talent to perform jobs
21:57
SPEAKER_01: that might have otherwise not been possible for them.
22:00
SPEAKER_02: That's right.
22:01
SPEAKER_02: And it's also like the quality of the job is, in my opinion, much more interesting.
22:06
SPEAKER_02: Like it's not doing the repetitive task.
22:10
SPEAKER_02: There is a very good reason why like OSHA exists, right?
22:15
SPEAKER_02: Over and over for eight hours, it is quite detrimental to your health.
22:20
SPEAKER_02: This is not something that we want humans doing.
22:22
SPEAKER_02: You want to be able to exercise your judgment.
22:25
SPEAKER_02: That is what humans are best at, your creativity.
22:28
SPEAKER_02: At least I'm biased, but I really love the idea.
22:30
SPEAKER_02: Of manufacturing not just being a mindless task, but it's like, oh, I'm orchestrating a line.
22:36
SPEAKER_02: And why is this robot not doing the work that I want to be doing?
22:41
SPEAKER_02: My machines are effectively robots.
22:43
SPEAKER_02: But this is what you use standard.
22:48
SPEAKER_02: How do I understand why this is happening?
22:51
SPEAKER_02: How do I use my creative endeavors to build the circuit boards that are wonderful
22:57
SPEAKER_02: and have a higher yield and so on?
22:59
SPEAKER_02: And so this is truly upskilling.
23:02
SPEAKER_02: You get humans in charge of things that they like doing rather than just a job.
23:06
SPEAKER_01: All that makes intuitive sense to me, but I'm curious about some specific examples of that.
23:11
SPEAKER_01: Obviously,
23:12
SPEAKER_01: People like their jobs more when they're more creative and fulfilling
23:15
SPEAKER_01: and they have potential for self-actualization.
23:18
SPEAKER_01: What does the prior world look like from a job perspective?
23:22
SPEAKER_01: And why do you feel so confident that the kinds of jobs that you are able
23:26
SPEAKER_01: to offer people through this combination of software
23:28
SPEAKER_01: and hardware design is more creative and more fulfilling?
23:32
SPEAKER_02: I'll give you an example of something that would happen on a line in Asia today.
23:37
SPEAKER_02: So you have a conveyor belt,
23:40
SPEAKER_02: you take a circuit board,
23:42
SPEAKER_02: you place a component, solder it, and then put it back on the conveyor belt.
23:46
SPEAKER_02: And you do this for 8, 9, 10 hours a day.
23:52
SPEAKER_02: This is uneconomical in the United States, and the person doing this needs to be trained.
23:58
SPEAKER_02: Another example is there is a screen on a machine,
24:02
SPEAKER_02: and you sit in front of this screen,
24:05
SPEAKER_02: or you stand sometimes,
24:06
SPEAKER_02: and the only thing that you do is you look at a picture and you say, good or bad?
24:11
SPEAKER_02: And that's it.
24:12
SPEAKER_02: That's your entire job.
24:13
SPEAKER_02: So this is what some manufacturing operations look like today.
24:16
SPEAKER_02: And there's very good reasons for this, right?
24:18
SPEAKER_02: Like the cost of taking that picture and making a wrong judgment call is very high.
24:23
SPEAKER_02: Imagine that you introduce a defect in your production run.
24:26
SPEAKER_02: Your customer will lose the trust that they have in you as a manufacturer.
24:31
SPEAKER_02: So it makes complete economic sense to pay somebody to just look at pictures all day.
24:35
SPEAKER_02: But these problems have become automatable now.
24:39
SPEAKER_02: To a point or to a degree where the person that used
24:42
SPEAKER_02: to be doing those jobs can now oversee the entire production run.
24:46
SPEAKER_02: And because we need so much of this capacity that we don't have right now,
24:50
SPEAKER_02: there's less than 6% of the total circuit board market is currently manufactured
24:55
SPEAKER_02: the United States.
24:56
SPEAKER_02: But more than 30% of the demand comes from American companies.
25:00
SPEAKER_02: So there's a huge gap and that gap is growing.
25:03
SPEAKER_02: We need more lines and more people that will oversee the line without doing
25:07
SPEAKER_02: the menial work, but actually understanding, oh, this machine is making a mistake.
25:12
SPEAKER_02: Why?
25:12
SPEAKER_02: Okay, let me go fix it.
