Latest / EDGE AI POD / Edge AI Investing Essentials

Transcript

Edge AI Investing Essentials · EDGE AI POD

• 0:07
  Speaker 1: We're going to talk today about investing in the edge.
• 0:09
  Speaker 1: What I really love about these events is that we have such an eclectic mix of presentations.
• 0:18
  Speaker 1: We have academic presentations, commercial presentations we learned about Dell Native Edge earlier today, millimeter Wave a little later on, and now we're going to talk about investment and really like for all of us here, you know it's all fun and games, but we also need some money to invest and to build these businesses, and so we've assembled a panel of folks that do that.
• 0:39
  Speaker 1: They put their money where their mouth is, they invest in this space, and we're going to talk about what they're looking for in investments, why they're investing, what do they see as sort of the big opportunities on the horizon?
• 0:50
  Speaker 1: So let me ask each of them to introduce themselves and we'll kind of go down the line.
• 0:56
  Speaker 1: Hans, if you want to get started, Sure, I'm Hans.
• 1:01
  Speaker 2: I'm with Momentum Ventures.
• 1:02
  Speaker 2: We're focused on startup seed to Series A companies in industrial automation.
• 1:11
  Speaker 2: Primarily.
• 1:11
  Speaker 2: Our funds are set up where each fund has an anchor investor and for us it's Rockwell.
• 1:18
  Speaker 2: So we're essentially investing on behalf of Rockwell across industrial automation, edge, ai, edge, impulse out there is one of our startups when we invested pretty early on.
• 1:32
  Speaker 3: Vitaly, I'm with a venture fund called Cyber Fund.
• 1:36
  Speaker 3: It's a thesis-driven fund.
• 1:39
  Speaker 3: We invest on the intersection between AI and crypto, and the main goal for us is to move the entire world towards this thing that we call cyber economy, which is a state of economy where the majority of the actions are performed by either robots or agents or all kinds of AI models.
• 2:10
  Speaker 4: Hi, my name is Riyaz.
• 2:12
  Speaker 4: I work for Murata in their corporate venture capital team.
• 2:15
  Speaker 4: Murata being a hardware company, our focus is to look into those enabling technologies, new technologies coming up that will enable more of these, not just AI Many verticals we identified Robotics is one of it, because that takes a lot of edge AI, a lot of edge communication, a lot of power technologies all of that.
• 2:40
  Speaker 4: So that's where I'm from, thank you.
• 2:44
  Speaker 5: My name is David Wyatt and, unlike the rest, I'm not really working.
• 2:48
  Speaker 5: I'm having fun.
• 2:49
  Speaker 5: So I used to be working at NVIDIA for nine years.
• 2:55
  Speaker 5: I was a distinguished engineer and before that, intel.
• 2:57
  Speaker 5: Open GPU is all about creating an open source, open standard infrastructure for interconnecting the next generation of AI, particularly those that are looking at the power efficiency like at the edge.
• 3:12
  Speaker 5: It's very important to us, but also thinking ahead beyond the end of Moore's law and how we can leverage things like time or temporal computing to be way ahead of the curve.
• 3:23
  Speaker 1: Cool, good, yeah, so I think one of the things that we probably want to touch upon here and this has been kind of a theme is around things that will sort of take the friction out of the edge AI, industry and, in this case, sort of the investment profile.
• 3:40
  Speaker 1: So maybe we can sort of go down the line or whoever you want to what order you want to talk about, like, what are you seeing as kind of the key to unlocking the scale of this industry when you're looking at investments?
• 3:57
  Speaker 2: uh, so um, the first thing we're looking for is customers, second is customers, and third is customers.
• 4:05
  Speaker 2: The number of lines of code you write doesn't matter.
• 4:07
  Speaker 2: It doesn't give a return.
• 4:08
  Speaker 2: You have to realize every VC is investing, especially institutional investors are investing on behalf of somebody else, so we're expecting a return.
• 4:17
  Speaker 2: This is not a science project, it's not a grant, and the only way that money comes back is when you have revenue.
• 4:25
  Speaker 2: And so you need to start with.
• 4:25
  Speaker 2: Whose job do you change?
• 4:26
  Speaker 2: Whose daily job do you change?
• 4:27
  Speaker 2: Who shows up every day and says these, this tool helped me with my job before and after.
• 4:33
  Speaker 2: Can you describe it in their language?
• 4:35
  Speaker 2: Second, who pays for you to change it?
• 4:38
  Speaker 2: Can they justify setting aside money for you based on everything else that they need to spend money on just for you?
• 4:46
  Speaker 2: And if you can answer that question, you're well beyond most startups, who mostly focus on the product, the widget.
• 4:52
  Speaker 2: It's not about your widget.
• 4:54
  Speaker 2: It's about who.
• 4:55
  Speaker 2: Whose problem are you solving and are they willing to give you money, and what is it going to take you to extract money out of them?
• 5:01
  Speaker 2: Because frequently the biggest competition is customer not doing anything.
• 5:04
  Speaker 2: Because you confuse the hell out of them, because frequently the biggest competition is customer not doing anything because you confuse the hell out of them with all the ai you brought to them.
• 5:10
  Speaker 2: So we focus on industrials.
• 5:12
  Speaker 2: We deal with a lot of plant managers and and people not tech right, they're trying to build something and get it out of the factory.
• 5:19
  Speaker 2: So focus on your customer focus on your customer.
• 5:32
  Speaker 3: Yeah, I fully agree with that.
• 5:33
  Speaker 3: Yes, so, but we like a cyber fund.
• 5:34
  Speaker 3: We we also understand that sometimes there is like an open source library that is supported by one or a couple of guys and this library is essential for like many, many things, and what we would, what we like to do, is sometimes we actually give grants to all of this, like individual contributors and like open-source contributors that are Supporting, you know, a tremendous infrastructure that is built for AI or beyond.
• 6:05
  Speaker 3: So that's why we are supporting Edge AI Foundation right, because we see a lot of talent that it puts together and with very, very specific, let's say, traits, because you guys are very limited in power and compute and you are very open to different alternative architectures for neural networks and stuff like this.
• 6:35
  Speaker 4: So coming from a corporate venture capital, slightly different for us.
• 6:40
  Speaker 4: All of the things handset and waitlist are all important, but we lean on our lead investors to take care of that.
• 6:47
  Speaker 4: From a CVC perspective, we look at the technology.
• 6:51
  Speaker 4: Is the technology of value?
• 6:53
  Speaker 4: Is it futuristic?
• 6:54
  Speaker 4: Is there a meeting point somewhere in future we can see, or is it really solving something?
• 7:00
  Speaker 4: Number one.
• 7:00
  Speaker 4: Number two look at the founders.
• 7:02
  Speaker 4: Are these founders having enough industry experience?
• 7:06
  Speaker 4: You know, many times a lot of technology come out of academy great, but they don't know what the industry challenges are or they never have that vision of you know you can get excited by a technology without knowing how it can get into production.
• 7:25
  Speaker 4: Or does it really meet the needs?
• 7:29
  Speaker 4: Like, for example, if you want to do a new transistor technology, great, but you have to see if it doesn't fit into the CMOS industry.
• 7:37
  Speaker 4: That's today.
• 7:37
  Speaker 4: If it's not, it's dead on arrival done.
• 7:39
  Speaker 4: So we look into more of those technology aspects and founders' experience.
• 7:43
  Speaker 4: So we look into more of those technology aspects and founders experience and the rest of it.
• 7:46
  Speaker 4: We lean on the lead investors like Hans and Mittli.
• 7:49
  Speaker 4: Whenever you work with CVCs.
