Back to Blog

Tec Trek Podcast: What are Chips?

Tune into our latest episode of Tec Trek! This week, TJ and Yatri are joined by Ambroise Popper of Nestwave to discuss microelectronic chips: what they are, how they're used, and how small they will allow the devices of the future to become.

Listen on Spotify and Apple Podcasts.


Episode Transcription

TJ: Welcome, ladies and gentlemen, back to "Tec Trek." My name is TJ and I'm here, as always with...?

Yatri: Yatri. How you guys doing?

TJ: And today, we're gonna be talking to another expert in another industry that is gonna be able to illuminate some really interesting facts to us. And today's question, what we wanna talk about today, is what are chips and what are they good for? Now, we're not talking about potato chips, clearly. We're not talking about french fries, if you're in the UK or in Europe, that might be a chip. We're talking about integrated circuits. We're talking about transistors, we're talking about MOSFET, and MOS circuits and all that good kind of stuff.

Chips are the basic component of pretty much any kind of electronic device. It's the building block that has led to modern electronics, especially computers. If you're thinking about it, you've got your motherboard, which is full of transistors in itself is a large chip, you've got the northbridge and southbridge on some motherboards. So you've got again, chips, the CPU is a MOSFET. It's an integrated circuit. You've got the RAM, you've got the GPU. Pretty much everything in a PC is using chips. Chips are integrated circuits, they're etched onto silicon wafers. Other materials are used MOS and MOSFET stands for metal oxide silicon, I think. It stands for metal oxide silicon. So there are other materials used, but the basic processes, your acid etching transistors on to pathways, bridges, if you will, on to silicon wafers.

Yatri: Used to be done by hand, not anymore. Although you still can, if you are feeling pretty bold.

TJ: I wonder if we're ever gonna get to a point, you know how we've gotten to a place with so much consumerism, it's like you get to the bespoke area too, right? Like we got things that are mass produced, but then there's always that one company like you've got automobile manufacturers, but you've got Rolls Royce who does it by hand, I hope we have a computer part manufacturer who comes out and like their big selling point is that they're hand-etching their chips, although that brings up an interesting point. So computer chips have embedded transistors, which are basically conductive opponents that control the flow of electricity, right? You've got a gate, on, off, on, off, and that's how the signal is sent. Well, modern chips, like so my PC, which we can't see here, obviously, this is audio, but I have an AMD Ryzen 5 3600 processor, it is 3.8 million transistors on the chip. And it's not very big, it is not a large piece of equipment. And we're looking at 3.8 million transistors, which is...

Yatri: And you contrast that with like the electronic calculators of the '80s and '90s, right?

TJ: Yeah, exactly, where a dense chip had 3000 transistors was, you know, that was bonkers. That's interesting, though, because that brings up a lot of interesting stuff about sort of the development of chips, right? Like, it's all about density and size, like going smaller and more dense, more transistors onto smaller footprints.

Yatri: We tend to forget that...we tend to have this idea that computers or electronic components are a bit of a black box. And, well, if you ever take these electronics apart, they do look like little black boxes, but they're physical pathways, and they're controlled by electricity. And they very much do something physical in the world. And so it's an interesting way to kind of remember.

TJ: Exactly. And I think that's the thing that's kind of like the sort of underlying unknown for a lot of people who don't deal with electronics in a sort of micro level. If you don't even think about it, it is's a physical gateway, sending electricity down a path or shutting electricity off down a path. And that's actually happening. And transistor count is one important thing, but it also scales, interestingly. There's a thing called Moore's law. It's Gordon Moore, he was co-founder, I think, of Fairchild Semiconductor, and then consequently was a founder of Intel in the, I believe, '75 or '65, something like that. And his thought, what Moore's law is, is that roughly every year, transistor count will double on...mainly we're talking about CPUs here. Now, that held true until, I don't know, maybe 2010. And then it started slowing down, which is why you've got things happening right now. And this is always something that interests me is the sort of battle between our big computer chip manufacturers, Intel and AMD, they're your...which...well now we've got companies like ARM coming in.

Yatri: Yes, like Qualcomm.

TJ: Yeah, Qualcomm. Exactly, mobile chips versus desktop chips, but Intel has been on their same process. Now, when I say process, that means the sort of method of fabricating the semiconductor, the chip. They've been on the same 14-nanometer process for many, many generations now. Whereas AMD is on a seven-nanometer process, which means, immediately, more density, which...and it's not exactly one for one scale, but more density usually means, A, more capability and, B, for the end user, better performance. I know that's a pretty blanket statement.

