Latest / EDGE AI POD / Beyond the Edge - Analyst Insights at AUSTIN 2025

Transcript

Beyond the Edge - Analyst Insights at AUSTIN 2025 · EDGE AI POD

• 0:07
  Speaker 1: All right, should we get started?
• 0:08
  Speaker 1: Yes, sounds good.
• 0:10
  Speaker 1: Bill, do you want to give a little intro on your Sure who you are and Is this thing working?
• 0:16
  Speaker 2: There we go Left mic.
• 0:17
  Speaker 2: Left mic.
• 0:18
  Speaker 2: Oh yeah, we're not in order.
• 0:21
  Speaker 2: Yeah, don't go by that.
• 0:25
  Speaker 2: I'm Bill Curtis.
• 0:27
  Speaker 2: I'm the analyst at More Insights and Strategy that tracks this kind of stuff.
• 0:32
  Speaker 2: My history goes back a long time, as you can tell by the gray beard.
• 0:37
  Speaker 2: Not enough gray beards on this panel One.
• 0:41
  Speaker 2: You should try it just for me.
• 0:44
  Speaker 2: There you go, why not?
• 0:46
  Speaker 2: It'll be 25 forever.
• 0:49
  Speaker 2: So for me, going back in time, it's ARM, and then, before that AMD, before that Dell, before that Halliburton.
• 0:57
  Speaker 2: And I actually did engineer systems for embedded systems back before we called them embedded systems.
• 1:09
  Speaker 4: My name is Stephanie Atkinson.
• 1:11
  Speaker 4: I have been running Compass Intelligence.
• 1:13
  Speaker 4: We turned 20 years old this year.
• 1:16
  Speaker 5: We're also an analyst firm, you turned 20 years old.
• 1:18
  Speaker 4: I do yes.
• 1:19
  Speaker 4: Can you tell no 21.
• 1:21
  Speaker 4: I got to be legal to drink the drink tonight, no 21.
• 1:25
  Speaker 4: I got to be legal to drink the drink tonight.
• 1:26
  Speaker 4: So we turned 20 years old this year and I also run Vmark, which is a marketing company for small and mid-sized businesses, Kind of where my two worlds collide is if I come across a small business that's also in the tech industry.
• 1:47
  Speaker 7: I'm your girl.
• 1:47
  Speaker 7: My name is Robert Quinn.
• 1:48
  Speaker 7: I'm a semiconductor industry influencer on LinkedIn.
• 1:51
  Speaker 7: I write and talk about the semiconductor industry.
• 1:53
  Speaker 7: I garner about 11 million eyeballs a year there.
• 1:57
  Speaker 7: I also do consulting and formally I've worked in the semiconductor industry.
• 2:02
  Speaker 7: I've worked as an equipment engineer, process engineer, traveling the world, having worked at all the fabs, from global to Skywater to Intel to Infineon and NXP and all the different fabs that are making these chips for us.
• 2:19
  Speaker 7: So having a deeper understanding of the technology of process engineering and chips and how we're making these chips, and I like to translate that information into palatable information for people to understand this technology, where it's going and what we're doing with it.
• 2:35
  Speaker 7: I was telling them earlier that I try to speak to my 6th grader the same way I speak to CEOs.
• 2:44
  Speaker 7: That's usually the best way to do it.
• 2:46
  Speaker 5: Same brain power right Well a funny story.
• 2:52
  Speaker 7: I was actually in Munich at the CEO roundtable and somebody said what do you think about neuromorphic computing?
• 2:57
  Speaker 7: We had seven of the biggest CEOs sitting in front of us at this roundtable.
• 3:01
  Speaker 7: Every single one of those CEOs looked at each other and said what's neuromorphic computing?
• 3:04
  Speaker 7: There you go.
• 3:05
  Speaker 7: I said we've got a problem.
• 3:07
  Speaker 7: They don't speak our language.
• 3:07
  Speaker 7: So trying to translate that into, to make it palatable information so people can understand what we're doing and how we're doing it and understanding this technology and the growth of it.
• 3:17
  Speaker 7: So but it's fun to write about, talk about and it's it's great to be here today and it's good to meet you.
• 3:25
  Speaker 5: Howdy, I'm Rob Tiffany.
• 3:26
  Speaker 5: I'm a research director at IDC.
• 3:28
  Speaker 5: I'm kind of newish to the industry analyst game, but you probably know me.
• 3:34
  Speaker 5: I did co-create the Internet of Things with Al Gore back in the early 90s, right after the ARPANET thing got released to everybody, and so that's been cool.
• 3:44
  Speaker 5: I did know this girl over here.
• 3:47
  Speaker 5: She used to interview me at Mobile World Congress for Windows Mobile stuff.
• 3:51
  Speaker 5: That's how long your business has been around, right.
• 3:54
  Speaker 4: Global Comm, huh, global Comm, oh yeah, back when.
• 3:58
  Speaker 5: Yeah.
• 3:58
  Speaker 5: So Pete and I spent a lot of time doing mobile stuff at Microsoft Windows Mobile, windows Phone and it didn't work out, unfortunately, but we gave it a good shot.
• 4:09
  Speaker 5: We had some tiles on a phone, so that was kind of cool.
• 4:12
  Speaker 5: But thank God for those phones because they made small little sensors and chips affordable.
• 4:18
  Speaker 5: The smartphone revolution had a side effect that helped all these people out here and all the devices.
• 4:24
  Speaker 5: That may not have happened if it wasn't for the might of Apple and Samsung and a lot of players with giant money pushing that forward.
• 4:30
  Speaker 5: So lots of love to those folks.
• 4:32
  Speaker 5: Oh, and cellular is kind of cool too.
• 4:36
  Speaker 5: Gosh helped build.
• 4:37
  Speaker 5: I designed Azure IoT.
• 4:39
  Speaker 5: That was fun building an IoT platform around the world.
• 4:42
  Speaker 5: I was one of the architects of that.
• 4:43
  Speaker 5: Then I built industrial IoT and digital twins at Hitachi, and so I was in Japan all the time, and then I got to go to Sweden all the time when I was at Ericsson doing connection management type stuff with the mobile devices over 5G.
• 4:57
  Speaker 5: So there you have it and here we are Awesome yeah.
• 5:01
  Speaker 1: Good bios, all right, and you're kind of all local Texans, I think right.
• 5:05
  Speaker 5: We are all local Texans.
• 5:06
  Speaker 5: So there you go From Redmond Washington.
• 5:11
  Speaker 2: True to our Austin location here.
• 5:12
  Speaker 2: Well, and I am the only Austin native here, all right.
• 5:15
  Speaker 1: I did go to school here.
• 5:16
  Speaker 1: I'm legit, that's not native.
• 5:18
  Speaker 2: All right, all right.
• 5:21
  Speaker 1: Sounds good Texas fight Sounds good Extra spite, so let's get into it.
• 5:23
  Speaker 1: So, like one of the things you know that we talk about all the time and we've been sort of through the trenches on things you know IoT the question I get a lot is how do we avoid edge AI from becoming what happened to IoT, which I would say?
• 5:39
  Speaker 1: The metaphor is like you know those box of cables you have in your garage, that's what happened to IoT.
• 5:44
  Speaker 1: Right, it became like this those box of cables you have in your garage, that's what happened to IoT.
• 5:46
  Speaker 1: Right, it became like this big box of cables.
• 5:46
  Speaker 1: And like how do we make sure Edge AI does not suffer the same fate as that big box of cables in your garage?
• 5:54
  Speaker 5: Kind of like a big nerd techie box of cables.
• 5:57
  Speaker 5: Yeah, everyone's got those box of cables right.
• 5:59
  Speaker 5: You're working in the garage Some connectors and stuff you don't know how to sell Parallel ports and stuff.
• 6:03
  Speaker 9: you don't know how to sell ports and stuff how to do marketing.
• 6:04
  Speaker 1: No, so like what?
• 6:06
  Speaker 1: How do we, how does edge ai avoid the same fate and actually kind of hit the right scale to get deployed and to kind of be a real business?
• 6:14
  Speaker 5: well, I'm going to tell you, do your edge ai stuff here, but don't do any of this stuff in front of actual customers, because it looks like a bunch of cables out here.
• 6:24
  Speaker 5: But this is the home crowd, right, but it has to do with your marketing and sales and simplicity and you're selling an outcome, right, and so you really shouldn't be talking about that stuff.
• 6:38
  Speaker 5: But that's what we did at IoT.
• 6:40
  Speaker 5: We kept talking about all the geek stuff, right, and thought that that was so important how we're connecting.
• 6:46
  Speaker 5: Do we have digital twins?
• 6:47
  Speaker 5: Are we using MQTT or AMQP?
• 6:51
  Speaker 5: Who cares?