• 7:52
  Speaker 5: They are more technology focused yeah, so I can give a quick story, maybe two different stories, okay, so one is anyone here heard about CUDA?
• 8:04
  Speaker 5: Okay, surprising.
• 8:07
  Speaker 5: You're welcome.
• 8:10
  Speaker 5: Okay.
• 8:10
  Speaker 5: So how did CUDA start?
• 8:12
  Speaker 5: It did not start with any customers None, okay, it did not start with a grand vision None, in fact, jensen didn't know about it.
• 8:24
  Speaker 5: He didn't like it with a grand vision None, in fact, jensen didn't know about it.
• 8:29
  Speaker 5: He didn't like it.
• 8:29
  Speaker 5: Okay, it was funded out of a skunk works in Dwight's pocket, basically pocket lint.
• 8:33
  Speaker 5: Okay, so key things were.
• 8:34
  Speaker 5: The need Driving this was the idea that the GPU had to be useful for something other than just gaming.
• 8:42
  Speaker 5: Right, because if you've got competitors in the same space as you, all fighting for the same customers, especially if one of them is Intel making GPUs, you are going to die unless you get ahead of the curve and make something that the customers will want.
• 8:59
  Speaker 5: Okay.
• 9:00
  Speaker 5: So I would differ from Hans in saying think not just what customers you can secure today, but think about what you can set the paving stones that you can put down on the dirt to make the way for the bitumen that you put down ahead of that for eventually putting some gas stations along the road.
• 9:18
  Speaker 5: But it starts with putting paving stones down before the gas stations are there.
• 9:22
  Speaker 5: That's kind of the front end of the curve putting paving stones down before the gas stations are there.
• 9:26
  Speaker 5: Okay, that's kind of the front end of the curve.
• 9:28
  Speaker 5: Okay, now to the second thing you know this is a great point about can you be, you know, interoperable with the current processes?
• 9:31
  Speaker 5: So a great academic that I know got up in front of a panel of investors and, just you know, talking through what she was proposing from academia, she touched on, oh, and it works with standard CMOS processes, and it completely went over the head of everyone in the audience, right?
• 9:49
  Speaker 5: The mere fact that she was the only person in the entire conference that even mentioned that was just blew me away.
• 9:54
  Speaker 5: It really is so important to be practical in what you're doing.
• 9:59
  Speaker 5: You've got to be thinking ahead and how you can make it not just interoperable but easily manufacturable without asking the manufacturers to do some leap of faith into something that doesn't exist yet.
• 10:09
  Speaker 5: That's a great science project, but if we're going to find near-term customers, we have to be thinking about, you know, what can we realistically do now while getting ready with an architecture that's ready for the future, for the future customers?
• 10:23
  Speaker 5: My two cents Good, good.
• 10:28
  Speaker 1: So yesterday we were talking about, you know, efficiency is the new currency, right?
• 10:31
  Speaker 1: It used to be.
• 10:32
  Speaker 1: Latency is currency, Now it's efficiency is the new currency.
• 10:35
  Speaker 1: Is the new currency?
• 10:35
  Speaker 1: Right, it used to be.
• 10:36
  Speaker 1: Latency is currency, Now it's efficiency is the new currency.
• 10:37
  Speaker 1: And so what are you seeing in the kind of investment horizon around efficiency that's really impressing you these days, Like you know, we've heard about you know, we all read about the deep seek stuff and all the small language models and things and everyone's now hot on how do we get more efficient, which is kind of why everyone's here.
• 10:53
  Speaker 1: So what are you seeing on the investment horizon?
• 10:56
  Speaker 1: Maybe we'll start with David, because it looks like you're ready to say something.
• 10:58
  Speaker 5: I don't want to be the one that just gets the last word.
• 11:00
  Speaker 1: No, no, go for it.
• 11:00
  Speaker 1: It's gonna be a bad trend if we start that one.
• 11:03
  Speaker 5: Okay, so let me give you three things.
• 11:06
  Speaker 5: Okay, number one has anyone heard of Louie?
• 11:10
  Speaker 5: Okay, one person.
• 11:13
  Speaker 5: All right, let me ask you a different way.
• 11:16
  Speaker 5: If you thought there was something that was 100 times more efficient than a GPU in inferencing, you would say, okay, so that's 2019 data, that's not recent.
• 11:29
  Speaker 5: That was the first.
• 11:30
  Speaker 5: Louis-hee chip was 100 times more efficient inferences per joule at a batch size of one, not doing thousands in a batch Batch size of one.
• 11:41
  Speaker 5: All right, so that's the leading edge in efficiency for the edge, as far as I'm concerned.
• 11:47
  Speaker 5: Second thing you have to think about what Jensen is doing for the edge.
• 11:51
  Speaker 5: His idea of supporting the edge is to put a GPU in the telephone pole Not just one, but many of them, and they're 10K each, and it's up a wooden pole in your neighborhood.
• 12:02
  Speaker 5: I'm sure they'll be there for a long time.
• 12:07
  Speaker 5: Third thing, in terms of efficiency DeepSeek's interesting, but did you know that everything that was done on that LLM had been done in an FPGA on 13 watts?
• 12:22
  Speaker 5: Okay, so again, academia, great source for inspiration.
• 12:30
  Speaker 5: Berkeley, sorry, university of Santa Cruz.
• 12:32
  Speaker 5: Jason is never going to forgive me for that one.
• 12:34
  Speaker 5: University of Santa Cruz found a way to do large language models without any matrix, multiply, right.
• 12:48
  Speaker 5: Everything that you think the GPU is useful and necessary for, gone Okay.
• 12:54
  Speaker 5: And because of that, you know, it meant switching to a different number system called ternary, which happens to be fantastic for the future of carbon nanotubes.
• 13:04
  Speaker 5: So, in the idea of beyond CMOS, getting into the next generation process, you know that's brilliant, but that's 13 watts to do what would normally, you know, take tens of kilowatts, right?
• 13:12
  Speaker 5: So that, I think, is the leading edge in efficiency.
• 13:16
  Speaker 1: Any other.
• 13:17
  Speaker 4: I think that 13 watt actually comes very close to the brain, human brain, which is like 20 watt, 25.
• 13:21
  Speaker 4: 25 watt.
• 13:22
  Speaker 4: Maybe less for me, but Because you are much more efficient Lazier.
• 13:27
  Speaker 4: Lazier.
• 13:28
  Speaker 4: So I think what was the question?
• 13:32
  Speaker 4: Again, I'm sorry, oh.
• 13:33
  Speaker 1: I think what was the question again, I'm sorry.
• 13:34
  Speaker 1: Oh, it was around like what is the new efficiency innovations you're seeing.
• 13:37
  Speaker 4: I think that all that goes right.
• 13:38
  Speaker 4: Necessity is the mother of innovation.
• 13:40
  Speaker 4: The problem happens when there is unlimited anything you know.
• 13:45
  Speaker 4: You can keep on adding GPU.
• 13:46
  Speaker 4: Then you don't have to innovate, which is a big problem happening today in our society.
• 13:51
  Speaker 4: Data centers are built by you know.
• 13:55
  Speaker 4: They said okay, I can buy a nuclear power plant, put it down, you don't have to innovate, you don't have to solve.
• 13:59
  Speaker 4: Same thing happened in automotive industry.
• 14:01
  Speaker 4: The mileage never increased, not because of a technology limitation.
• 14:05
  Speaker 4: You didn't have to.
• 14:06
  Speaker 4: I think by forcing restrictions yourself is the only way you can innovate and, in a way, the good thing about CVs is that's what we are looking at Like in terms of Murata.