25:14
SPEAKER_02: And this is actually quite interesting,
25:15
SPEAKER_02: and it can be guided in a very interesting way, in my opinion.
25:19
SPEAKER_01: So it's funny,
25:19
SPEAKER_01: as you were just talking,
25:20
SPEAKER_01: I realized it became hard for me to listen closely after a certain moment,
25:25
SPEAKER_01: and it was when you said defect.
25:26
SPEAKER_01: And so I think for someone who operates in the policy legal sphere,
25:29
SPEAKER_01: like once you say defect,
25:30
SPEAKER_01: that is just like the blaring thing,
25:33
SPEAKER_01: in part because in product liability law,
25:36
SPEAKER_01: basically there's a version of strict liability that applies and it requires
25:40
SPEAKER_01: a showing that there's a defect in the manufacturing process.
25:43
SPEAKER_01: So I'm curious if Given that concern that you're obviously like very aware of,
25:49
SPEAKER_01: how does your company,
25:49
SPEAKER_01: the technology that you implement, how does it address those kinds of situations?
25:53
SPEAKER_01: It sounds like you have confidence that the process
25:55
SPEAKER_01: that you're setting up is not just creating more fulfilling jobs,
25:58
SPEAKER_01: but also could identify those defects and remedy them early in the process.
26:02
SPEAKER_01: How does that part of it work?
26:03
SPEAKER_02: You can imagine that if I am looking at pictures all day for 10 hours,
26:08
SPEAKER_02: towards the end of my shift, I will be tired.
26:12
SPEAKER_02: That is very quantifiable.
26:14
SPEAKER_02: You can basically check what the defect rate of a given line is.
26:20
SPEAKER_02: And you can track it over time because you can actually test the boards
26:23
SPEAKER_02: after they have been manufactured.
26:26
SPEAKER_02: And what we see is that by automating the parts that do the defect qualification,
26:31
SPEAKER_02: it is quite hard technically to do.
26:33
SPEAKER_02: You need better cameras.
26:35
SPEAKER_02: You need a camera that will sweep around.
26:37
SPEAKER_02: You need very sophisticated algorithms to find those defects.
26:42
SPEAKER_02: But the outcome is a much higher quality, more automated process.
26:48
SPEAKER_02: The Japanese manufacturer have a saying that anytime you touch a circuit board,
26:53
SPEAKER_02: And so your goal as a human is to basically supervise and make sure that each
26:57
SPEAKER_02: of those steps it's doing its algorithmic job correctly, not to be
27:08
SPEAKER_02: I'm curious about kind of the arc of the company to date
27:11
SPEAKER_02: in taking what sounds like it was this initial really strong concept, but then
27:36
SPEAKER_01: Manufacturing is quite difficult.
27:51
SPEAKER_02: It's not quite hard, but it is also incredibly rewarding.
27:54
SPEAKER_02: Seeing the line go from a small prototype,
27:58
SPEAKER_02: 5, 10, to 100, to 500, to 5,000, 10,000, and see these units just keep going.
28:06
SPEAKER_02: It's just that as the models become
28:22
SPEAKER_02: We can switch more of our company time in fine tuning the manufacturing.
28:27
SPEAKER_02: And we are at the point where it is a no brainer for us to say, let's buy a second line.
28:32
SPEAKER_02: It's a no brainer for us to say, let's do more ambitious things.
28:35
SPEAKER_02: Let's go do some chemistry,
28:37
SPEAKER_02: which we usually didn't do,
28:38
SPEAKER_02: because these are the problems that are creating bottlenecks in
28:41
SPEAKER_02: the American supply chain.
28:43
SPEAKER_01: Our team has had the idea,
28:46
SPEAKER_01: we've talked about the concept of a digital supply street,
28:49
SPEAKER_01: so not trying to physically recreate a supply street in a place like China,
28:54
SPEAKER_01: but actually trying to create sort of a digital version of it.
28:56
SPEAKER_01: And I think that goes Is that something that you're seeing play out in practice?
29:00
SPEAKER_01: For sure.
29:12
SPEAKER_02: I think that when you say supply street,
29:16
SPEAKER_02: what I actually hear is there is a very strong ecosystem of partnerships.
29:21
SPEAKER_02: This is both true in Japan, they call it a kiretsu, I believe, and in China.
29:25
SPEAKER_02: If you look at Shenzhen, it is wonderful.
29:27
SPEAKER_02: Yeah.