Yatri: You also see an increase in electrical efficiency as well. So you see changes in voltages applying and which means that the same amount of work not only gets done faster, but takes less power to do, arguably.

TJ: Exactly. Well, so my computer, I rebuilt it last year. The previous processor was way less powerful. It was also AMD, but it was...we won't even mention the name. It was in the dark days of AMD's desktop processing, it was an FX processor. But it was way less powerful and it was way more power hungry. Like it was so inefficient, the heat that it put off, which that's another thing that's important to realize, I think, about integrated circuits and chips, if you will, is like it's a physical interaction of electricity, heat is put off, lots of heat. This chip, at idle, runs 35 Celsius, that's when my computer's doing nothing. I've spiked it, not on purpose, I've spiked it up over 90 C, doing a stress test, which is pretty warm...

Yatri: You definitely don't want to run that for very long.

TJ: Yeah, soon as I saw, shut down, fire strike. I was like, "Oh, God get out of there. Oh, no, what have I done?" So again, like, we've kind of talked about this a little bit. But chips are using electricity to send signals. And that's how they communicate, which is then interpreted by a receiving device, which is how electronics, in a nutshell, work.

Yatri: Back in the early days, used to use vacuum tubes for this. In fact, actually, if you look online on YouTube, and search for transistor gateways out of wood, there's a lot of people who will carve these out of wood to kind of demonstrate how these gates work. Singularly, like with marbles, singularly and in conjunction, forming different types of more complex units at a various different types of gates. And that's not to mention what you can do with redstone building in Minecraft, right?

TJ: Yeah, exactly. Practical real world application is how we're gonna use it at a Minecraft server.

Yatri: True factors are a thing.

TJ: Yeah, exactly. Well, it's also interesting, too, because I think everybody knows that old adage about, well, back in the '40s, computers took up an entire room. Well, yeah, that's because, again, we were using a different type of physical...

Yatri: Your basic building components are a lot bigger.

TJ: Large. Yeah, exactly, exactly. It's like a hot wheel versus a micro machine, what is just smaller, it's, like, that's just how it is. And so if you use a lot of something to build, then it's gonna be larger. Well, and that goes for sort of all technology. There's larger computers. One of my favorite things, like, as far as like tech history is looking at old memory, right? Like now, random access memory, RAM, on your computer, it's a chip. It's a solid state memory that's being accessed. The Apollo computer, for instance, use magnetic rope memory. And so you have washers with this magnetic rope. And if it went inside the rope, it was a one or if it went inside the washer, it was a one, if it went outside the washer, it was a two. And that's how it communicated and that's how it was accessed and those things were like braided by hand. Again, bespoke computing, that's the next big thing. Made of wood, fine materials.

Yatri: It's amazing to look at, like, how far we've come in so many different ways too. So like, again, the decrease in size enables us to do so many more things. I mean...

TJ: And it's such a short amount of time.

Yatri: Yeah.

TJ: Like it's not been that long.

Yatri: It's not surprising to think of basic things, for example, lightning and USB-C cables that basically just look like the cable have some fairly advanced circuitry in line. And things that you wouldn't expect or think to really need or require any kind of integrated circuitry just has it because you can do so much with it, like building and failsafes or adding digital readouts, things like that.

TJ: Yeah, I mean, you got listening devices, refrigerators, GPS trackers, I mean, any electronic device is gotta have a chip, it's got to have a brain at the heart of it. All right. Well, that's enough of us rambling, because we are not the experts. We simply are interested. So what we're gonna do now, we're gonna bring on our expert, and we'll be right back guys.

All right guys, and we are back. And now I would love to welcome our expert guest. His name is Ambroise Popper. He's the CEO of Nestwave, a provider of geolocation, IP and cloud services for IoT modern makers. He's worked in France and Silicon Valley in the wireless IoT and AI fields, including strategy and business development, at Quantenna, a leading chipset supplier, and prior to that, he co-founded Sequans, a 4G chipset vendor.

Welcome, Ambroise. It's such a pleasure to have you.

Ambroise: Thank you very much. It's my pleasure to be here today with you guys.

TJ: Fantastic. Tell us a little bit about your career and your personal history.

Ambroise: Yeah, thanks. So I'm originally from France, although I was lucky to live as a kid in the UK. So I was fortunate to learn English at a young age. So I always wanted to have more of an international career. So I was able to actually go twice to Silicon Valley during my career. So the first time was in 2005. So actually, I had been in France, working as R&D engineer, initially for a U.S. startup called Pacific Broadband. In those days in the 2000s, it was working on new generation DOCSIS, so cable, providing superfast broadband, that was really important. And we were trying to get enough throughput to your home so people could do video on demand. Remember, in the year 2000, people will say, "Yeah, video on demand is the next big thing." And then everybody call it a failure, because it probably only took off maybe by 2010, 2011, when Netflix really became popular. It probably took 10 more years than people were hoping.