• 14:19
  Speaker 4: We are looking at a holistic view.
• 14:23
  Speaker 4: We have SDG, un, sdg goals in our, you know, in our guideline.
• 14:28
  Speaker 4: How are you going to build a society if you are ignoring these things?
• 14:33
  Speaker 4: You know, getting a GPU in everybody's desk doesn't solve anything in society.
• 14:38
  Speaker 4: You have to look at the entire world.
• 14:40
  Speaker 4: You know your neighbor dying and you have a GPU.
• 14:43
  Speaker 4: It doesn't help you as a society.
• 14:45
  Speaker 4: We look at the society as a whole and that is where we look at.
• 14:50
  Speaker 4: How do we?
• 14:52
  Speaker 4: You, for example, I'll tell you typically, computing started with Intel 4004, right, 4004, when they invented they created that processor for a calculator.
• 15:03
  Speaker 4: And then we found out doing everything in a numerical world had an efficiency Even at that time.
• 15:10
  Speaker 4: Yes, convert everything into numbers.
• 15:12
  Speaker 4: You know work everything in mathematics.
• 15:14
  Speaker 4: Bring it back.
• 15:15
  Speaker 4: Too bad for the slide rule industry?
• 15:17
  Speaker 4: Yeah, no, I think there is a.
• 15:20
  Speaker 4: You know you need to.
• 15:20
  Speaker 4: At some point we need to look back and see is it the right thing to do?
• 15:24
  Speaker 4: You know, sometimes, yes, because the cost of doing otherwise is higher.
• 15:28
  Speaker 4: If you want to change everything, do in physical computing.
• 15:31
  Speaker 4: The physical computing is a word from Georgia Tech, by the way.
• 15:34
  Speaker 4: Do in physical computing.
• 15:36
  Speaker 4: The physical computing is a word from George Attack, by the way.
• 15:39
  Speaker 4: They use this instead of converting.
• 15:40
  Speaker 4: Like, for example, you have a, you need to have a multiplication of 5 by 3.
• 15:42
  Speaker 4: Well, you can convert that into.
• 15:44
  Speaker 4: You know binary logic.
• 15:47
  Speaker 4: You can put a lot of gates, a lot of transistors, and then you can do it.
• 15:51
  Speaker 4: Or, if you have a 3 ohm resistance, you put a 5 ampere current across it and measure it Done how many transistors?
• 16:00
  Speaker 4: Zero, and if you think about it, so much power today is because of these transistors.
• 16:06
  Speaker 4: When you know, as if we kind of talk about, it's a big achievement.
• 16:11
  Speaker 4: We have 65 million transistors in this chip.
• 16:13
  Speaker 4: Great.
• 16:14
  Speaker 4: That means every time these transistors switch, you are taking so much power.
• 16:20
  Speaker 4: Think about it.
• 16:21
  Speaker 4: Why is this, all this megawatt of power, going into a GPU?
• 16:24
  Speaker 4: Is there a physical motion anywhere?
• 16:26
  Speaker 4: It's just pure heat.
• 16:29
  Speaker 5: Yeah.
• 16:30
  Speaker 5: So just to, I think, emphasize what you're saying is that the audience and we in general have to be self-disciplined in thinking about what efficiency and how we could be more efficient.
• 16:39
  Speaker 5: Okay, and the reason I think you're touching on here is that you know, in nature, biology and nature figured out you have to be efficient.
• 16:48
  Speaker 5: I mean, you have been crawling across the desert, you haven't eaten in days, you haven't drank any water.
• 16:53
  Speaker 5: Now you're at the pool and there's a flash of yellow and black in the corner of your eye.
• 16:58
  Speaker 5: You jump because you inferenced oh shit, tiger and you survived.
• 17:03
  Speaker 5: Right, so you survived the Darwin test.
• 17:05
  Speaker 5: So where is the Darwin test for this industry?
• 17:09
  Speaker 5: It's not.
• 17:11
  Speaker 5: I think the key point is by making more GPUs and making nuclear power plants to keep them going.
• 17:16
  Speaker 5: It's us in the room.
• 17:18
  Speaker 5: We have to think how to fulfill the role that isn't being fulfilled here.
• 17:21
  Speaker 5: There is no nature that's driving this evolutionary process.
• 17:24
  Speaker 5: We are responsible for it.
• 17:27
  Speaker 4: Did I cover the key things?
• 17:28
  Speaker 4: Thank you, yes, yes, so basically, that's when an AI is a key area.
• 17:31
  Speaker 4: Look at, you know, if you really want to solve Edge AI is a key area.
• 17:33
  Speaker 4: Look at you know, if you really want to solve, what problem are you solving?
• 17:35
  Speaker 4: It's not about running LLM in a GPU and putting a battery.
• 17:39
  Speaker 4: It's what is a key problem you're solving.
• 17:42
  Speaker 4: I always say AI's big test is can you put it in industry and have this machine start and run on its own?
• 17:48
  Speaker 4: Now you're talking about 100% accuracy.
• 17:51
  Speaker 4: You know you can't make mistakes.
• 17:53
  Speaker 4: That's where AI's real test happens.
• 17:56
  Speaker 4: And again, you know you put more and it has to be low power because you can't power all these edge AI devices.
• 18:04
  Speaker 4: If you talk about 50 billion, you know this is what Cisco talked about.
• 18:07
  Speaker 4: Right, I owe to 50 billion devices.
• 18:09
  Speaker 4: Easy to say.
• 18:10
  Speaker 4: How are you going to power these 50 billion devices?
• 18:13
  Speaker 4: Is it going to be batteries?
• 18:14
  Speaker 4: And then, if these batteries are going to trash, you know it's a big problem.
• 18:19
  Speaker 4: You know we have to look at holistically.
• 18:20
  Speaker 4: Thank, you.
• 18:24
  Speaker 3: So, um, I have two mental models for for efficiency, like first of all, I'm from finland.
• 18:30
  Speaker 3: Finland is a very cold country, so compute is a byproduct of heating process.
• 18:38
  Speaker 3: Like you take a GPU, it converts with almost 100% efficiency all of the electricity into heat, so, and I'd like this heat to be used for the district heating or something like this.
• 18:53
  Speaker 3: So, and basically, compute is free.
• 18:59
  Speaker 3: If you would stop wasting the heat, so if there is a way to reuse the heat, the compute is absolutely free.
• 19:07
  Speaker 3: So that's the first thing.
• 19:09
  Speaker 3: The second thing is I like Star Wars and there is this destroyed R2D2, right, and I was as a kid I was always afraid that it's gonna die.
• 19:20
  Speaker 3: So why is that?
• 19:23
  Speaker 3: I mean, I could back it up, right, so I could download the firmware from it and and migrated to some other droid, because there are billions of these droids that were manufactured, but unfortunately that is not the case for some reasons.
• 19:37
  Speaker 3: So the reason is all of our computers are digital and we pay a tremendous price for this.
• 19:45
  Speaker 3: We pay with energy.
• 19:48
  Speaker 3: Like, digital computers are perfect computers.
• 19:51
  Speaker 3: You could take a program from one computer to another and it's gonna produce the same exact results all the time in the first and second computer.
• 20:01
  Speaker 3: What I believe happened is that R2-D2, he has some sort of like a mixed signal in perfect computer, so it embeds like a neural network inside and you could train R2-D2 to do something, but you cannot extract the program and port it to some other robot Because the computer that he has is imperfect.
• 20:28
  Speaker 3: So this is why R2-D2 is mortal, he could die.
• 20:33
  Speaker 3: R2-d2 is mortal, he could die.