29:42
SPEAKER_02: A16z A16z
29:57
SPEAKER_02: You are becoming friends, right?
29:59
SPEAKER_02: It's just like, oh, hey, can you hook me up with a battery, please?
30:03
SPEAKER_02: I really need it for this project.
30:04
SPEAKER_02: And my customer would be unhappy if we couldn't source this as well.
30:09
SPEAKER_02: I think that that is incredibly powerful and something that is not talked about.
30:13
SPEAKER_02: Being able to not only improve the process for every vertical,
30:17
SPEAKER_02: but also improve the network is a huge lever, even at a policy level.
30:23
SPEAKER_02: You cannot just invest in circuit boards and expect everything to be done.
30:27
SPEAKER_02: You need to invest in silicon, in substrates, in the machines makers.
30:32
SPEAKER_02: Getting machines, a huge bottleneck for me.
30:35
SPEAKER_02: The machine manufacturers cannot make them fast enough for me to buy them.
30:39
SPEAKER_02: So these are the things that I really care about when I think about supply chain.
30:43
SPEAKER_01: You're leading us down,
30:43
SPEAKER_01: I think,
30:44
SPEAKER_01: a really productive path to focus a little bit more on the policy environment
30:48
SPEAKER_01: that needs to exist in order for you to be successful.
30:52
SPEAKER_01: You just alluded to a comprehensive approach to the manufacturing process
30:57
SPEAKER_01: and you need policy support for each phase of that.
31:00
SPEAKER_01: What are the kinds of policies or policy initiatives that are desirable for you
31:06
SPEAKER_01: to see in order for you to be successful?
31:08
SPEAKER_02: For what it's worth, I think the U.S. is actually doing an incredible job.
31:13
SPEAKER_02: There was a decision that was very deliberate to say the U.S.
31:17
SPEAKER_02: needs to become a more industrial country for many reasons,
31:22
SPEAKER_02: like national security,
31:24
SPEAKER_02: economic prosperity,
31:25
SPEAKER_02: the realization that software is going to become part of the AI ecosystem,
31:31
SPEAKER_02: but hardware is not quite there yet and it needs to be helped.
31:35
SPEAKER_02: And I think that those decisions are really pushing the entire ecosystem in
31:39
SPEAKER_02: the right direction.
31:40
SPEAKER_02: Because at the end of the day,
31:41
SPEAKER_02: as a company,
31:42
SPEAKER_02: the thing that makes my decision-making clear is a very objective direction that
31:53
SPEAKER_02: the entire ecosystem is going towards.
31:55
SPEAKER_02: If there is a back and forth where one year, yes,
31:59
SPEAKER_02: we need to industrialize,
32:00
SPEAKER_02: the next year is like,
32:01
SPEAKER_02: yeah, maybe we do not, that is incredibly detrimental to making business decisions.
32:06
SPEAKER_02: I cannot decide to invest $10 million in equipment.
32:10
SPEAKER_02: Yeah.
32:25
SPEAKER_02: I truly believe that there is a generational opportunity for this.
32:30
SPEAKER_02: So when I think of policy, I'm not a policy expert, so I cannot really speak in details.
32:34
SPEAKER_02: But any policy that gives me like a degree of certainty that it's not going to be reversed.
32:40
SPEAKER_02: It's
32:46
SPEAKER_02: a
32:50
SPEAKER_02: huge way for me to say, yes, this makes complete sense.
32:53
SPEAKER_02: Bipartisan decisions are great for us to make decisions that are not going
32:56
SPEAKER_02: to be reversed in one year or two years.
32:59
SPEAKER_01: How do things look for you as a startup relative to larger companies in the ecosystem?
33:05
SPEAKER_01: I assume there are So how does that dynamic affect you?
33:29
SPEAKER_02: So for us specifically,
33:31
SPEAKER_02: I think we actually are in a very unique position because the larger companies
33:37
SPEAKER_02: that play in our space are not American.
33:41
SPEAKER_02: There's very few large American companies that are building American manufacturing.
33:46
SPEAKER_02: And so it is reluctantly that they are bringing their manufacturing facilities to the US.
33:53
SPEAKER_02: I can cite a few examples,
33:54
SPEAKER_02: but the core idea is investing in CapEx in the United States makes very little sense
33:59
SPEAKER_02: if you don't believe that you are going to be there to stay.