• 6:51
  Speaker 5: Nobody cared.
• 6:53
  Speaker 5: In the end, nobody cares.
• 6:54
  Speaker 5: But we had all these endless conferences where that's all we talked about.
• 6:57
  Speaker 5: The customer doesn't care.
• 6:59
  Speaker 5: So, remember, the customer just cares about some positive outcome.
• 7:03
  Speaker 5: And if you never mention your edge AI and your solution, also, what things really worked well in IoT.
• 7:09
  Speaker 5: Everybody built these platforms saying, come, use my platform for all your stuff.
• 7:13
  Speaker 5: But the stuff that really succeeded were people who built just singular products like the Nest thermostat or the Ring doorbell.
• 7:24
  Speaker 5: They built an end-to-end thing and we're selling this deal and you don't even know what it's all about.
• 7:27
  Speaker 5: Those guys had success.
• 7:28
  Speaker 5: Other people maybe not so much, but you guys are smarter than I am.
• 7:32
  Speaker 5: You got a better answer than I do.
• 7:34
  Speaker 2: No, you're absolutely right, and that points to what the problem really is with IoT Diversity it's diversity.
• 7:42
  Speaker 2: If you want to make an IoT solution that scales across the entire industry, then you're talking about something like a phone or a PC that's big enough and powerful enough to support a whole bunch of different applications.
• 7:55
  Speaker 2: But that's not the way we design IoT devices.
• 7:58
  Speaker 2: No, we design IoT devices or OT devices operations technology in IIoT terms or OT devices operations technology in IIoT terms.
• 8:05
  Speaker 2: We design these things for specific tasks and so, therefore, every one of them is custom.
• 8:13
  Speaker 5: Now, custom sounds real good, that's a dirty word.
• 8:15
  Speaker 2: It is because it destroys scale.
• 8:18
  Speaker 2: Yeah, it means cost and it means no scale.
• 8:20
  Speaker 2: That's what happened to IoT.
• 8:21
  Speaker 2: That's why your question is spot on, that we don't want this to happen to the world of the edge technologies we're talking about here.
• 8:29
  Speaker 2: Now that they're AI powered, you can't afford the luxury of writing everything in C and bare metal engineering.
• 8:39
  Speaker 2: You need a platform that's got a real OS and something that other people have to support, with OTA updates.
• 8:47
  Speaker 2: Then you want to write application code on top of that.
• 8:50
  Speaker 2: Until we get there, we're still going to be in the IoT world of customization.
• 8:58
  Speaker 4: I think the biggest thing that I notice is I specialize in understanding what businesses want and how they consume and adopt and scale technology.
• 9:06
  Speaker 4: So the applicability of the solution and a lot of what we saw with IoT was we're going to build this mass solution that's going to provide 20 different, it's going to serve 20 different things right.
• 9:19
  Speaker 4: But the biggest thing is that they didn't really lead with the customer, with the business value, the value proposition or the outcome.
• 9:26
  Speaker 4: As Rob mentioned and as we start to think about technology in general, if we keep going down that path, we're not going to find the customers.
• 9:34
  Speaker 4: We're not speaking their language.
• 9:35
  Speaker 4: They don't want to hear about you know gadgets and technology.
• 9:40
  Speaker 4: They want to know what you're going to do for their business and how it's going to save them time and how it's going to improve their quality right If it's production or manufacturing or industrial, how it's going to help their workforce, simplify you know different processes or application back office, front office.
• 9:57
  Speaker 4: So we have to really get to the point of where we're really speaking the language of the customer, of the industrial customer or an enterprise customer, or in many cases, it might even be a city.
• 10:08
  Speaker 7: I like the way you talk about that box of things.
• 10:11
  Speaker 7: Edge AI is going to fill that box of things.
• 10:15
  Speaker 7: We're going to all replace our devices.
• 10:19
  Speaker 7: We're going to be replacing all of our devices because we're going to require new hardware and so all those devices that we currently carry in our pockets our phones, our laptops, our everything well, everything's going to be living on the edge in our devices, which means the old stuff is going to go in that box.
• 10:35
  Speaker 7: So what we can do to prevent that box from having that box?
• 10:39
  Speaker 7: Well, that box is fixing to get really full really fast yeah.
• 10:43
  Speaker 1: Let me give you one of the one of the things we always talk about is scenarios, and you know we'd say connecting AI to the real world.
• 10:49
  Speaker 1: That's my definition of edge AI is edge AI in the real world, where the data is being created.
• 10:53
  Speaker 1: And so let me give you this like non-hypothetical scenario and you tell me how edge AI can help solve it.
• 10:59
  Speaker 1: So we've all had experience with the McDonald's ice cream machines, right?
• 11:03
  Speaker 1: Those things never freaking work.
• 11:04
  Speaker 1: So, like, how could Edge AI solve the McDonald's ice cream machine dilemma?
• 11:09
  Speaker 1: Anybody have any ideas?
• 11:11
  Speaker 1: I don't think McDonald's is in the audience.
• 11:13
  Speaker 1: Speak freely.
• 11:16
  Speaker 7: I think we had a little discussion about this earlier.
• 11:18
  Speaker 7: I was talking about how we're going to be putting sensors on everything right and moving forward.
• 11:23
  Speaker 7: What is the AI is data, analyzing data, and so what does that mean?
• 11:28
  Speaker 7: That we're going to be doing lots of MEMS, lots of manufacturing of MEMS, sensors and I love seeing all the sensors devices that we're doing out here and the new creation of sensors that we're coming up with.
• 11:40
  Speaker 7: But everything is going to get that sensor, including the McDonald's ice cream machine.
• 11:46
  Speaker 7: So, yes, will that, will that that device will, will will analyze that that motor inside of that device and make sure that it is, uh, it is running it at a certain frequency and when it doesn't, the ai is going to analyze that data and then send that data back to the shop saying, hey, you've got to replace this ice cream machine so it is available for people to use on Saturday, because I know you go into the McDonald's every single time.
• 12:13
  Speaker 7: What happens?
• 12:13
  Speaker 7: Ice cream machines down.
• 12:14
  Speaker 7: Doesn't work.
• 12:16
  Speaker 5: GP, I've never had ice cream at McDonald's before because it's always broken.
• 12:20
  Speaker 5: You know, there's this other stuff called computer vision.
• 12:25
  Speaker 5: Maybe there should be a camera looking at the people sabotaging those machines.
• 12:30
  Speaker 5: I think that's what's really going on here Conspiracy, it's not a machine failure uptime thing.
• 12:36
  Speaker 4: And you know, I think the funny thing about this is sometimes it doesn't require a sensor, it's just a phone call.
• 12:40
  Speaker 4: You know, sometimes it's just a manual alert or something really basic.
• 12:46
  Speaker 4: I think sometimes we'll get to the point where we are over analyzing, we want to put sensors on everything and for something like that, and maybe it is predictive maintenance, I don't know, I don't know what the main machine was working is there?
• 12:57
  Speaker 4: is there, like this horizontal number one reason why these things are failing?
• 13:01
  Speaker 4: If that's the case, then maybe it is a sensor that's alerting someone to be dispatched to repair that.
• 13:07
  Speaker 2: Yeah that's the technical problem, but there's another problem that we can address with AI, and that's the people that are operating the McDonald's.
• 13:18
  Speaker 2: The people operating the McDonald's are good at making hamburgers for you and french fries, but they really are not technologists, they're not engineers.
• 13:26
  Speaker 2: They don't know how to operate or how to maintain this stuff.
• 13:28
  Speaker 2: So having a machine that's smart enough to help them do that and to talk them through the process of analyzing and repairing problems and making sure that they don't screw up accidentally, making the stuff that's in the ice cream machine unsafe to eat Stuff like that is really valuable.
• 13:47
  Speaker 2: So it's not the individual sensors and stuff that's important too, but it's a higher level function sitting on top of the whole thing operating the ice cream machine and maintaining it in lieu of having an engineer sitting there Sounds like a solution for a humanoid robot.
• 14:02
  Speaker 1: maybe.
• 14:02
  Speaker 5: Bill, I thought every company was going to become a software company, right.
• 14:07
  Speaker 4: And the fast food places are using lots of robots these days.
• 14:13
  Speaker 5: So maybe we should take human out of the equation.
• 14:14
  Speaker 5: I just said, remember, software is eating the world.
• 14:17
  Speaker 1: Yeah, and they stared at me like I'm predicting that at Jensen's keynote at GTC, he's going to have a humanoid robot making ice cream.
• 14:26
  Speaker 4: I think that's what's going to no, they're dancing now and doing backflips, yeah, but they're not making ice cream that's not productive.
• 14:32
  Speaker 1: Making ice cream is productive to bill's point adoption.