• 20:36
  Speaker 3: So I think that the trade-off is between having an immortal and inefficient computer and mortal and inefficient computer.
• 20:50
  Speaker 1: Maybe immortality is the new currency then?
• 20:52
  Speaker 2: The ultimate efficiency immortality is the new currency, then it's the ultimate efficiency.
• 20:59
  Speaker 2: So we don't really invest down at the CPU levels.
• 21:06
  Speaker 2: We don't really think of efficiency in terms of am I going to be efficient in how my model is going to run?
• 21:08
  Speaker 2: Our companies do build models that run on very, very small devices.
• 21:13
  Speaker 2: You can go and talk to Edge Impulse out there and they'll tell you everything about what it means to be efficient on running a model.
• 21:24
  Speaker 2: We have a company in the UK that's doing essentially adaptive AI where they do minimal amount of training and they can deploy a model, for example, on a drone where if it hasn't been trained how to react let's say, if a propeller prop pops off it'll automatically figure out what it needs to do and it'll get back on track.
• 21:44
  Speaker 2: They're now applying that same technology down to I'm still getting all the industrial automation terms but essentially the drives that drive a conveyor belt in a factory or a logistics warehouse.
• 21:58
  Speaker 2: They're now deploying those models onto these drives.
• 22:01
  Speaker 2: The acronym is VFDs, where the model essentially makes the drive and the conveyor belt and whatever it's plugged into completely autonomous.
• 22:13
  Speaker 2: So any outside conditions it can slow down, down, speed up, depending on whatever is going on.
• 22:18
  Speaker 2: It can coordinate with the robot that might be doing something on the line.
• 22:21
  Speaker 2: Uh, so you?
• 22:24
  Speaker 1: know, we to me when I think of efficiency.
• 22:27
  Speaker 2: We think of in, in industrial sectors, which is the efficiency of you know.
• 22:31
  Speaker 2: Can I improve the efficient productivity of all the inputs?
• 22:34
  Speaker 2: Right that go into making my output power widgets, whatever it is.
• 22:40
  Speaker 1: Makes sense, let me throw in.
• 22:44
  Speaker 1: Or, david, did you have another one on that one?
• 22:46
  Speaker 5: I was just going to say productivity, efficiency, productivity, yeah.
• 22:49
  Speaker 1: so efficiency for manufacturing in terms of resources used to create the widget.
• 22:54
  Speaker 2: Yeah, labor, materials, capital, energy, that efficiency.
• 22:58
  Speaker 1: That makes sense.
• 22:58
  Speaker 1: I was actually talking to Abhijit before this about the stack is kind of like this big sandwich Remember the Dagwood sandwich?
• 23:05
  Speaker 1: You've got all the pieces in there.
• 23:07
  Speaker 1: You've got bread and the salami and the pickles.
• 23:09
  Speaker 1: That's a tech stack that we have right now for Edge AI.
• 23:14
  Speaker 1: When you look at the stack, which part of that sandwich is interesting to you?
• 23:18
  Speaker 1: Like, where do you see the upside?
• 23:20
  Speaker 1: Is it down at the middle or is it up toward the cloud?
• 23:23
  Speaker 1: Where in the sandwich is your sweet spot?
• 23:27
  Speaker 1: Or is there a sweet spot?
• 23:29
  Speaker 1: I don't know.
• 23:31
  Speaker 5: Where do you want to start Me?
• 23:32
  Speaker 1: Yeah, sure why not Okay.
• 23:37
  Speaker 5: So when do you want to start Me?
• 23:38
  Speaker 5: Yeah, sure, why not?
• 23:38
  Speaker 5: Okay.
• 23:38
  Speaker 5: So OpenGPU is about filling in the missing space and defining the hardware, an interoperable hardware.
• 23:41
  Speaker 5: So think of the 1990s there were 36 different graphics vendors, now there's three, but then there were 36, and they're all going in different directions.
• 23:52
  Speaker 5: And it wasn't until OpenGL came in and basically SGI gave away an interoperable, open source and open standard.
• 24:03
  Speaker 5: That's the key missing piece, not just open source, open standard.
• 24:05
  Speaker 5: Then there was something for hardware designers to focus on.
• 24:09
  Speaker 5: So actually the last entrant into that 36 graphics vendor space was can anyone guess?
• 24:16
  Speaker 5: It was NVIDIA.
• 24:20
  Speaker 5: Nvidia was the last one into the graphics space, but they were the first one to pick up an open standard, open source, open GL and run with it.
• 24:29
  Speaker 5: And because of that, coincidentally, Carmack had said I'm not going to deal with all these 36 vendors, you get your shit together behind OpenGL, that's it, that's all I'm taking, and Quake and Doom, you know, lined up behind OpenGL and that was it.
• 24:43
  Speaker 5: The industry took off and so the key thing is defining.
• 24:47
  Speaker 5: From my point of view, it's the key thing is to define interoperable hardware.
• 24:51
  Speaker 5: That means hardware interfaces that can enable people doing the interesting stuff in software and frameworks and applications to leverage and not be one-off throwaway efforts, and that many people can innovate beyond that and so it can perpetuate, just as OpenGL has.
• 25:09
  Speaker 4: So Murata started in Kyoto in 1944, and Kyoto had a lot of potteries and that time there was one person called Muratwa and he found out some university did a capacitor with ceramics.
• 25:27
  Speaker 4: So there it started.
• 25:30
  Speaker 4: You know there's ceramic, was there like Kyoto has a lot of.
• 25:33
  Speaker 4: You know Kyocera came from Kyoto ceramics.
• 25:35
  Speaker 4: You know so much ceramics there and that's where it started.
• 25:39
  Speaker 4: And we still keep the same thing.
• 25:41
  Speaker 4: So to your question.
• 25:42
  Speaker 4: We look at the metal.
• 25:43
  Speaker 4: Even if I go to Murata today in Japan, talk to ten people, six of them are material scientists Because it all enabled by you know.
• 25:54
  Speaker 4: If you talk about efficiency, you of them are material scientists Because it all enabled by you know, if you talk about efficiency, you have to change the metal or you can run more and more on the same metal.
• 26:02
  Speaker 4: I mean you can do everything on x86.
• 26:04
  Speaker 4: Maybe you need so many x86.
• 26:07
  Speaker 4: Or you can go for parallel processing or a metric point.
• 26:10
  Speaker 4: All of that Same way.
• 26:11
  Speaker 4: We look at the enabling technologies at the bottom.
• 26:15
  Speaker 4: So most of the Murata products are designed by us, the material developed by us and manufactured by us.
• 26:22
  Speaker 4: So we still take the same approach when we look into the future technologies.
• 26:26
  Speaker 4: What is an enabling technology.
• 26:29
  Speaker 4: There is a company that does, for example.
• 26:32
  Speaker 4: Whenever you talk about IoT, there are multiple things needed.
• 26:35
  Speaker 4: Edge AI is a glorified version of IoT I would say.
• 26:39
  Speaker 4: These days you need to put AI everywhere, so we call it Edge AI.
• 26:43
  Speaker 4: Iot had the same concept.
• 26:45
  Speaker 4: You need to have some computing, you need to have a sensor, you need to have an actuator, you need to have communication.
• 26:50
  Speaker 4: These are the four aspects.
• 26:53
  Speaker 4: So everywhere you can save power, everywhere you can have innovation, like in communication, has been one big problem for IoT in terms of battery.
• 27:02
  Speaker 4: You know, in the original IoT we didn't run AI on it, so the biggest power consumption was happening on communication and it was not just burning.