34:03
SPEAKER_02: It is quite refreshing, actually, to be a small company.
34:06
SPEAKER_02: We can move in a very nimble way, and we can bring new infrastructure online.
34:12
SPEAKER_02: Even as we speak,
34:14
SPEAKER_02: the amount of demand compared to the amount of supply in
34:18
SPEAKER_02: the circuit board manufacturing industry, there is a huge gap.
34:23
SPEAKER_02: And we see it with existing manufacturers, which are doing amazing work.
34:27
SPEAKER_02: Like US manufacturers are at 100% capacity.
34:30
SPEAKER_02: Like lead times keep going up.
34:32
SPEAKER_02: And so as a small company, like we have a software DNA, right?
34:37
SPEAKER_02: Like we've built hardware.
34:38
SPEAKER_02: But we are also incredibly good at writing software.
34:41
SPEAKER_02: So it could seem counterintuitive that we are bringing more manufacturing capacity online.
34:45
SPEAKER_02: Like, why are you doing this?
34:47
SPEAKER_02: We literally cannot find enough capacity for our customers.
34:50
SPEAKER_02: The only way to do it is to bring it online ourselves.
34:52
SPEAKER_02: And like larger companies are trying to do it,
34:55
SPEAKER_02: but we will obviously outcompete on speed, right?
34:58
SPEAKER_02: Like we'll do whatever it takes.
34:59
SPEAKER_02: We'll buy the machine, use them, fix them ourselves.
35:01
SPEAKER_02: We'll like ask for the demo units that are in universities and we will ship them to us.
35:05
SPEAKER_02: Like we will get the job We don't need to influence necessarily policy because
35:10
SPEAKER_02: the policies,
35:10
SPEAKER_02: if made with common sense decisions,
35:13
SPEAKER_02: are obviously pushing in the direction of we need more manufacturing in the U.S.
35:17
SPEAKER_02: So I hope this will be an uncontroversial opinion.
35:34
SPEAKER_01: If only we could trust the policy process to deliver common sense solutions.
35:38
SPEAKER_01: Can you talk a little bit more about what Diode looks like today?
35:41
SPEAKER_01: So one thing that we often experience
35:44
SPEAKER_01: in policy conversations is policymakers will have a view of what
35:47
SPEAKER_01: a company looks like because many of the companies that they interact
35:50
SPEAKER_01: with regularly have that look.
35:52
SPEAKER_01: That is a company that has an established, sizable, often legal team.
35:57
SPEAKER_01: And then within that legal team, Might have a head of policy, a chief policy officer type role.
36:03
SPEAKER_01: Some of them have lobbyists that are specific to the House of Representatives,
36:06
SPEAKER_01: specific to the Senate, specific to Republicans, specific to Democrats.
36:11
SPEAKER_01: Some teams will have someone who works directly with an administration.
36:14
SPEAKER_01: Sometimes they'll have different teams based on whether
36:16
SPEAKER_01: the administration is Republican or Democrat.
36:18
SPEAKER_01: Curious about kind of the physical org chart nature of what your company looks like
36:23
SPEAKER_01: at this phase.
36:25
SPEAKER_02: I am now taking notes because we're all engineers.
36:29
SPEAKER_02: We have a really nice office in Brooklyn that is also our factory.
36:33
SPEAKER_02: It's about 10,000 square feet.
36:35
SPEAKER_02: And we have the engineers sitting on their tables looking at our manufacturing line.
36:40
SPEAKER_02: And when they are done coding, they stand up and they run the manufacturing.
36:44
SPEAKER_02: And so if you want to take a snapshot of our company,
36:50
SPEAKER_02: it's like I am in the process of negotiating a real estate deal in San Francisco.
36:56
SPEAKER_02: And I cannot put more than a certain amount of desks because otherwise I cannot call
37:01
SPEAKER_02: that space an industrial space.
37:03
SPEAKER_02: But I need the desk for my engineers who will also be running the line.
37:06
SPEAKER_02: And so this is how my company looks.
37:08
SPEAKER_02: We have no head of policy yet.
37:11
SPEAKER_02: I am...
37:13
SPEAKER_02: Very hopeful that I will not need to have lobbyists.
37:18
SPEAKER_02: What I want to focus on is on manufacturing.
37:21
SPEAKER_02: And I actually do think that these policies will be fairly uncontroversial.