So I started off really technical working on the chips to do DOCSIS and bring it to your home. And then with part of that team of Pacific Broadband based in Paris, we founded this company called Sequans Communications. At the time, it was doing something called WiMAX. Maybe you heard of WiMAX, it was a high speed wireless technology, which had some success for a while. And then it was kind of replaced by 4G LTE. So in the years 2009, 2010 2011, in the U.S., Sprint was deployeing WiMAX. It was a extremely successful with Android smartphones, in particular from HTC.

TJ: I was a Sprint user, I remember it.

Ambroise: You remember then?

TJ: Oh, yeah.

Ambroise: So that HTC phone, that was using the chip from Sequans. And that was those days, iPhone just came out. And the first version, maybe you remember, was only supporting 2G EDGE. So it's actually quite slow in terms of throughput. And Sprint took that opportunity with WiMAX to bring in kind of a first 4G smartphones based on Android. And then unfortunately, for us at the time, at Sequans, Sprint decided to put the efforts on 4G LTE, so they could go and work with the iPhone. And unfortunately, Sequans had to kind of stop most of its sequence in WiMAX. But the good news is that Sequans, we then started to develop 4G chips. And we started looking at 4G, not just for smartphones, but for low-power IoT. And from that point on, which was like 2011, I spend most of my career working on IoT and low-power chips.

TJ: That's awesome. That's very cool. Yeah, what I love hearing about your career trajectory is you've stayed right in front of sort of the emerging technology. Like that's so fascinating to me, like working on video on demand that far beforehand. And then transition, like that' I love the idea of just always being on the bleeding edge. I'm sure it's challenging. But...

Ambroise: It is challenging. And sometimes when you're ahead of the curve, you're not the one that gets the most reward out of these technologies.

Yatri: I'd say you're the pioneers, right? I mean, you're the...

Ambroise: We're the pioneers. And sometimes, it's someone bigger that arrives maybe when the market is more ripe can benefit more. But anyway, I find it fun to be ahead of that curve. And at Sequans, we're finding a lot of that low-power IoT stuff. And now, maybe you've heard of technologies like narrowband IoT or LTM, which are deployed in the U.S. to bring in really cellular technology to IoT devices. And I think that's really an extremely interesting field. And that has a lot of future, although it is taking a little bit more time to take off than we were hoping for, that market is really significant today. And it's ramping year after year. And it's gonna go into so many places like automotive, logistics, manufacturing, industry 4.0, and a lot of stuff is happening via IoT space.

So back to what I've done is now, I moved gradually from being an engineer working on the communication side, algorithms to working more in standards, meanings, and then I moved more to product management, marketing, business development, management. And at Quantenna, when I was the second time in the U.S., I was working on Wi-Fi, super high speed Wi-Fi to go above one gigahertz and working on what people now call Wi-Fi 6. So again, that was kind of bleeding edge technology and I was looking at acquiring new companies for Quantenna or investing in startups.

So that was super interesting because I could really look at what is the new future technology that is out there, what is really interesting. And we looked at all sorts of things, not just in chips, like super interesting software technologies. And after that company was acquired by ON Semi, one of the companies that you mentioned, a big chip guy, I decided to move back to France. I was a cold consultant for a while, one of my clients was Nestwave. And they offered me to join full time as a CEO last summer. So that's how I ended up CEO of Nestwave.

TJ: That's a pretty awesome job offer to come in.

Ambroise: So actually, I was one of the co-founders for Nestwave in 2014, 2015, but I was only very remotely involved because I had other full time jobs. But I always stayed very close to my company. It was dear to my heart. I'm good friends with the founder, Rabih. And there was an opportunity for me to join as a CEO. So I said, "Bingo," I was really happy to take that job. And what is interesting with Nestwave was we're not really developing the chips, but we are selling the technology that goes onto the chip. So most of us come from a chip background. And we're extremely close to the chips, mostly IoT chips, but it could be chips also going into broadband, or agriculture or other kind of applications.

TJ: And I think it's brilliant seeing what you guys are doing at Nestwave because it solves such a direct need. The idea that...GPS, as we become more and more reliant on it, like as we become more and more reliant on geolocation, we expose its weaknesses. And so being able to sort of go after those weaknesses and try to patch it, like I think the marketing on the website mentions something about like better reception in urban canyon and indoor coverage and that sort of thing, that is something that I know plagues us, like, in our side, quite a lot.