• 27:13
  Speaker 4: You know, computing is purely burning.
• 27:14
  Speaker 4: This is doing some actual work, so we can give that benefit.
• 27:18
  Speaker 4: So one investment we did is in a backscattering technology.
• 27:22
  Speaker 4: Can you like?
• 27:24
  Speaker 4: For example, when you drive on the tollway, we have this small sticker.
• 27:29
  Speaker 4: There's no battery in it.
• 27:30
  Speaker 4: It actually operates off the transmitter power.
• 27:34
  Speaker 4: Same thing can happen in Wi-Fi, which can solve a lot of problems backscattering, and those are the things we look at.
• 27:42
  Speaker 4: I mean, what is the enabling technology?
• 27:44
  Speaker 4: Is there anything that Murata can help in terms of new material, new processes or new miniaturization things.
• 27:53
  Speaker 4: That's what we look into.
• 27:56
  Speaker 3: So we are looking for a super simple thing.
• 27:59
  Speaker 3: It's great founders.
• 28:01
  Speaker 3: So and the reason why we are looking for the great founders here is just the pure fact that you guys are typically super limited in power, in compute, and have to be creative to build something.
• 28:21
  Speaker 3: That's why we are here.
• 28:23
  Speaker 3: We don't know whether you're going to apply your skills to build like an edge AI, or you're going to design a model that is extremely efficient and going to run it in the cloud.
• 28:37
  Speaker 3: We don't know.
• 28:38
  Speaker 3: Extremely efficient and going to run it in the cloud, we don't know.
• 28:41
  Speaker 3: But we know for sure that this crowd is extremely open to the new ideas, to the new architectures and to the new way of doing things like curating data sets and so on.
• 28:51
  Speaker 3: So we are looking for the great founders.
• 28:56
  Speaker 2: Actually, that's a really good place to start.
• 28:57
  Speaker 2: So I've been doing this for about four years.
• 29:00
  Speaker 2: When I first got into it, I asked a friend of mine who's a career VC in Seattle, how do you decide what to invest in?
• 29:10
  Speaker 2: And he told me he's invested, pre-seed, all the way up to growth stage.
• 29:16
  Speaker 2: He even built a company that turned into a unicorn and he said it didn't matter where he invested, it was always the team, Always the team.
• 29:24
  Speaker 2: It's not your idea, it's the team.
• 29:27
  Speaker 2: The idea is how you're going to make money, how we're going to make money together right, and make it successful.
• 29:34
  Speaker 2: But it ultimately comes down to the team.
• 29:36
  Speaker 2: The best teams are ones where there's a hacker and a hustler.
• 29:39
  Speaker 2: So remember when I said start with your customer.
• 29:42
  Speaker 2: You need somebody who's going out there and looking for the opportunities, trying to figure out you know what's the minimal thing that I need to build, even in hardware?
• 29:51
  Speaker 2: What LOIs can I get?
• 29:53
  Speaker 2: That'll actually, somebody will say, yep, this will fix the problem, I'll give you money when you have it.
• 29:57
  Speaker 2: Those are the things that you want.
• 29:58
  Speaker 2: Those are the types of teams that we're looking for, right, when you know, you understand the technology.
• 30:04
  Speaker 2: I was just on a call this morning where the hacker is a hardcore physicist, PhD, and then the hustler is an American sales guy, the PhD is a German guy, and perfect combination, because one guy can talk your ear out in the science if you need to go down to that detail, but the other person is focused on customers and he's making sure that the only things they're working on are the things that are going to help them drive their business forward so they can keep moving it.
• 30:43
  Speaker 2: It.
• 30:43
  Speaker 2: But just in terms of where we invest our funds I mentioned earlier are set up so that each fund has a strategic company, so a big company that invests in that fund, so investing on their behalf.
• 30:50
  Speaker 2: The fund before this was with the Advantech.
• 30:53
  Speaker 2: The current fund is with Rockwell Automation and last week we announced NS Solutions, which is the IT outsourcing arm of Nippon Steel.
• 31:04
  Speaker 2: So we generally are investing in things where it's going to make production efficient.
• 31:11
  Speaker 2: It's going to make material movement more efficient, reduce the amount of waste.
• 31:18
  Speaker 2: So most of the edge AI tends to be autonomous operations.
• 31:24
  Speaker 2: So we have companies that are autonomously operating bioreactors, furnaces, those types of things.
• 31:32
  Speaker 4: Yeah, I think it's important to state here that there's one difference.
• 31:35
  Speaker 4: One of the differences between a corporate venture fund versus a typical venture is that we are not looking at the financial model, whereas most of the, if you go to the normal ventures, they look at.
• 31:48
  Speaker 4: They don't care what your product is, they look at what is your financial model is in 10 years or in two years?
• 31:55
  Speaker 4: What is the valuation going to be?
• 31:56
  Speaker 4: How many?
• 31:56
  Speaker 4: How many customers from a strategic investment, from a corporate venture?
• 32:00
  Speaker 4: We look at the technology more, of course, founders, because it has to be scalable and everything, not necessarily customers, because sometimes the corporate which we see is looking at, well, this technology, aligned with something that we do internally, maybe it can be complimenting to us, maybe we can get you customers because we know the problem, we know we have connection with customers than you do.
• 32:26
  Speaker 4: So I would say talk to both.
• 32:29
  Speaker 4: Talk to CVCs also if you are a startup, because sometimes that will give you ideas.
• 32:36
  Speaker 1: I like the hacker and hustler model.
• 32:38
  Speaker 1: I think that's like the Wozniak jobs thing, and and we've seen that pattern before, so I want to keep that in mind.
• 32:44
  Speaker 1: I wanted to also open it up to folks in the audience if we had some folks with questions.
• 32:49
  Speaker 7: Robert yeah, thank you so much.
• 33:01
  Speaker 7: I appreciate you know all the insights.
• 33:02
  Speaker 7: Um, I actually have two questions, if that's okay.
• 33:04
  Speaker 7: Um, so first question is um, you know, customers was brought up a lot customers and and, of course, the founding team or the people who are building this company.
• 33:10
  Speaker 7: How valuable do vcs consider survey data?
• 33:16
  Speaker 7: Um, alpha testers?
• 33:18
  Speaker 7: Uh, they're not yet customers, but they are in fact utilizing the product and could in fact convert into customers.
• 33:25
  Speaker 7: And what would maybe be a considerable threshold that VCs might be looking at and how would you kind of define that as a first question Interesting?
• 33:36
  Speaker 2: Anybody, yeah, hi.
• 33:39
  Speaker 2: So, as Riaz has pointed out, we're an institutional investor, we're a traditional VC where, again, we're an investment class, where we're investing somebody's money where they're expecting a return and obviously we get paid when there's an exit or something like an acquisition, right, but that's when we really get paid.
• 33:58
  Speaker 2: So that's why the focus is on financial.
• 34:00
  Speaker 2: But in terms of stage, so we're focused on seed series A and in the sectors that we play in industrial, a lot of the edge stuff.
• 34:12
  Speaker 2: The previous funds were very focused on edge.
• 34:15
  Speaker 2: We realized that it requires patient capital.
• 34:20
  Speaker 2: You're not building a SaaS company or a consumer company.
• 34:23
  Speaker 2: It requires patient capital.
• 34:25
  Speaker 2: The go-to market ecosystems are very different from traditional tech and so we're generally looking for signals and we invest in seed series a.
• 34:38
  Speaker 2: At seed, we want to see several POC's where, if I called a customer, the customer would tell me exactly what you told me that you're doing for them.
• 34:47
  Speaker 2: It could be a POC, it doesn't matter, but that shows that there's a need.