37:27
SPEAKER_02: If there are no policies that incentivize American manufacturing,
37:31
SPEAKER_02: the United States will be left behind.
37:33
SPEAKER_02: It's already behind China from a perspective of manufacturing capacity.
37:38
SPEAKER_02: And we need this, right?
37:41
SPEAKER_02: The robotics wave that's coming in the next few years and
37:45
SPEAKER_02: the AI compute requirements are so staggeringly high that if
37:50
SPEAKER_02: the US does not bring manufacturing capacity online, it will be a disaster.
37:55
SPEAKER_02: What I hope to do as a company,
37:57
SPEAKER_02: what we hope to achieve is to give those technical means of production.
38:04
SPEAKER_02: And you're making me a bit worried here, but I'm hoping that the policy will follow.
38:08
SPEAKER_02: But we also want to have a showcase for any policymaker that you can do it.
38:15
SPEAKER_02: This is much more important to me than any lobbying.
38:18
SPEAKER_02: Here's a working example of a factory that's producing circuit boards in
38:22
SPEAKER_02: a very economical way in Brooklyn.
38:24
SPEAKER_02: If you can do it in Brooklyn, I need to be able to do it in Ohio.
38:28
SPEAKER_02: I can do it in Tulsa.
38:30
SPEAKER_02: I can do it across the
38:31
SPEAKER_01: United States.
38:32
SPEAKER_01: Brooklyn and San Francisco,
38:33
SPEAKER_01: I think a lot of people would think if you're doing real estate deals and trying
38:37
SPEAKER_01: to be cost effective about it, those are not the two places that you would choose.
38:40
SPEAKER_01: I mean, literally some of the highest real estate, most expensive real estate in the country.
38:44
SPEAKER_02: You would be absolutely right,
38:46
SPEAKER_02: but specifically for a printed circuit board assembly,
38:50
SPEAKER_02: the actual real estate footprint that you need to produce a million boards
38:54
SPEAKER_02: a year is not that high.
38:55
SPEAKER_02: It's about 4,000 square feet.
38:57
SPEAKER_02: Incredibly efficient.
39:00
SPEAKER_02: Just to frame it for anybody who's listening and is not very familiar
39:06
SPEAKER_02: with circuit boards,
39:07
SPEAKER_02: unlike smelting or steel production or even CNC machining, these things are tiny.
39:14
SPEAKER_02: Most circuit boards are the size of your hand.
39:17
SPEAKER_02: And so if you need to produce even 100,000, they will fit on a crate that I can ship on a plane.
39:23
SPEAKER_02: You don't even need freight, right?
39:25
SPEAKER_02: Obviously, it really helps because the machines that we need are fairly large.
39:29
SPEAKER_02: But these are very high value items.
39:32
SPEAKER_02: You can think about them almost like as if the silicon is like a diamond on
39:35
SPEAKER_02: a jeweler's ring,
39:37
SPEAKER_02: the circuit board is the either platinum or gold,
39:40
SPEAKER_02: like actual ring that will hold the gem.
39:43
SPEAKER_02: It is a very high value item.
39:44
SPEAKER_02: And so the unit economics really work out.
39:49
SPEAKER_02: And it is much more advantageous for me to be close to my customers,
39:54
SPEAKER_02: which are the startups in San Francisco,
39:56
SPEAKER_02: which are the companies doing aerospace in Los Angeles,
39:59
SPEAKER_02: and be able to ship them aboard the same day than it is for me to be in,
40:04
SPEAKER_02: you know, Taiwan.
40:05
SPEAKER_02: If I need a ton of real estate,
40:08
SPEAKER_02: the obvious conclusion is why don't you outsource it to somewhere else?
40:12
SPEAKER_02: Because the US as a whole has a fairly high real estate cost and labor costs.
40:16
SPEAKER_02: But that's not the point.
40:17
SPEAKER_02: My internal goal for the company is if we are profitable in Brooklyn NSF,
40:22
SPEAKER_02: just imagine how much we can scale this when we go to millions instead of hundreds
40:26
SPEAKER_02: of thousands, which we can do in these two geographies.
40:29
SPEAKER_02: And then as an aside, we hire very smart people.
40:32
SPEAKER_02: There's a lot of smart engineers on the coast.
40:35
SPEAKER_02: Our factories don't look like what you would think of a factory.
40:38
SPEAKER_02: We have a very nice view of the Brooklyn Bridge.