Ambroise: Yeah, definitely. Definitely. Yeah, I mean, actually, location can be quite challenging. Although I mean, GPS is a great system. It's very robust. And if you imagine it was deployed in the early '80s, the core technology, I mean, coming back to chips, the chips that people had in the initial GPS receivers were extremely basic.

TJ: I'm assuming there was something like 8086s almost like Intel 8086 equivalent.

Ambroise: Yeah, probably, to be honest, I wouldn't know for sure. But I mean, the infrastructure that was deployed by people who invented that system are extremely smart. And you can make a GPS receiver with very relatively basic hardware compared to what we can do nowadays, and completely independent of any cellular technology. You can just turn on your GPS receiver on top of the mountain, in the middle of the ocean, and it works. And GPS itself has not changed that much, but GPS receivers have improved a lot, especially by combining with cellular technology, what people call the assisted GPS. So, for instance, in your smartphone, it uses the same core GPS technology, but it uses the data from 4G to help you dramatically. So you can get a position much faster, which can just download the position of a satellites from your smartphone superfast. But despite that, you're right to point out that it doesn't work that well, in many cases, especially, I don't know where you live in the U.S., but if you're in a big city with...

TJ: Our headquarters are in midtown Manhattan, so we have that problem at our main office also.

Yatri: Tall buildings, congested infrastructure in general. And when it's raining, you have so much feedback, it becomes a difficult thing to solve.

Ambroise: Yeah. And the GPS signal, when you think about it, it's crazy that your smartphones, they communicate through base station cell towers, which are not too far away from you down on Earth, and Wi-Fi, you have access point in your home. You're communicating over maximum several miles. The GPS, you're downloading, I mean, not downloading, but you're using signals from satellites, which are thousands of miles away. And it gives you something super precise. And you could think that's kind of crazy. Why can't I use signals that are closer to me for location? And that's one of the ideas behind Nestwave. These satellites signals are super rebust, super reliable, you have everything there.

But actually, you can use both terrestrial signals, cellular signals, Wi-Fi signals, or Bluetooth beacons that people deploy in the airport, whatever, those signals which are closer to you can actually complement what you have from the satellites to give you the best position possible. And you probably know that if you're using a smartphone, the smartphone already does that. They combine the GPS signal or the signal with the Wi-Fi, because there's databases of Wi-Fi access points around the world. If there's Bluetooth beacons at point, they combine all these things, because the GPS only in indoors, it's not gonna work at all. And in Manhattan downtown, it's just gonna bounce off the building.

TJ: Well, it also allows you to get that... And this is one thing that's interesting, because it allows you to get... so everybody does this, I'm assuming or maybe I'm just, you know, bad with directions.

Ambroise: All the smartphones do that.

TJ: And I'm saying so if you're walking around, if I'm walking around Manhattan, I need to get somewhere, I'm gonna pull up my directions. And I'm gonna watch where I'm walking on my map, and by my cell phone, using Wi-Fi, Bluetooth beacons, it's using satellites, it's able to give me the illusion that I am seeing every step I'm taking. Because it's constantly finding a connection to bounce information and get the information.

Yatri: It's so different from even just a few years ago, where you'd be walking, and all of a sudden, it'll jump you a block over or...

TJ: It jumps you a block. Yeah, exactly. Because it's...

Yatri: You're walking east, but the map is showing you walking west because it's progressively getting better reception or worse reception, depending on what's going on.

TJ: Yeah, depending on...yeah. Exactly.

Ambroise: Yeah, it's improved a lot, but it's still not perfect. And actually, just to pursue a little bit on the smartphones, and then I'll talk about what we do, which is not the smartphones. Google made an announcement a couple of weeks ago, about using the technology where they do 3D mapping of buildings in downtown Manhattan, so they can see how the GPS signals are bouncing around to improve the accuracy on Android and smartphones. So I'm not sure if that's deployed already, when it will be. But we posted the news on our LinkedIn, referring that news because we found it super interesting, although it's not at all developed by Nestwave.

So the point is, to understand what we're doing, so you understand smartphones, by definition, they're smart, because it's part of a name, and they combine all resources. Now when you go to IoT, so I guess we're familiar with IoT, but IoT, here, I'm talking about typical example, tracking your bike. Bikes get stolen all the time. So more and more people, they put a little tracking device on their bike, so they can know, once a day or once an hour, where is my bike just to make sure it is where it's supposed to be.