• 34:52
  Speaker 2: And then this is where the hustler side comes in, that you have the ability to actually not only close the customer.
• 34:57
  Speaker 2: If you have a paid POC, it means that you can also extract money out of the customer and you're going to go figure out how to get money from more of them.
• 35:09
  Speaker 2: So at Seed, it's mostly POCs.
• 35:10
  Speaker 2: By Series A, we're expecting companies to show up with some path to some of those POCs turning into contracts, multi-year contracts.
• 35:22
  Speaker 2: Series A is also where there's a brick wall for most startups, and I'll give you some data from last year.
• 35:30
  Speaker 2: So traditionally, companies that raise a seed series and when I say traditionally, this is kind of even 2021, 2022, during the heyday of valuations Only about a quarter of the companies that raised a seed were able to raise a series A.
• 35:49
  Speaker 2: Most of the time at series A, you need to have some revenue model.
• 35:53
  Speaker 2: Nobody's going to give you money if you don't have some customers with revenue.
• 35:58
  Speaker 2: You're going to hit a brick wall At that point.
• 36:01
  Speaker 2: If you're just building product, it is a science project and that's it, and then yeah, so we expect some contracts by the time you are at least some path to contracts.
• 36:15
  Speaker 5: So I would challenge you to think about objectively as you can, how sticky is all of that?
• 36:23
  Speaker 5: So is there any reason why they wouldn't switch tomorrow to something else?
• 36:29
  Speaker 5: And the reasons might be your hardware, the platform, you're interoperable and they need interoperability.
• 36:37
  Speaker 5: The reasons might be you've sold them on a future vision that's much bigger than just you now.
• 36:41
  Speaker 5: There's lots of different angles, but basically it boils down to how sticky can you make the connection be?
• 36:48
  Speaker 5: And I would take a different view If you can come up with something that is an example, a proof, a proof of concept.
• 36:55
  Speaker 5: That's incredibly valuable.
• 36:56
  Speaker 5: If someone's running with it and they say, yeah, we need you to come back and we're going to pay you.
• 37:01
  Speaker 5: Okay, now, the converse of that is also at some point.
• 37:05
  Speaker 5: So you should be thinking.
• 37:09
  Speaker 5: What I'm saying is you should be thinking about how to make the solution sticky but also interoperable.
• 37:12
  Speaker 5: So sell them on the future, okay, and not just you're going to try and lock them into some solution.
• 37:17
  Speaker 5: People are going to detect that and they're going to run away.
• 37:19
  Speaker 5: Not just you're going to try and lock them into some solution.
• 37:20
  Speaker 5: People are going to detect that and they're going to run away.
• 37:24
  Speaker 5: Firstly is think about selling them on the future and then also think about what can you do to come up with a story where you say, okay, now you're stuck and you love it.
• 37:35
  Speaker 5: Sorry, you're stuck, but the key thing is you love it.
• 37:39
  Speaker 5: How about you give us some money so we can keep and expand and service you?
• 37:43
  Speaker 5: So what I'm saying is the other part of the equation is when you can ask them for money or for help or support, then that's a really big proof point.
• 37:51
  Speaker 5: It doesn't have to be lots of revenue or lots of customers, but if someone's willing to put down a big chunk of change because they need you that's a really significant thing.
• 38:09
  Speaker 7: I have one more question, but if anyone else wants Good, okay.
• 38:12
  Speaker 7: What is considered when deciding to reinvest in a founder who previously lost you money and quite possibly might have a history of wins and losses and or just losses, but has demonstrated, you know, immense resilience and or perseverance in their task to succeed?
• 38:31
  Speaker 7: Have you invested in a previous loss and then won?
• 38:36
  Speaker 4: Again that very subjective.
• 38:37
  Speaker 4: You know, just because it was a loss does not mean it was a wrong idea.
• 38:43
  Speaker 4: Sometimes the ideas are premature.
• 38:45
  Speaker 4: So anybody remember?
• 38:47
  Speaker 4: What is the company Steve Jobs did after first time, leaving Apple?
• 38:51
  Speaker 4: Next, next, next was a failure, but he kind of brought it back to Apple.
• 38:57
  Speaker 5: Steve would disagree.
• 39:00
  Speaker 4: He will disagree.
• 39:00
  Speaker 4: What I'm saying is sometimes things are premature.
• 39:04
  Speaker 4: Windows and Net PC, I mean so many products had come.
• 39:09
  Speaker 4: Sometimes it's not the right environment.
• 39:10
  Speaker 4: They misjudge some environmental factors.
• 39:14
  Speaker 4: You know for a particular technology to succeed, there's so many factors need to come together.
• 39:18
  Speaker 4: Sometimes you misjudge but your technology.
• 39:19
  Speaker 4: You know For a particular technology to succeed, there's so many factors need to come together.
• 39:20
  Speaker 4: Sometimes you misjudge but your technology.
• 39:25
  Speaker 4: You know, if someone is strong in technology, especially from a CVC point of view, if I'm looking at a technology, maybe I will, because maybe I'll say well, last time they did it because they didn't know how the market is, or they didn't have the right complementary technologies, maybe we can help him.
• 39:39
  Speaker 4: So it's very subjective.
• 39:47
  Speaker 2: So you'll notice that VCs run after second-time founders.
• 39:52
  Speaker 2: It doesn't matter what you did the first time, because you learned a bunch of lessons on somebody else's dime.
• 39:57
  Speaker 2: So it's a and you know a lot of these lessons are expensive, right?
• 40:03
  Speaker 2: So if you're a second-time founder, it doesn't matter whether you succeeded or failed the first time.
• 40:09
  Speaker 2: You're in a much better place than somebody going out for the first time.
• 40:14
  Speaker 2: And somebody going out for the first time show up with a customer list.
• 40:16
  Speaker 2: It doesn't matter, even if it's just an idea.
• 40:25
  Speaker 5: Yeah, you have to be able to reflect on the.
• 40:26
  Speaker 5: The person who's failed many times has to be able to really convincingly reflect on why they failed and what they learned.
• 40:29
  Speaker 5: And they have to demonstrate an ability to pivot, because steve jobs pivoted more than once in the life of next, as your example.
• 40:37
  Speaker 5: Firstly pivoted from doing hardware to doing just software NextOS.
• 40:40
  Speaker 5: And then right, what was sold to Apple is in fact, the OS that powers your phone.
• 40:46
  Speaker 5: Right, it became Darwin, became macOS.
• 40:50
  Speaker 5: Everything runs on top of essentially what was, you know, started with NextOS because Apple needed something beyond macOS 9 and they were way behind Microsoft and Jobs pivoted and sold them a software company and a future vision, right?
• 41:06
  Speaker 5: So the ability to recognize your mistakes, learn from them, but also recognize the opportunities is something that a failure can lead to, or it can lead to drinking.
• 41:16
  Speaker 5: You just have to differentiate which class of person you're working with.
• 41:20
  Speaker 5: Thank you very much Good questions.
• 41:25
  Speaker 1: Oh, I think there's a question in the back there.
• 41:27
  Speaker 1: Oh, I'm sorry.
• 41:33
  Speaker 6: This is kind of piggybacking on the previous question what is the biggest mistake you've seen startups make in their early stages?
• 41:39
  Speaker 6: And, to follow it up, what advice do you have For, like future entrepreneurs, that they don't make that mistake?
• 41:45
  Speaker 4: I would say honesty.
• 41:46
  Speaker 4: Honesty is absolutely necessary.
• 41:50
  Speaker 4: If something is not working, open up, share that what is not working, don't try to hide things and Keep in mind the VC community has a back channel of checking things.