• 14:35
  Speaker 7: Right, what is, what is adoption going to be?
• 14:37
  Speaker 7: And and how, how is, uh, how is corporate america people, how you know, how are we going to adopt all this new technology?
• 14:46
  Speaker 1: Right, right, yeah, there has to be a willingness.
• 14:48
  Speaker 1: We were talking to some partners out there about there's no shortage in every Fortune 500 company.
• 14:54
  Speaker 1: They have a list of really difficult problems to solve, but the solution that's using AI has to sort of fit inside the return on investment envelope that they have, right.
• 15:06
  Speaker 1: So you might come up with this cool solution and it's got this you know depth sensing, lidar thing going on and they're like that's cool but it's way too expensive and I can't.
• 15:16
  Speaker 1: I need I need a hundred thousand of those and I can't do it.
• 15:19
  Speaker 1: So that's one of the things is like you know what's the challenges of really the differences between commercializing the tech versus you know, developing tech?
• 15:27
  Speaker 1: We saw some great talks today, for example, from lots of great you know universities, right, talking about a lot of the kind of theoretical things, a lot of the research that they're doing, but at the end of the day, these things have to get commercialized.
• 15:40
  Speaker 1: So what do you see that's going to help accelerate, maybe, commercialization of edge AI, like what are the factors or technologies or trends out there?
• 15:48
  Speaker 1: I already have an idea what the answer is.
• 15:50
  Speaker 1: I'm wondering if your answer is the same as mine.
• 15:54
  Speaker 5: It's like dead silence up here.
• 15:55
  Speaker 5: Monetize something, what?
• 15:57
  Speaker 5: That's no fun.
• 15:59
  Speaker 5: If there's money at the other end of the tunnel, that'll help accelerate what you're talking about.
• 16:04
  Speaker 5: Yeah, find a real problem to solve, go ahead.
• 16:08
  Speaker 4: I usually will tell the customers that I'm working with.
• 16:10
  Speaker 4: I'm like, go ahead and work with one customer and solve that problem and if it's something that is affordable and they see that in a year it's going to pay for itself or it's got to save them money, it's got to improve efficiency, it's got to improve quality, but in the end, it's got to be affordable for them.
• 16:33
  Speaker 4: So the metrics the financial metrics, I think are very, very important part of the commercialization, Sure, and I also think that you can build a use case, but it may not be economical Right and I think there's a lot of that that's happening today and that's a lot of what happened with IOT as well.
• 16:50
  Speaker 4: It wasn't, it wouldn't scale Right and it wasn't affordable.
• 16:54
  Speaker 4: And the customization you know that term when you say custom.
• 16:58
  Speaker 4: Although this is where we are today, we have to build things for specific industries and try to look at some of the factors that might go across maybe two or three industries and really work on our focus on those three areas, Because if we try to launch something and we're launching it across seven or eight vertical markets, you're not gonna make money and you're not gonna find those wins that can take you to the next level.
• 17:23
  Speaker 2: That's right and that's why the diversity IoT solutions, I think, kind of works to the disadvantage of AI diffusion.
• 17:34
  Speaker 2: But there are some big ones, there are some big rocks that are going to move things a long way.
• 17:42
  Speaker 2: For example, imagine what happens when large enterprises start connecting their ERP systems, their business improvement systems, their supply chain management systems to increasingly intelligent edge devices that are all connected.
• 18:00
  Speaker 2: That really provides the ground truth for turning those applications from reactive to proactive.
• 18:08
  Speaker 2: That can generate enough ROI to power the entire industry and everybody in this room.
• 18:15
  Speaker 2: That one use case, that one concept of using smart AI to provide the ground truth for AI-powered enterprise applications.
• 18:27
  Speaker 5: Is that how we get intelligent devices?
• 18:28
  Speaker 5: They have to have AI in them, right?
• 18:30
  Speaker 5: Because we've done this stuff forever.
• 18:33
  Speaker 5: We have lots of dumb devices.
• 18:34
  Speaker 5: In fact, we have way more dumb devices in the world than smart ones.
• 18:37
  Speaker 1: Way more disconnected.
• 18:38
  Speaker 2: Yeah, but this is going to create demand for the dumb devices as well, because AI is everywhere.
• 18:43
  Speaker 2: Apply AI to, for example, a supply chain problem and some really small ambiently powered 10-cent cost device.
• 18:56
  Speaker 2: Really, a dime 10-cent cost device stuck on a product making its way through a supply chain can give you an immense amount of value.
• 19:05
  Speaker 2: The value is actually realized by the fact that you know where the palette or the product is and you have the AI in big blank in the cloud somewhere that's going to figure out ooh, this isn't where it's supposed to be.
• 19:19
  Speaker 1: I heard a scenario about it's kind of similar to the supply chain, using ambient IoT to sort of track produce, for example, and the freshness of produce and things like that.
• 19:30
  Speaker 1: And then what's cool is then the person said and this is where sort of they jumped.
• 19:33
  Speaker 1: The shark was like oh, and then you connect this system to like an LLM and so you can talk to your bananas and ask them, like how fresh are you and where were you picked, and you know all that other stuff.
• 19:42
  Speaker 1: So you know it's sort of like if you bend it it's funny, but if you break it it's not funny.
• 19:47
  Speaker 2: Yeah but that example is actually doable.
• 19:51
  Speaker 2: It is, if you think about the ground truth data making its way up to the point where you can run an LLM.
• 20:00
  Speaker 1: Yeah, and you could talk to your family.
• 20:01
  Speaker 5: I'm going to start a new grocery store tomorrow where people walk up and talk to the bananas.
• 20:05
  Speaker 5: I think that's going to be a hit.
• 20:06
  Speaker 5: Are you fresh?
• 20:08
  Speaker 5: There you go.
• 20:09
  Speaker 2: How are you feeling today?
• 20:10
  Speaker 2: How are?
• 20:11
  Speaker 5: you feeling Don't eat me.
• 20:16
  Speaker 1: No, the bananas become sentient at that point.
• 20:18
  Speaker 1: And then there's a whole other issue.
• 20:21
  Speaker 1: That's not good.
• 20:23
  Speaker 5: Maybe you're kind of now.
• 20:25
  Speaker 5: You've got me going down a rabbit hole here, because when you talked about putting the dime thing on a package, it reminds me of the early days of RFID and it used to be too expensive.
• 20:35
  Speaker 5: And so you do it at the pallet level, right, but I'm imagining at the pallet level of the bananas I might have enough money for one AI thing, for a pallet or a whole section in the grocery store, but not at the banana level.
• 20:49
  Speaker 1: We haven't gotten there yet.
• 20:50
  Speaker 5: We haven't gotten there yet the banana level.
• 20:53
  Speaker 1: But I'm sure, like robert you know, like the advances in semiconductors and sensors and stuff, we'll get to the banana level at some point.
• 20:59
  Speaker 1: Right, so we're definitely.
• 21:01
  Speaker 7: You talk about marketability and like being able to make a product.
• 21:05
  Speaker 7: I've I've been in through the startup world.
• 21:07
  Speaker 7: I've worked in the world of startup for the last four years and, uh, going through the Series A and raising funds and working with all that and I've seen incredible.
• 21:17
  Speaker 7: I have clients that have incredible technology and so many people think that if you build it, they will come.
• 21:23
  Speaker 7: That's not the case.
• 21:24
  Speaker 7: Yeah, it's never been that.
• 21:25
  Speaker 7: That is not the case.
• 21:26
  Speaker 7: I've seen the most incredible technology semiconductor industry technology out there that is absolutely out of this world, revolutionary technology that nobody's using.
• 21:36
  Speaker 7: Nobody has.
• 21:37
  Speaker 7: But yet the guy sold two units in two in 20 years.
• 21:40
  Speaker 7: Sure?
• 21:41
  Speaker 7: because, he just doesn't have the ability to market his products to do.
• 21:46
  Speaker 1: There's so many more aspects to right than just well, they say in tech you can sell 20,000 of anything.
• 21:51
  Speaker 1: Basically, that's pretty much, you know.
• 21:54
  Speaker 1: I mean, I worked on the Zoom, so you know.
• 21:57
  Speaker 5: You sold 20,000 Zooms yeah at least how many.
• 22:00
  Speaker 5: Kens.
• 22:02
  Speaker 1: Probably less.
• 22:04
  Speaker 4: I was going to say, too, is.
• 22:05
  Speaker 4: I think I want to always caution everyone in our industry because, just like with the IoT hype, we're now in that AI hype, so all of the failed IoT companies are now going to be AI companies.
• 22:18
  Speaker 4: It's happening, it's already happening, and we see a lot of vaporware, we see a lot of trials and POCs, but we don't really see any meat no scale, no implementation.
• 22:30
  Speaker 4: Like what are you doing?