40:40
SPEAKER_02: And we want people to be excited about going and manufacturing things.
40:44
SPEAKER_02: And we don't want to close people in a basement.
40:46
SPEAKER_02: That's not the goal.
40:48
SPEAKER_02: And you can do both.
40:49
SPEAKER_02: You can do both and you can be incredibly profitable and deliver a ton of value.
40:52
SPEAKER_02: So this is the philosophy behind our first offices.
40:55
SPEAKER_02: But we are also like exploring Texas.
40:57
SPEAKER_02: I love Austin.
40:58
SPEAKER_02: It's a great place.
41:00
SPEAKER_02: Like there's a lot of like real estate.
41:02
SPEAKER_02: There are parts of the production that we are looking into
41:05
SPEAKER_02: that will be much more like square footage inefficient, unlike assembly.
41:10
SPEAKER_02: So I'm not,
41:11
SPEAKER_02: you know,
41:11
SPEAKER_02: categorically opposed, but you should be able to do both if you're efficient enough.
41:15
SPEAKER_01: We've been talking a lot about the process that is involved
41:18
SPEAKER_01: in making better circuit boards,
41:20
SPEAKER_01: making them more efficiently, making them faster, empowering American manufacturing.
41:24
SPEAKER_01: From a consumer's perspective,
41:25
SPEAKER_01: it might be great to have better circuit boards,
41:27
SPEAKER_01: but what they really want are the products that result from better circuit boards.
41:31
SPEAKER_01: In your vision,
41:32
SPEAKER_01: if your company succeeds,
41:34
SPEAKER_01: how is that going to affect the products that consumers are using?
41:37
SPEAKER_02: Let's take a very concrete example.
41:39
SPEAKER_02: I don't know how many of your listeners have read sci-fi,
41:43
SPEAKER_02: but I'm sure some of them are familiar with Isaac Asimov.
41:47
SPEAKER_02: In the 1960s,
41:49
SPEAKER_02: when you thought about the future,
41:50
SPEAKER_02: what you thought about was humanoid robots that would walk around.
41:55
SPEAKER_02: And if you go to San Francisco,
41:57
SPEAKER_02: some of our customers, when I go visit them, live in that reality today.
42:02
SPEAKER_02: Their office has a bunch of robots that walk around and talk to you.
42:06
SPEAKER_02: And That type of innovation and bringing the future closer is not just a mechanical process.
42:15
SPEAKER_02: There needs to be some inspiration.
42:16
SPEAKER_02: There needs to be some creativity.
42:19
SPEAKER_02: There needs to be the perception of fast progress that makes everybody at
42:23
SPEAKER_02: the company innovating in robotics excited.
42:26
SPEAKER_02: That magic gets lost any time that you wait six weeks for your circuit board
42:31
SPEAKER_02: to come back and your robot is not walking in your factory
42:34
SPEAKER_02: or your robot is not seeing because we haven't delivered the eyes
42:38
SPEAKER_02: or your robot is not hearing.
42:40
SPEAKER_02: All of these things compound very, very quickly.
42:42
SPEAKER_02: So what we try to do is we want to unblock a very important
42:49
SPEAKER_02: and sometimes overlook We work a lot with aerospace, robotics, some automotive.
43:10
SPEAKER_02: But those products make their way into the life of our consumers.
43:14
SPEAKER_02: We do satellites that check the weather.
43:17
SPEAKER_02: If you want better weather predictions,
43:20
SPEAKER_02: we want to give those circuit boards as fast as we possibly can.
43:24
SPEAKER_02: And I think our lives just become better and better as that technology comes online.
43:29
SPEAKER_02: Today, we cannot do things like toys,
43:32
SPEAKER_02: for example,
43:32
SPEAKER_02: which I would love to be able to do,
43:34
SPEAKER_02: but we're not quite cost competitive enough that it makes economic sense for
43:38
SPEAKER_02: a company to say,
43:39
SPEAKER_02: I'm going to build a very low margin product on our manufacturing process.
43:44
SPEAKER_02: But as we scale volume, we will.
43:46
SPEAKER_02: And it would be kind of cool to have toys made in the United States.
43:50
SPEAKER_02: David, thanks so much for coming on the podcast.
43:53
SPEAKER_02: It's my pleasure.
43:54
SPEAKER_02: Thank you, Matt.
43:55
[music]