• 42:02
  Speaker 4: Dishonesty will never be pardoned, okay.
• 42:09
  Speaker 4: So always be honest.
• 42:11
  Speaker 4: Don't try to hide things.
• 42:12
  Speaker 4: When something is not working, explain it.
• 42:15
  Speaker 4: I mean, maybe somebody can help.
• 42:17
  Speaker 4: If you try to cover up and do things, you're gone.
• 42:23
  Speaker 3: I think that's inability to reflect on your past performance and inability to act really quick to the changes and pivot your business, because I think it's almost inevitable for a majority of the startups to actually have an evolution of their initial ideas, and this inability to kind of give up on the initial direction is really killing a lot of startups, I think.
• 42:56
  Speaker 2: Yeah, so there's actually research on this that came out 12 years ago, why startups die or fail.
• 43:01
  Speaker 2: You can actually Google it Visually.
• 43:04
  Speaker 2: Did an infographic?
• 43:06
  Speaker 2: Just Google visually and why startups fail and what they found was they failed because they scaled one or more or five things too quickly.
• 43:14
  Speaker 2: Number one was product.
• 43:16
  Speaker 2: That's essentially what Vitaly here described.
• 43:20
  Speaker 2: We see a lot of founders who are just so focused on the idea.
• 43:23
  Speaker 2: When a customer says this is not useful for me, they blame the customer.
• 43:28
  Speaker 2: When anybody comes in, they blame the customer.
• 43:30
  Speaker 2: You're a science project.
• 43:33
  Speaker 2: You're not a business.
• 43:34
  Speaker 2: You're a science project.
• 43:36
  Speaker 2: Number two was scaling customers too quickly.
• 43:39
  Speaker 2: The reality is your initial customers, your product is pretty shitty.
• 43:42
  Speaker 2: You're selling a vision and when the customer comes in, they're gonna start complaining.
• 43:47
  Speaker 2: And now you're dealing with customer complaints and that distracts you from other things building a business.
• 43:52
  Speaker 2: The third thing was you start growing your ecosystem too fast.
• 43:56
  Speaker 2: You start signing up partners everywhere and we see this a lot as well.
• 43:59
  Speaker 2: You need again.
• 44:06
  Speaker 2: You need to start with the customer whose job you change and work backwards from there.
• 44:08
  Speaker 2: Who do they have a relationship?
• 44:08
  Speaker 2: So we see companies will go sign up a relationship with Microsoft and Amazon, but if you're selling in a factory, they have zero control over the budgets for what goes in the factory.
• 44:19
  Speaker 2: So, wrong channel, you're gonna waste your time.
• 44:23
  Speaker 2: The fourth thing was they brought on too many people.
• 44:26
  Speaker 2: If you're very early, you need a small team so you can get to a point where you at least get some traction.
• 44:32
  Speaker 2: As you start bringing people in, you get more ideas.
• 44:35
  Speaker 2: People start doing their own thing.
• 44:37
  Speaker 2: You start getting distracted.
• 44:38
  Speaker 2: More ideas people start doing their own thing, you start getting distracted.
• 44:44
  Speaker 2: The fifth thing which surprised me was that you raised more money than you needed, and what it does is it creates all the other problems that we went through before.
• 44:49
  Speaker 2: So there's multiple things that kill startups, but your only source of truth is going to be your customer.
• 45:01
  Speaker 2: Nobody else.
• 45:03
  Speaker 2: Nobody else.
• 45:03
  Speaker 2: We're all opinion here.
• 45:04
  Speaker 2: None of us are buying your product.
• 45:06
  Speaker 2: You're not buying your product.
• 45:07
  Speaker 2: Go talk to the person who's going to part with money and make sure they're willing to part with money and make sure you, as the first investor in that business.
• 45:15
  Speaker 2: It's worth your while to go spend 10 years building this, because founder depression is real and nobody talks about it.
• 45:22
  Speaker 2: But now there's communities that are popping up like this.
• 45:25
  Speaker 2: It's not.
• 45:26
  Speaker 2: Nobody talks about the shitty side of being a startup.
• 45:29
  Speaker 2: So I was a founder.
• 45:30
  Speaker 2: I sold my company to ge.
• 45:31
  Speaker 2: That's why I'm talking from experience, uh, so yeah, listen to the market and sorry one more time.
• 45:39
  Speaker 5: Your question was how do you Sorry?
• 45:51
  Speaker 6: My question was like if you have noticed the biggest mistake that startups make in their early stages, and what's your advice for future entrepreneurs to be avoided?
• 46:01
  Speaker 5: Okay.
• 46:02
  Speaker 5: So as I was thinking about that and listening to the answers, I think a couple of things here are really good.
• 46:07
  Speaker 5: One was about the pivoting and recognizing when you've got to change from the original vision and things like that, and then all of it was honesty.
• 46:14
  Speaker 5: But I want to touch on the guy that just left His spidey sense, is saying spidey sense, right.
• 46:22
  Speaker 5: Is saying oh, they like us, we're good for data, but is that going to lead to money?
• 46:28
  Speaker 5: How Right?
• 46:30
  Speaker 5: He isn't sure.
• 46:32
  Speaker 5: So I'm saying intellectual honesty with yourself is the first step, and then an additional, being honest with your investors and, you know, being ready to pivot and all the rest of it.
• 46:44
  Speaker 5: But you know, at the beginning I want to disagree a little bit with some of the other points in that, as much as it is your role as a founder to have the vision to be doing something that no one else thinks is necessary otherwise it would have already been done, okay.
• 46:58
  Speaker 5: Okay.
• 46:58
  Speaker 5: So it's your gut and your instinct and your intuition that leads to creating that innovation, Okay, but you also have to be a little bit bipolar.
• 47:08
  Speaker 5: You have to be able to okay, that didn't quite work or doesn't feel like we're going to make money on that path.
• 47:14
  Speaker 5: Stop think, get advice, talk to different people, pitch, pitch, pitch, listen, listen, listen and then listen again and then ask some more questions and then figure out what's going wrong.
• 47:29
  Speaker 5: And sometimes reflecting on what's being said can help you figure out the mistakes aware and what is actually the mistake.
• 47:36
  Speaker 5: I hope that helps.
• 47:39
  Speaker 4: Quick one point I would like to add.
• 47:41
  Speaker 4: This is what Jack Ma told the Chinese students, or someone, when he was addressing.
• 47:46
  Speaker 4: He said all this AI coming, don't be tools.
• 47:50
  Speaker 4: Go and learn music.
• 47:52
  Speaker 4: Learn art, because I think the way I took it is don't believe in Excel sheets only.
• 47:59
  Speaker 4: Art is still your own intellectual capability.
• 48:03
  Speaker 4: Rely on it, develop it.
• 48:06
  Speaker 4: Rely on it.
• 48:07
  Speaker 4: Don't just go with the Excel sheets.
• 48:11
  Speaker 3: The advice, I think, is don't die, because that's the only thing that is between your startup and success.
• 48:27
  Speaker 3: So just don't die, and it's going to be successful at some point in time.
• 48:34
  Speaker 2: So there is data on this too.
• 48:35
  Speaker 2: So the number one reason startups die is they run out of cash.
• 48:41
  Speaker 2: Number one reason is run out of cash, and if you're bootstrapping it, it's typically time right, life happens.
• 48:47
  Speaker 2: You need to go get job, salary, whatever, but number one reason is you ran out of cash.
• 48:51
  Speaker 2: And why you run out of cash are those five reasons I talked about earlier.