• 22:31
  Speaker 4: Where are you making money?
• 22:33
  Speaker 4: Do you have a customer?
• 22:34
  Speaker 4: Those are the hard questions you have to ask some of these partners or other companies that you might be talking to.
• 22:41
  Speaker 4: And the other area is the gen ai hype.
• 22:44
  Speaker 4: Right, every, every device that we have our smartphones, our laptops, our computers um, we have ai, ai.
• 22:52
  Speaker 4: It's attaching to every software out there today, but customers are not going to pay for it in the end, they are going to expect it to be part of that software application.
• 23:03
  Speaker 4: Now, what do you do?
• 23:05
  Speaker 4: You're going to get them to pay for it another way.
• 23:07
  Speaker 4: There's always tricks to that from a sales and marketing perspective, but they're not going to say, oh, I'm going to tack on $2 a month and I'm going to start paying for that.
• 23:15
  Speaker 5: Did you say that the AI PCs didn't create a new super cycle for Microsoft?
• 23:19
  Speaker 5: Absolutely not, it didn't happen?
• 23:21
  Speaker 4: It's not happening.
• 23:22
  Speaker 4: You know they may pay more for that device because of all these new features yeah, or maybe the bananas get expensive, no.
• 23:32
  Speaker 4: So that's kind of where we are right now, and I think that we have to caution ourselves and also a lot of the AI that's happening out there is not AI, it's truly machine learning and it's not reaching that level of where we we will get to.
• 23:45
  Speaker 1: Let me ask a little question about the cloud.
• 23:47
  Speaker 1: And you know there's the cloud and then the edge.
• 23:50
  Speaker 1: You know some people say that the the edge and edge AI is kind of the edge of the cloud.
• 23:55
  Speaker 1: I would say it's actually the inverse the cloud is kind of the big part of the edge.
• 24:00
  Speaker 1: That's my bias.
• 24:01
  Speaker 1: But what is the role of the cloud and how can we leverage the cloud in edge AI better?
• 24:08
  Speaker 1: Where's the architecturally?
• 24:10
  Speaker 1: How do cloud DevOps and things like that help us?
• 24:12
  Speaker 1: To your point, bill, about how we create a little more standardization in the ecosystem to get deployments out faster that's a fantastic question.
• 24:20
  Speaker 2: That's exactly the right question that we should all be asking in this room.
• 24:23
  Speaker 2: Now.
• 24:23
  Speaker 2: There are two edges.
• 24:26
  Speaker 2: By my definition anyway, you can argue there are 20, but let's just think about two.
• 24:32
  Speaker 2: One is the IT edge.
• 24:34
  Speaker 2: In an enterprise perspective, that's where you can use cloud-native techniques and you can use traditional IT programming all the way out to the edge.
• 24:42
  Speaker 2: The IT edge, you're using the same languages, tools, programming techniques, everything all the way out to that point.
• 24:53
  Speaker 2: Beyond there there'd be dragons.
• 24:55
  Speaker 2: It's a different world.
• 24:57
  Speaker 2: The world of operational technology is completely different.
• 25:01
  Speaker 2: You might say, oh no, no, no.
• 25:03
  Speaker 2: What we want to do is make these little devices just like little bitty IT devices.
• 25:08
  Speaker 2: Well, that doesn't work, and we've spent the last decade in the world of IoT trying to prove that, and we have proved it.
• 25:13
  Speaker 2: It doesn't work.
• 25:13
  Speaker 2: And we've spent the last decade in the world of IoT trying to prove that, and we have proved it.
• 25:16
  Speaker 2: It doesn't work.
• 25:17
  Speaker 2: So OT is always going to be different.
• 25:21
  Speaker 2: So fortunately, now we're seeing a change even in the big cloud service providers and the ERP guys and all that.
• 25:28
  Speaker 2: They're no longer trying to push their heavyweight stuff down into the OT world.
• 25:33
  Speaker 2: Instead, they're beginning to define more intelligent interfaces and lighter weight stuff.
• 25:41
  Speaker 2: I was talking with the guys at Edge Impulse out there earlier today.
• 25:42
  Speaker 2: They have a really good demo about using AWS on the back end of their AI-enabled microcontroller stuff Really well actually it's microcomputer stuff the one that I was looking at but anyway it was an excellent demo and it shows that the in this case green grass stuff that gets put on the edge on the OT equipment is way smaller than it used to be back in the days when that stuff was all written in Java Very heavyweight stuff and it required re-architecting the OT devices.
• 26:19
  Speaker 2: Now they have a much lighter weight approach and Microsoft is doing essentially the same thing.
• 26:24
  Speaker 2: They're pulling back on the heavyweight stuff they used to put on the devices, replacing it with hey, let's just grab the data, let's do MQTT, let's have a fairly simple and lightweight interface on the device.
• 26:36
  Speaker 2: So, taking a step back from that, I'm saying there's two edges.
• 26:40
  Speaker 2: The IT edge stops at the IT boundary and we're in the early days now of defining the interface between the OT stuff and the IT stuff and I think that's real progress.
• 26:51
  Speaker 4: Not so fast, One second.
• 26:53
  Speaker 4: I was going to say something.
• 26:54
  Speaker 2: I was waiting for that.
• 26:55
  Speaker 2: I know he's about to disrupt.
• 26:56
  Speaker 4: So I wanted to say last week I saw something about the carriers are talking about edge of network, right.
• 27:04
  Speaker 4: So it's like everyone has to put the word edge in something.
• 27:07
  Speaker 4: So now it's edge of network, so microcells, smaller regional cell sites, that kind of thing.
• 27:11
  Speaker 4: So that's their edge.
• 27:12
  Speaker 4: So there are more flavors coming.
• 27:15
  Speaker 1: We're trying not to talk about telco at this conference.
• 27:17
  Speaker 5: I know Don't talk about telco Trying to avoid that they're still trying to figure out how to make money.
• 27:20
  Speaker 1: We're detoxing from telco here, but go ahead, rob yeah.
• 27:23
  Speaker 4: Go Rob.
• 27:24
  Speaker 5: Well, they tell you about the cloud.
• 27:25
  Speaker 5: I remember back when Dave build.
• 27:28
  Speaker 5: Azure, and it was tough, man.
• 27:30
  Speaker 5: You know we're in the room with Ray Ozzie, remember.
• 27:32
  Speaker 9: We were figuring out the whole deal and it's like this is going to be big.
• 27:39
  Speaker 5: And then there's all those PCs and whatever out there.
• 27:42
  Speaker 5: There was this kind of talk down to kind of view that the cloud was a superset of the edge and it controlled the edge and the edge was just a part of the whole continuum.
• 27:51
  Speaker 5: But I certainly saw the other side, where edge people are like well, maybe I never need to go there or maybe the cloud is just for a broader view of things and the edge might be a more localized view.
• 28:02
  Speaker 5: And yes, it could be the IT OT thing like you're seeing in manufacturing and factories.
• 28:07
  Speaker 5: But there's lots of other weird edges and I remember when an edge didn't mean there was the IoT device, the endpoint, and that was not the edge.
• 28:17
  Speaker 5: There was this edge thing that was kind of like a gateway.
• 28:20
  Speaker 5: I remember thinking is this like a Cisco router or something at the edge of the factory or something I don't know?
• 28:25
  Speaker 5: And a lot of manufacturers like Dell and HPE made edge ruggedized little PCs and they would talk to machines close to the data.
• 28:33
  Speaker 5: But we didn't think of the edge as being the machines.
• 28:35
  Speaker 5: But the people keep getting smarter OEMs who are making OT devices, machines.
• 28:42
  Speaker 5: At some point they'll bake some of that stuff in and then it'll probably get really weird, and some of them are already doing it.
• 28:48
  Speaker 5: What's your smartwatch?
• 28:53
  Speaker 6: And then what's?
• 28:54
  Speaker 5: my smart phone.
• 28:56
  Speaker 5: Well, my smart phone is connected via Bluetooth and this is an edge gateway to route my watch to the interwebs, right, but also?
• 29:06
  Speaker 5: But I've got an intelligent edgy kind of thing here too which I used to just think.
• 29:11
  Speaker 5: You know, I don't know, I don't know, I just can't live with the two edges.
• 29:19
  Speaker 2: When you're writing software for the watch, though, you're not using the same techniques you are when you're writing the application in the cloud that uses it.
• 29:25
  Speaker 2: And that's my point is that the techniques that you use to build this stuff and the architecture internally is different.
• 29:31
  Speaker 5: You don't have Kubernetes on your smartwatch.
• 29:35
  Speaker 2: I guess it's theoretically possible but, I, wouldn't recommend it.
• 29:39
  Speaker 4: I think the bottom line is, as we really start to think about AI in operations and production in cities, in real-world applications, is that there will be specific applications and use cases that will require some kind of hybrid edge cloud architecture and we don't have to send everything out to the cloud.
• 30:00
  Speaker 4: But the AI does mean that we're going to be dependent on cloud architecture and I think that that's happening now with edge.
• 30:08
  Speaker 4: You know, I think edge is really driven around low latency, real time, actionable intelligence.
• 30:14
  Speaker 4: Right, we want to make some, we want that information to provide some level of intelligence that's going to alert us, it's going to automatically do something.
• 30:23
  Speaker 5: But it didn't always have to be AI, did it?
• 30:25
  Speaker 5: No Like, remember when you're the first baby steps of making your endpoint edge whatever thingy smarter, of making your endpoint edge whatever thingy smarter, the first thing was filtering, or hey, if the new packet of data from taking a sample from the sensors looks like the one I just sent, maybe I won't send it.
• 30:42
  Speaker 5: The big data guys will be angry because they want all the data, but maybe this is noise.
• 30:47
  Speaker 5: And then you kind of went ahead and so like when I was doing all that agriculture stuff and I've got moisture sensors and things like that, you can make some decisions right there.
• 30:57
  Speaker 5: To decide I need to actuate the irrigation system for this block of an apple orchard or for something like that.
• 31:04
  Speaker 5: And even then it was still if this than that.
• 31:06
  Speaker 5: It was still threshold stuff, but it was actually happening on the device.
• 31:10
  Speaker 5: Do we get?
• 31:10
  Speaker 5: Does that count?
• 31:11
  Speaker 5: Does that not count?
• 31:11
  Speaker 5: Maybe not?
• 31:12
  Speaker 5: You know expert systems used to be called ai and that was essentially rules engines, right?
• 31:18
  Speaker 5: So don't feel bad, all you rules engine people out there, it's still ai.
• 31:22
  Speaker 5: You're probably already calling that in your sales presentation.
• 31:25
  Speaker 1: So whatever, yeah well, I guess one of the questions is like I mean, you were saying there's the ot it thing and but also there's, you know, everyone who's building these kind of edge systems, edge AI systems, need to think about how they're going to plug into a bigger cloud architecture, right, I mean, because ultimately, these things are going to connect.
• 31:43
  Speaker 1: Very few things are completely disconnected all the time, right?
• 31:47
  Speaker 1: So it's like how do you think about how you're a first class citizen in some sort of cloud architecture?
• 31:52
  Speaker 5: Oh, I got one.
• 31:53
  Speaker 5: Go for it.
• 31:54
  Speaker 5: Building management systems.
• 31:55
  Speaker 5: I've been in skyscrapers and they're using, like Niagara or different systems like that.
• 31:59
  Speaker 5: Guess what.
• 32:00
  Speaker 5: That whole sucker has to work offline in that skyscraper and cannot be dependent on a cloud, right?
• 32:06
  Speaker 5: Because?
• 32:07
  Speaker 6: guess the.
• 32:07
  Speaker 5: Internet goes down all the time or power goes out and they still got to manage all the elevators.
• 32:12
  Speaker 5: Yeah, there's like occasionally connected systems.
• 32:14
  Speaker 5: It's occasionally, that's what.
• 32:16
  Speaker 6: I used to say in mobile.
• 32:17
  Speaker 1: Energy and gas right and mining and you know these other things.
• 32:19
  Speaker 1: But ultimately, like if you're managing that kind of system, you want to manage it holistically, from the cloud to the edge right, and so your edge devices and even down to your sensor level.
• 32:28
  Speaker 1: They need to show up as like resources in an overall kind of framework of resources.
• 32:33
  Speaker 2: Show up, yeah, but you don't necessarily need to manage them.
• 32:35
  Speaker 2: In other words, because you're connecting an OT device to an IT system does not mean that that IT system needs to update that device and be responsible for its security in perpetuity.
• 32:45
  Speaker 2: Instead, that's the purview of the OT world and that's the mistake that everybody's always made about IoT the OT world and that's the mistake that everybody's always made about IoT Thinking that the OT device is going to be part, an integral part, of the IT universe.
• 33:00
  Speaker 2: And it's just not.
• 33:02
  Speaker 5: Is that those OT guys with the hard hats?
• 33:04
  Speaker 5: Yeah, writing Python.
• 33:06
  Speaker 2: Well, no, they don't have to.
• 33:07
  Speaker 2: That's the whole point.
• 33:08
  Speaker 2: Okay, you're not going to use cloud native stuff down there, you're going to do it some other way.
• 33:11
  Speaker 2: Well, and the?
• 33:12
  Speaker 4: cybersecurity guys are going to just be screaming at everyone, right, because of risk in hacking, and it's like so IT ends up becoming involved, no matter what.
• 33:22
  Speaker 2: Yeah, but nothing really changes.
• 33:24
  Speaker 2: In other words, you have to solve those problems.
• 33:25
  Speaker 2: It's how you do it where we always make the mistake.
• 33:28
  Speaker 2: You don't do it using the same techniques that you used on PC-based infrastructure, from cloud down to the servers in the field.
• 33:37
  Speaker 2: Instead, you have to come up with systems that do that, that are appropriate for the class of hardware that you're using, and that's what's missing.
• 33:46
  Speaker 5: But part of your scale thing too is maybe the concepts could be the same, but it's not the same technology because it's not appropriate.
• 33:53
  Speaker 5: I can do Kubernetes if I wanted to up here, but it doesn't make sense to do it on a little device.
• 33:58
  Speaker 5: I might want to use WebAssembly.
• 34:00
  Speaker 2: Wasm.
• 34:01
  Speaker 5: Absolutely.
• 34:02
  Speaker 5: I think that makes a lot more sense.
• 34:03
  Speaker 5: Does that make sense to everybody?
• 34:04
  Speaker 4: WebAssembly Wasm.
• 34:06
  Speaker 5: Wasm Okay, there you go.
• 34:14
  Speaker 1: You put COBOL on your web assembly, nice.
• 34:15
  Speaker 1: All right, let me ask.
• 34:16
  Speaker 1: You can pay me later.
• 34:16
  Speaker 1: Yeah, there you go, let me.
• 34:17
  Speaker 1: Let me ask a little.
• 34:17
  Speaker 1: A harder question.
• 34:18
  Speaker 1: I think is, like you know, a lot of times technology gets advanced through different Circumstances environmental circumstances, historic circumstances, sociological circumstances.
• 34:29
  Speaker 1: So Edge AI right now is becoming a thing.
• 34:32
  Speaker 1: It's getting deployed in water systems and agriculture, as you were saying, and building maintenance and manufacturing and vehicles and cars and things.
• 34:41
  Speaker 1: What's the thing that you think would really catapult Edge AI sort of to the next thing?
• 34:47
  Speaker 1: What's the black swan or some sort of event that's going to really accelerate Edge AI in terms of development and deployment?
• 34:55
  Speaker 5: Show me the money.
• 34:56
  Speaker 5: Go ahead Serious people.
• 35:00
  Speaker 4: I mean, I honestly think that we're back to the same scenario that we saw in IoT is the money's actually not on the consumer side?
• 35:07
  Speaker 4: You know, consumer devices, devices it's in the industrial sector.
• 35:10
  Speaker 4: So operations, production, oems that are actually building the equipment or that might be sitting in a factory floor.
• 35:19
  Speaker 4: With the automotive space, that's an interesting area because of the V to X and V to V, when we get to a place that we're really starting to see the vehicles communicate to other vehicles and with the traffic systems and with, you know, delivery trucks getting a green light as they go through.
• 35:36
  Speaker 4: There are so many different scenarios that are changing.
• 35:38
  Speaker 4: But that automotive space is really interesting because there's a lot of growth in terms of, you know, the chassis assembly, the dashboard, the infotainment systems, the level of sensors are now on a vehicle and what they're going to continue to leverage that technology for is now it's like they're a little bit more of a standalone, but the communication to the transportation systems, the lighting systems, the city we are not there yet.
• 36:08
  Speaker 4: We're very early in the games.
• 36:09
  Speaker 4: I think there's there's a lot of great examples that are out there with different companies, but there's there's a growth opportunity there for those that get it right.
• 36:18
  Speaker 5: Um, and I think the business models haven't really shaken out yet, so that we're early in the game I think I figured out the conclusive answer to your question are you ready for the jaded, dystopian view of all this?
• 36:30
  Speaker 5: Please All right, you know you're pitching the VC right and you got your technology and your edge AI and your stuff like that and we're really wanting to accelerate.
• 36:38
  Speaker 5: And the VC says, how many employees is your technology going to replace?
• 36:44
  Speaker 5: And you're like, oh, that's awkward, I don't want to have that discussion.