• 48:55
  Speaker 2: You scale those too quickly because in a startup, when you burn money, if you decide to create one feature, you have to dedicate time and money to figure out how you're going to sell that feature.
• 49:06
  Speaker 2: You're going to go find a customer.
• 49:07
  Speaker 2: You're going to need to support that feature.
• 49:09
  Speaker 2: It's not for free.
• 49:10
  Speaker 2: It's not just adding a feature, throwing it over the wall.
• 49:13
  Speaker 2: Every decision has an impact in another part of the business where you need to spend money, and most founders especially if you're a technical founder they just focus on the product, not on the cash burden and the customers.
• 49:25
  Speaker 2: So they find themselves frequently in a situation where your pre seed seed.
• 49:29
  Speaker 2: You might find enough technical people who are enameled, enamored by the technology to give you money, but you're gonna hit this brick wall.
• 49:38
  Speaker 2: At Series A, the number I didn't give out was last year.
• 49:47
  Speaker 2: It went from 25% to 12%, and most of the companies that we're seeing at Series A are pretty mature, coming in with almost a million in ARR.
• 49:51
  Speaker 2: So they're pretty far along, and if you're still making excuses for why you don't have customers and revenue, you're going to find it hard to raise money.
• 49:59
  Speaker 1: They say don't confuse a business.
• 50:00
  Speaker 1: You're not in a business if you're spending money.
• 50:05
  Speaker 1: You're in a business if you're making money.
• 50:07
  Speaker 1: Exactly so don't confuse yourself, just because, you're spending money, it doesn't mean you have a business.
• 50:12
  Speaker 1: It's when you're making money when you actually have a business.
• 50:15
  Speaker 3: You could have it in any color, as long as it's black, that's right.
• 50:18
  Speaker 1: That's right, exactly.
• 50:20
  Speaker 1: I want to get to Avijit's question out there.
• 50:22
  Speaker 1: Maybe this will be our final question.
• 50:29
  Speaker 8: Thanks, really good discussion.
• 50:30
  Speaker 8: I'm going to build on what Pete said, given the conversation we had as I looked at a lot of the companies out there.
• 50:37
  Speaker 8: And then to your point about working back from the customer and solving a problem.
• 50:41
  Speaker 8: What happens is, as a startup, you tend to solve a particular problem where there are a surrounding set of problems, whether it's in a vertical stack or a horizontal flow and the startup tends to be good at solving that one particular thing but then gets stuck because the rest of the things are not being solved by them.
• 51:01
  Speaker 8: But as a VC, if you can step back and look at the entire chain, whether it's the value chain or it's up and down the stack, and you could assemble a coalition of the willing, if you will, that are willing to partner with each other and solve it end-to-end, up and down the stack, wouldn't that give you more optionality, because each startup can solve a particular problem, but when pulled together they can solve a holistic problem to your point.
• 51:25
  Speaker 8: Would that strategy work for you where you step back, look at this and assemble things?
• 51:30
  Speaker 8: They can go complete standalone, piece by piece, but in some situations, if you add it all together, you could pull all pieces forward.
• 51:38
  Speaker 8: Is that something you could think about?
• 51:40
  Speaker 5: how you invest, you mean like an initiative.
• 51:44
  Speaker 5: Yeah, I mean Okay yeah.
• 51:49
  Speaker 5: Yes?
• 51:49
  Speaker 5: Short answer is yes.
• 51:50
  Speaker 5: Open GPU is a standard as the ability to fund projects, especially in research, also has the ability to certify, also has the ability to invest in companies that pick up the standard and have something really promising to deliver with it.
• 52:11
  Speaker 5: In other words, opengpu Ventures, right?
• 52:14
  Speaker 5: Yes, as a VC sometimes you have Sorry, as an investor, I shouldn't say VC as an investor sometimes you have the ability to do all the things to create a standard, to set up the path for the future, to put down the stones, to set up the petrol bows and charge people for the petrol along the way.
• 52:29
  Speaker 5: Everything is possible.
• 52:30
  Speaker 5: Yes, bigger picture helps, right.
• 52:33
  Speaker 5: But to the earlier point, yeah, perhaps we are in a really great setting here, with the edge in mind.
• 52:40
  Speaker 5: We are thinking ahead by being power efficient.
• 52:43
  Speaker 5: So you are in the place to think what all is necessary.
• 52:47
  Speaker 5: Okay, so use the chance to create your own initiative.
• 52:52
  Speaker 4: Completely agree with you, and especially, that's true with CVCs, because CVCs invested in you knowing the technology.
• 52:57
  Speaker 4: Is true with CVCs, because CVCs invested in you knowing the technology, being fascinated by the technology, not by the financial model.
• 53:04
  Speaker 4: So, if you are struggling, what the CVC?
• 53:06
  Speaker 4: Look at it.
• 53:07
  Speaker 4: Well, I like this technology, I like the founders.
• 53:10
  Speaker 4: I need him to succeed.
• 53:12
  Speaker 4: How do I help?
• 53:13
  Speaker 4: I'll probably directly help you, creating an ecosystem like that, or I'll connect you to people.
• 53:21
  Speaker 4: You know what.
• 53:22
  Speaker 4: You should probably work with this.
• 53:23
  Speaker 4: That's what we do because we want our investment to succeed.
• 53:26
  Speaker 4: That's very singular.
• 53:29
  Speaker 4: Go on, man.
• 53:32
  Speaker 5: Sorry.
• 53:32
  Speaker 5: I just want to add Sorry, when I'm saying think of initiative, I'm trying to say what he said more articulately, which is, think bigger picture and think where you can help a partner and then help someone else, and then blah, blah, blah.
• 53:45
  Speaker 2: So bigger picture.
• 53:46
  Speaker 2: So we do that and actually do that intentionally.
• 53:50
  Speaker 2: So at Jim Pulse out there is a horizontal platform goes across.
• 53:55
  Speaker 2: We have companies that have built smart sensors doing different things.
• 54:01
  Speaker 2: So when we look at companies, we also look at how can we stitch them together, particularly when we're selling into factories.
• 54:10
  Speaker 2: Because, again, the ecosystems are different the persona that they're targeting, the budget that they're targeting.
• 54:19
  Speaker 2: If it's the same, frequently it makes sense to put them together and take them in.
• 54:24
  Speaker 2: So we have vision systems going with somebody who's maybe managing a casting process.
• 54:29
  Speaker 2: So we have a company that is just focused on metal casting and another company that's focused on vision systems doing anomaly detection in metal factories.
• 54:38
  Speaker 2: So those types of things.
• 54:40
  Speaker 2: So we do do that.
• 54:41
  Speaker 2: But for us, also because we have LPs like Rockwell Automation, we actually take it a step further and actually figure out what inside of Rockwell's product ecosystem, their partner ecosystem, who else can the startup enable to do more?
• 54:58
  Speaker 2: And is the target customer the same as well?
• 55:01
  Speaker 2: So, again, the persona at the other end.
• 55:03
  Speaker 2: So we not only do a look at how we stitch our startups together, but also tie it back into our LPs.
• 55:12
  Speaker 3: Good, yeah, we do the same, but sometimes the supply chains are so long that it's mission impossible.
• 55:22
  Speaker 3: So you could do this in certain areas, but for some use cases, the supply chain is extremely long.
• 55:29
  Speaker 3: Actually, like Murata, I think your supply chain is extremely long.
• 55:33
  Speaker 3: So if you have something that would benefit from this along supply chain, then you should probably go to CVC, because these guys have everything.
• 55:45
  Speaker 1: Good, well, we are at time, but thank you all, let's give a round of applause to these guys.