• 36:49
  Speaker 5: That is absolutely how that stuff gets going.
• 36:52
  Speaker 5: It's this horrible thing for me to say, but it is exactly how it's going to happen.
• 36:56
  Speaker 5: Did you see Jensen Huang at CES telling you the IT department was going to become the HR department for all your agentic AI employees?
• 37:05
  Speaker 5: It could happen, right, Absolutely it can happen, and so that's one way that it'll accelerate it and people start believing that stuff.
• 37:16
  Speaker 4: And it's also some of the economists and big thinkers out there.
• 37:19
  Speaker 4: They're saying our population is not growing at a fast enough rate to employ the companies, the spots that we need, so we will have to use robotics.
• 37:30
  Speaker 4: We will have to use robots to fulfill.
• 37:32
  Speaker 4: Yeah.
• 37:32
  Speaker 4: We, if we look at the long-term growth of the population.
• 37:35
  Speaker 4: So, sometimes you have to take a step back and look at the economics and some of the macro factors that we're just not considering.
• 37:43
  Speaker 5: You're so right.
• 37:43
  Speaker 5: I mean, even before this whole, about midway through last year, last summer, I noticed every event.
• 37:50
  Speaker 5: I went to every CEO.
• 37:51
  Speaker 5: Every other word was agentic, AI, agentic, AI, and it's going to be agents and they're going to do all that stuff.
• 37:56
  Speaker 5: You know who was at the party a little bit earlier than all that?
• 37:59
  Speaker 5: That robotic process automation?
• 38:01
  Speaker 5: Folks like UiPath, RPA what do you think that was doing?
• 38:06
  Speaker 4: RPA.
• 38:06
  Speaker 4: Yeah, just basic software, mundane tasks.
• 38:08
  Speaker 5: Remember the 1990s we built tens of millions of Win32 apps and people are clicking.
• 38:13
  Speaker 5: And now RPA does that, and what's it doing?
• 38:15
  Speaker 5: It's replacing an employee.
• 38:17
  Speaker 5: In fact, I know companies that put that RPA agent into Active Directory at their company as if it was an employee.
• 38:24
  Speaker 5: So this is not me just making stuff up.
• 38:27
  Speaker 5: It's not a stretch.
• 38:28
  Speaker 5: I think this agentic stuff yet.
• 38:31
  Speaker 1: So this is a callback to the McDonald's ice cream machine.
• 38:34
  Speaker 1: It is Humanoid robot.
• 38:35
  Speaker 1: It is.
• 38:36
  Speaker 5: It is absolutely, but anyway that'll get excited.
• 38:39
  Speaker 5: So make sure when you're doing your VC pitch, tell them 10,000 employees gone.
• 38:43
  Speaker 5: I hope nobody's recording this, okay.
• 38:50
  Speaker 2: We built a world that requires a lot of maintenance.
• 38:52
  Speaker 2: It really does Like the ice cream machine, yeah, but when I talk to CIOs at some big enterprises, they tell me they can't hire enough people.
• 39:02
  Speaker 2: They can't hire the skill sets that they need.
• 39:04
  Speaker 2: They're willing to pay.
• 39:06
  Speaker 2: The talent isn't there, and I think that's where the agents are really going to come into play.
• 39:13
  Speaker 2: So will it replace jobs?
• 39:14
  Speaker 2: Yeah, but it will also give us the skills that we need and amplify them to the point where we can.
• 39:18
  Speaker 2: We can make even more productivity out of the stuff we already have in the semiconductor industry the skills is a huge issue right now.
• 39:27
  Speaker 7: We're I so about two years ago I started a nonprofit organization called SEMI Semiconductor Industry Mobile Education Unit.
• 39:33
  Speaker 7: The plan is to build a 52-foot semi, an 18-wheeler, that goes to schools around the nation that gives these kids this aha moment of I learned what semiconductors are and give them this Disney World experience.
• 39:43
  Speaker 7: It's just a.
• 39:44
  Speaker 7: We're still fundraising and growing that, but we have to bring this spark to the kids.
• 39:51
  Speaker 7: Right now in semiconductor industry.
• 39:53
  Speaker 7: We're doubling the size of the industry in the next eight years.
• 39:57
  Speaker 7: The average age for senior engineers is 52 years old and there's zero students in the pipeline to fill these roles.
• 40:03
  Speaker 7: It's a big issue.
• 40:05
  Speaker 7: Everybody, from the president of the United States and everybody else is looking down at this going.
• 40:07
  Speaker 7: This is a big problem.
• 40:08
  Speaker 7: It's a big issue.
• 40:08
  Speaker 7: Everybody, from the president, united States, everybody else, is looking, looking down at this going.
• 40:11
  Speaker 7: This is a big problem.
• 40:11
  Speaker 7: But now, like what you're saying about, we're gonna need to reskill everybody, and you don't, you do.
• 40:16
  Speaker 7: We don't just need to train the new guys, we need to reskill the old guys, and so this is a big issue and and and trying to train everybody and get everybody up to park.
• 40:25
  Speaker 7: Yeah, and how this is going to change, change so many of the.
• 40:28
  Speaker 1: Well, it's like we mentioned this morning, like our Edge AI One Million Initiative, try to get a million people educated on Edge AI so they can be in the workforce and build some of the stuff that we need.
• 40:37
  Speaker 4: I was going to say I serve on Texas A&M University's engineering technology department.
• 40:45
  Speaker 4: I serve as an executive advisory board member and that's generally the most common.
• 40:49
  Speaker 4: They have an engineering school, very big and number one, but the biggest conversations we have every year or we meet twice a year is they continue to modify and change and try to incorporate the curriculum to fill those skills gaps.
• 41:08
  Speaker 4: Yeah, incorporate the curriculum to fill those skills gaps and it's not happening at the pace at which they need to, even just with the conversations that we're having now.
• 41:15
  Speaker 4: Where they're getting a lot of that is with their capstone projects and working with big companies that are, let's say, in Texas that can come to the university and help work on real live projects with the students, and so that's helping some.
• 41:29
  Speaker 4: But the biggest gap that they see and that they see that we need, that the businesses are asking for, is we need our technologists, our engineers, our developers, our coders, to come out with a little bit more balance of understanding the business use cases, the applicability to the enterprise, and that's an area that's the soft skills that they're really working on, shifting the curriculum a little bit more for the engineering department and for some of the other, the software engineering areas.
• 42:02
  Speaker 4: That's a big area of focus, I know, for A&M, and so it might be similar.
• 42:06
  Speaker 4: There might be similar conversations happening in other colleges.
• 42:10
  Speaker 1: Yeah, great, hey, I wanted to open it up too for the audience.
• 42:13
  Speaker 1: If there's folks with questions, kind of burning questions for this esteemed panel, now is your chance.
• 42:20
  Speaker 1: Let's get those hands up.
• 42:22
  Speaker 1: People, there you go.
• 42:23
  Speaker 5: There's Ryan back there there you go there's a plant, all right, he's a groupie.
• 42:31
  Speaker 5: You can really ask us anything, not even related to this.
• 42:36
  Speaker 10: Okay, I have a lot of experience in the OT space and I think the biggest gap is finding OT sensor providers, solution providers and getting them to connect the dots with software solutions.
• 42:48
  Speaker 10: I think it's one thing I run into every time I go to pursue it on the hardware side is really big NRE that kind of feels like we're funding and building their own products.
• 42:59
  Speaker 10: So I'd love to hear any kind of solutions to kind of minimize that NRE and still be able to move quick to solve these OT problems.
• 43:06
  Speaker 10: What's an NRE Non-recurring?
• 43:10
  Speaker 2: expense.
• 43:11
  Speaker 10: Thanks.
• 43:14
  Speaker 2: Can I take that?
• 43:14
  Speaker 2: Yeah, that's an excellent question.
• 43:17
  Speaker 2: How do we get those?
• 43:27
  Speaker 5: hardware sensor people to know about the software guys or work about the software guys or work with the software folks.
• 43:34
  Speaker 10: Yeah, there's always an urgency for OT.
• 43:37
  Speaker 10: This is broken today.
• 43:38
  Speaker 10: And they have really high scalable solutions.
• 43:41
  Speaker 10: But now especially, I'm seeing, with more hardware coming out, I'm seeing more NPUs instead of GPUs and there's always that kind of customization that you're going to see to get the software to run on the hardware, to even get the data ingested right.
• 43:55
  Speaker 10: So I don't know how to, kind of, you know, solve that gap to where I'm so jaded.
• 44:01
  Speaker 10: Ryan.
• 44:02
  Speaker 5: You know, no, seriously you could have asked this question 20 years ago or 10 years ago.
• 44:07
  Speaker 5: This is not a new thing of trying to get those folks to get on board with the software.
• 44:11
  Speaker 5: How are we going to do that?
• 44:12
  Speaker 5: And we still haven't done it.
• 44:14
  Speaker 5: Or we do it in little pockets, right?
• 44:16
  Speaker 5: You know some point, solutions that are like that, and you're right, you get frustrated.
• 44:20
  Speaker 5: How is this not solved?
• 44:21
  Speaker 5: Yeah, we were doing this stuff forever ago.
• 44:23
  Speaker 5: How is this not solved?
• 44:25
  Speaker 5: You is not solved.
• 44:30
  Speaker 5: You know, maybe talk to the guy.
• 44:31
  Speaker 5: You know there's that guy in office space who would talk to the software folks.
• 44:33
  Speaker 5: The customers would talk to him.
• 44:34
  Speaker 5: You know the software folks don't know how to talk to customers.
• 44:35
  Speaker 5: The people person.
• 44:36
  Speaker 5: He's a people person.
• 44:37
  Speaker 5: Yes, exactly that's what we need.
• 44:39
  Speaker 5: That's what we need is a people person.
• 44:55
  Speaker 2: Yes, I was going to say it's a little rude to the semiconductor industry so I was kind of hesitant.
• 44:58
  Speaker 2: But when you ask a semiconductor company for a product that you're going to use to build a bigger product, you can ask them to supply the entire thing, the whole software stack, everything.
• 45:05
  Speaker 2: Give me the OS, give me a way to update that OS.
• 45:09
  Speaker 2: Give me a way for us to apply patches to that thing.
• 45:12
  Speaker 2: Give me a complete platform that I can build on.
• 45:15
  Speaker 2: Now, that's not traditionally the role of a semiconductor company.
• 45:18
  Speaker 2: It sort of is.
• 45:26
  Speaker 10: But what's broken here is really the supply chain, yeah, and I guess I'm just.
• 45:33
  Speaker 2: I don't see a lot of software resources at those OT hardware side.
• 45:40
  Speaker 5: Be there yeah, they need a platform to build on next time.
• 45:42
  Speaker 5: You're there right flipping over tables.
• 45:43
  Speaker 5: Yeah, yeah, I try not to.
• 45:43
  Speaker 5: Who else come on?
• 45:44
  Speaker 5: There's got to be hundreds of questions here.
• 45:45
  Speaker 5: This is a AMA going here.
• 45:46
  Speaker 5: Come on.
• 45:46
  Speaker 5: Anyone, anyone, nobody cares.
• 45:52
  Speaker 9: Oh, there's one.
• 45:52
  Speaker 9: Okay, you had an interesting comment you said earlier on the need for a standardized platform for Internet of Things systems.
• 46:09
  Speaker 9: Can you elaborate more on those thoughts?
• 46:12
  Speaker 2: Is that for me?
• 46:12
  Speaker 2: Yeah, I guess so.
• 46:13
  Speaker 2: Yeah, if so, I did not say standardized.
• 46:16
  Speaker 2: We can review the tape, but I guarantee you I did not say standardized.
• 46:19
  Speaker 2: Yikes, somebody said standardized I said uniform or at least not done by the end developer.
• 46:26
  Speaker 2: You're either developing applications or you're developing system code.
• 46:29
  Speaker 2: When you mix the two and when you make a company responsible for building an entire piece of equipment, all the way down to the metal, that's what kills our productivity.
• 46:39
  Speaker 2: That's what kills scale for everything we're talking about.
• 46:42
  Speaker 2: At the edge, that's where your NRE is, is in assembling all that stuff.
• 46:47
  Speaker 2: And oh, you want another product.
• 46:49
  Speaker 2: Oh, we have to do that all over again.
• 46:51
  Speaker 2: You want to change vendors?
• 46:53
  Speaker 2: Oh, we have to change everything.
• 46:54
  Speaker 2: You know again, you don't really need an industry standard that's all-encompassing.
• 46:59
  Speaker 2: I think those things happen over time.
• 47:01
  Speaker 2: But we need to fix the supply chain first and figure out who's going to do that.
• 47:05
  Speaker 1: So the supply chain needs to kind of go up the stack a little more, to be more of a solution provider, so that there's less of a gap to commercialize.
• 47:15
  Speaker 4: I don't know if you guys saw the news today, but I was privy to an NDA meeting last week with Qualcomm.
• 47:20
  Speaker 5: Oh, tell us all about the NDA thing.
• 47:23
  Speaker 4: Well, it's now out today.
• 47:24
  Speaker 4: But Qualcomm has really been working.
• 47:27
  Speaker 4: They've done a lot of reorg around their IoT embedded side of the business, which really is more enterprise, tech and industrial, and they recently just did a I mean, they had a rebrand Dragon Wing is their new brand, not the Snapdragon, but it's kind of a spinoff.
• 47:45
  Speaker 4: But that's part of what they're working on is building out.
• 47:50
  Speaker 4: I mean, I told them it's one thing to build out a whole partner list of hardware vendors, solution vendors Everyone says that they are a platform vendor, but then there's software and software as a service vendors.
• 48:03
  Speaker 4: But they're really going to look at and taking a deeper look at okay.
• 48:06
  Speaker 4: Okay, if we're serving a manufacturing facility or a smart factory and whoever, who are we bringing to the table?
• 48:14
  Speaker 4: As opposed to just saying, hey, I'm just gonna create a logo sheet, right, which we've seen a lot of, which is okay, you're not really doing anything with them, but you say they're a partner, but you don't have anything live deployments with these vendors, so it's now moving into the next level where they're actually going to start pulling.
• 48:27
  Speaker 4: Partner, but you don't have anything live deployments with these vendors, so it's now moving into the next level where they're actually going to start pulling some pieces together.
• 48:34
  Speaker 2: Yeah, and they've been working on this for a while.
• 48:35
  Speaker 2: Last year they bought a little company called foundriesio and they supply a Yocto-based web-updated or internet-updated Linux distro that you can use pretty much as is.
• 48:49
  Speaker 2: So I think they're on this page where you buy the silicon from them and it comes with all the software you need to use it, including updating it in perpetuity.
• 49:02
  Speaker 1: Cool, any last questions.
• 49:05
  Speaker 1: Samir, we got go for it.
• 49:08
  Speaker 3: Last question, samir, we got, go for it, maybe more for comments, and that would kind of ask a reaction.
• 49:15
  Speaker 3: I strongly do not believe that the Edge can be unified or that anyone can do.
• 49:22
  Speaker 3: Unified platform and even your example about Foundry you have 10 things like Foundry, you have canonical, you have the embedded have canonical, you have the embedded kit you have.
• 49:30
  Speaker 3: Per definition, the edge is a mess.
• 49:33
  Speaker 3: The edge is a wide, wide west.
• 49:35
  Speaker 3: And why so?
• 49:36
  Speaker 3: Because you have a ton of different devices.
• 49:39
  Speaker 3: What a truck has in common with a coffee machine and a medical device and a sensor, and you have way more variation in between those devices than you have in between two servers or two gateways or whatever.
• 49:53
  Speaker 3: So I think, per definition, it's like an engineer dream to have unified platform, even for a JI.
• 50:00
  Speaker 3: They won't be.
• 50:01
  Speaker 2: I completely agree with you and that's why I think I said just a minute ago no, there's not going to be a single standard that enables this.
• 50:09
  Speaker 2: But if we have a change in the supply chain where people that make the hardware that's required to support that diversity give you a complete set of tools, everything you need, then that will help a lot, Because then people developing applications can just develop applications.
• 50:27
  Speaker 5: That's where you get scale and the mess he describes.
• 50:28
  Speaker 5: What does that tell you?
• 50:30
  Speaker 5: Be a consultant when there's mystery, there's margin.
• 50:37
  Speaker 2: Well, that's just me.
• 50:39
  Speaker 6: I am from the OT world as well industrial automation and NVIDIA is making a big push into partnerships with industrial companies.
• 50:47
  Speaker 6: They're all over some of the industrial shows like Automate, but I'm not really seeing any of these other players doing that same thing.
• 50:54
  Speaker 6: Any thoughts on their approach to the ecosystem and kind of getting the hardware and the industrial partners to develop on their platform versus trying to develop, I guess, full solutions?
• 51:06
  Speaker 6: And I don't know I don't have enough background on this organization as to why nobody from NVIDIA is here, for instance, or that sort of thing.
• 51:14
  Speaker 6: Is that a different pool altogether?
• 51:17
  Speaker 1: No, I think it's like I mean folks were saying here.
• 51:20
  Speaker 1: The trend is, and maybe one of the answers to make edge AI scale better, is to have technology partners go farther up the stack right To give people, instead of 50% of the solution, give them 80% or 85% of the solution and not let them sort of kind of go in all these weird directions.
• 51:38
  Speaker 1: I think NVIDIA is like with CUDA infrastructure.
• 51:41
  Speaker 1: I mean, they built up a multi-million developer community around CUDA and that's a huge moat for them as a defense against other, you know, platforms.
• 51:50
  Speaker 1: But I think they did it because they invested in the stack, went way farther up the stack than a lot of competitors did.
• 51:57
  Speaker 4: But what I would say about that is what they did was waken up the competition, and now competition is like oh shit, I'm behind.
• 52:06
  Speaker 4: What do I need to do?
• 52:08
  Speaker 4: Do I need to buy Intel?
• 52:09
  Speaker 4: Do I need to partner with Broadcom?
• 52:11
  Speaker 4: So what it did was just kind of disrupt, you know just the overall competitive landscape.
• 52:19
  Speaker 4: So as a result of that, I do see that we're going to see some major changes over the course of the year and you're going to see some of their top competitors looking at what do we do and how do we roll out our services, and maybe you'll see some big announcements at Mobile World Congress.
• 52:37
  Speaker 4: So I do expect to hear some big things, but it'll start trickling in from that.
• 52:42
  Speaker 5: Here's my jaded but probably true answer to your question.
• 52:45
  Speaker 5: Nvidia has been a rocket ship.
• 52:47
  Speaker 5: Probably true answer to your question NVIDIA's been a rocket ship and that rocket ship is based on their GPUs for training AI models and doing inference.
• 52:55
  Speaker 5: And then you kind of saw it at the CES thing.
• 52:58
  Speaker 5: It's like all of a sudden, how do we keep this rocket ship going?
• 53:02
  Speaker 5: How do we keep that stock price?
• 53:03
  Speaker 5: Just killing it, Because it's kind of chilling a little.
• 53:06
  Speaker 5: We need to broaden where we live.
• 53:08
  Speaker 5: Let's invade manufacturing and let's get all those people hooked on CUDA, you know, and have manufacturing and robots.
• 53:17
  Speaker 5: All be NVIDIA.
• 53:18
  Speaker 5: They have to broaden their footprint to sustain the lift that they've enjoyed for the last several years.
• 53:25
  Speaker 5: That's what I think.
• 53:27
  Speaker 2: Exactly, and it's physical AI.
• 53:28
  Speaker 2: It's their big bet.
• 53:29
  Speaker 2: Yeah, that's right, and I think they placed that's what I think Exactly, and it's physical AI.
• 53:30
  Speaker 2: It's their big bet.
• 53:30
  Speaker 2: Yeah, that's right, and I think they placed it wisely and I think they're doing a good job of that.
• 53:34
  Speaker 5: Yeah yeah, you're right.
• 53:37
  Speaker 1: Okay, I think we're kind of at time, I think the next oh, we got another one.
• 53:41
  Speaker 1: Oh wait, boop Go for it.
• 53:43
  Speaker 1: One more question.
• 53:51
  Speaker 8: Yep, I had a question for the group about the lower end of the middle market or small businesses.
• 53:58
  Speaker 8: A lot of what you're talking about applies to big businesses.
• 54:01
  Speaker 8: What about mom and pop businesses out there, small regional chains?
• 54:06
  Speaker 8: Is that a market opportunity?
• 54:09
  Speaker 8: Are they all going to crash because they can't compete?
• 54:12
  Speaker 4: I would say the majority of the excitement around that is software as a service, just what we've been seeing for quite some time.
• 54:19
  Speaker 4: Where AI becomes a feature, they're not paying for it.
• 54:25
  Speaker 4: Small and mid-sized businesses are not going to pay for it.
• 54:28
  Speaker 5: But Satya said SaaS is dead.
• 54:28
  Speaker 5: What you're going to see?
• 54:28
  Speaker 5: More not paying for it.
• 54:29
  Speaker 4: Small and mid-sized businesses are not going to pay for it.
• 54:30
  Speaker 4: But Satya said SaaS is dead.
• 54:30
  Speaker 4: What you're going to see more people paying for on the small and mid-sized business side are professional-type companies, marketing companies, those that are serving small and mid-sized businesses, because the tools that they're going to be using they will pay to ramp up their business or to provide new services.
• 54:51
  Speaker 4: Because what AI is going to do for marketers, for example, because I'm in the marketing industry it will create new revenue streams for them because they're able to essentially ramp up their workforce without having people.
• 55:03
  Speaker 4: So the volume that they have today of I'm going to develop X number of websites, I'm going to have X number of customers to support their social media PR campaigns that kind of thing that ramps up with the tools that are out there that are available.
• 55:20
  Speaker 4: But I still caution folks because a lot of us test some of the Gen AI stuff and it's not accurate and it's hit or miss.
• 55:30
  Speaker 4: So you just have to be very, very careful if you're using it.
• 55:34
  Speaker 2: Yeah, that's a great question.
• 55:35
  Speaker 2: In the world of IOT today, the big players that are shipping millions of products have a big advantage because of all the NRE associated with with the development, with having to integrate all this stuff that's required to run the application code, which is what they get paid for.
• 55:53
  Speaker 2: What we're talking about here, about really treating IoT devices as more platforms, really benefits the little guy more than the big guy, because they have a more solid platform that they can build on.
• 56:07
  Speaker 2: They have to write less code to get their product up and running and other people then are all set up to support it, to do the OTA updates and everything else it takes to make a whole product.
• 56:19
  Speaker 2: So I think if we really do make this turn as an industry and start to think of small devices, edge devices, ot devices, personal devices as platforms, the little guy is going to benefit probably more than the big guy.
• 56:34
  Speaker 4: The only other industry that I saw that's been kind of interesting is the insurance industry, because you do have a large base of SMBs in the insurance industry or it's their franchise, and so buy here, pay here.
• 56:47
  Speaker 4: For example that's a small one that's been around forever we put a sensor on on the automotive, on the vehicle, and it's if we have someone who's low credit but and if they're not paying their bill, we can turn that vehicle off and allow them.
• 57:00
  Speaker 4: They won't be able to start the car, right, but there are.
• 57:03
  Speaker 4: There are other examples in the insurance space.
• 57:05
  Speaker 4: So I think there might be some little pockets here and there where you know it makes sense.
• 57:10
  Speaker 4: And real estate may be another.
• 57:13
  Speaker 4: So some of the building sector, because you have a lot of SMBs in the building sector, that might be another area.
• 57:19
  Speaker 5: I think you're right about the SaaS Different point AI solutions as a service that the little person with not a lot of money can take advantage of.
• 57:28
  Speaker 5: And there'll be millions of these little things and they will seem like the apps on your iPhone of today, but they will be little AI SaaS things.
• 57:36
  Speaker 5: You own a print shop and there's stuff for the video camera and there's stuff for your point-of-sale machine and there's stuff for all that and there'll be little things and they'll be as a service and they'll be inexpensive for SMBs to be able to stomach that kind of capability without needing tons of money.
• 57:55
  Speaker 7: I question, though what is the adoption rate?
• 57:57
  Speaker 7: I talked to a guy the other day.
• 57:59
  Speaker 7: He asked me what ChatGPT was.
• 58:03
  Speaker 7: I was like really you don't know what Chacho PT is, but what is truly like.
• 58:11
  Speaker 7: I mean small business.
• 58:12
  Speaker 7: I deal a lot with small business and what is the adoption rate going to be like for those businesses?
• 58:18
  Speaker 7: Right, and you know the expense and the cost and can they stomach that cost?
• 58:24
  Speaker 1: Yeah, yeah, definitely no, I mean, it's going to be, you know, as the cost will always come down, there will always be, you know, as the cost will always come down, there'll always be, you know, more affordable solutions, but they have to be solutions that really add value, especially in the SMB market.
• 58:49
  Speaker 1: More, you know we're seeing it like, even with the security cameras, right, I mean, and ring doorbells and things like that, those have become much more commonplace and we'll start to see, you know, peel and stick cameras that will measure, you know, dwell time and waiting time and all these other things, and small businesses that will help them out and they'll be as a service and you'll buy them on a monthly basis and cameras will come free and stuff Camera as a service, stuff like that We'll get there.
• 59:11
  Speaker 1: We're kind of running out of time.
• 59:13
  Speaker 1: So any other critical questions?
• 59:16
  Speaker 1: Or does everyone want to go to Rainy Street and get a beer?
• 59:18
  Speaker 1: I think that's kind of the next step.
• 59:20
  Speaker 1: Okay, cool.
• 59:21
  Speaker 1: Well, I want to thank the panel.
• 59:22
  Speaker 1: Thank you very much, Robert.
• 59:24
  Speaker 1: Rob, Stephanie and Bill Appreciate that Round of applause.
• 59:28
  Speaker 1: Thank you.