Now, that tracking device is gonna be quite small. It's gonna be battery based, not a big battery, like a smartphone, they're gonna recharge overnight, you're not gonna recharge the battery on your bike tracker every night. You might recharge it every year, or you might want to never charge it if you can. So you have to consume very low power on these devices. It has to be quite cheap because you're not gonna put $100 piece of equipment on your bike tracking, unless it's a super expensive bike. And it has to work in all of these environments because if you're leaving your bike in a closed-door garage at night, it's not gonna get to the sattelites.

So most of the IoT tracking devices today, I mean, up to today, they've been using GPS primarily. And it still consumes a lot of power because those GPS chips were designed for smartphones, it's not gonna work indoors. So the idea behind what we do in Nestwave is to say, you need a new technology for all this trackers. And it's not just your bike, it could be your scooter, if you're still talking about city transportation, but think about Amazon shipping you a parcel to your home. Today, they don't tell you precisely where it is, they may tell you from time to time, the parcel has left the warehouse, it's in the truck for delivery, or it's been delivered to your house. But if you could have a tracking device that tells you exactly where it is, you will know when to be home to reception it or whatever. But today, it's too expensive. And with Nestwave, we hope we can make these small tracking devices to cost only a few dollars. And they can be used on many more applications.

TJ: Yeah, it just broadens the applications. And I think something you said earlier that's really interesting, that's something that sort of applies to us a little bit as well, is the ability to automotive, and warehouse, and logistics, where you're tracking single items through a process.

Ambroise: Exactly. That's extremely important.

TJ: Yeah, and that's gonna be such an opportunity and it also is something that I think has very easily illuminated tangible benefits to industries that right now don't have that kind of visibility. Right now, they're relying on somebody walking around, checking a part as it moves down the assembly line, checking a package as it moves through their sorting process. And to be able to automate that, I think, is gonna be absolutely massive going forward.

Yatri: I think this is actually really...this is a really interesting point, just because we talked a lot about trying to be ahead of the wave, and to see what's coming and what is sort of important. And there's a thought process that's involved there. So obviously, these are...we touched on several things just now that are going to be large, are already showing some promising developments. Could you tell us a little bit about how current innovations with integrated circuitry right now are leading to those changes? What kind of like physical examples know, if you wanna talk about some of the specialized software that runs on these things? We could talk about how that works.

Ambroise: Yeah, yeah, that's very important. So if I focus on the field that I know the most, which is really IoT, what you have with IoT is you have all sorts of devices capable of generating a ton of data that comes mostly from sensors. So, you know, measuring temperature, vibrations, cameras, and analyzing movements, I mean, all sorts of sensors. So I think one of the major challenges and where there's a lot of innovation is how you can do smart processing of those data on the edge. And I think that's really one of the key areas.

So at Nestwave we're doing just part of that, which is collecting location data on the edge. But in my previous experience, I was working with companies collecting, for instance, looking at vibrations on machines in factories to detect when they're gonna be failing. So analyzing temperature, vibrations, listening even to acoustics for anything that sounds suspicious to do productive maintenance and tell the technician, "You should check that AC unit out there, because it's been behaving slightly different."

TJ: That's awesome. So how do you is the idea of that gonna be made? Is the plan to like have like a baseline database of like, this is what it should sound like, and you've got the frequency, and we're just matching. And if it mismatches, it's oh, that is awful.

Ambroise: I mean, it's a little more sophisticated in the sense that...

T.J: Yeah, of course.

Ambroise: ... people use machine... I mean, it's typically applications for AI and machine learning that can detect regular patterns. And they are able to combine different elements. It's not just the sound is off its frequency, it's maybe a combination of a sound doing this, the temperature doing that, and the vibration doing this, they have detected that based on the examples they analyze, this is gonna lead to a failure within 24 hours, in general. So that's an example of area of innovation, which I find fascinating. And there's lots of chips able to do these things. And people are ARM, for instance, is doing that, or companies like STMicro in France, they're adding engines on this low-power chips, some kind of machine learning engines, giving you the capability of doing that analysis.

TJ: So combining that with what we were just talking about, the idea of logistics, right, not only now, can you track your process, can you track your products or your process, you can also now track the machines that make your process be physically happening, and make sure that their maintenance is up to date with, in theory, low power, easy to access, easy-to-use solutions. Like that's...

Ambroise: Plus you need the location, of course.

TJ: Yeah. Oh, yeah, exactly. Yeah, yeah, we got to be able to find it, once it's broken.

Ambroise: You've got to be able to find it. But I found that that is an area of high innovation, which is people call it Edge AI or Edge Machine Learning. And I was lucky to work a little bit in that space. But I believe that's really a space where it's gonna be, in the next 5, 10 years, bigger and bigger. And even at Nestwave, even though today, our location does not really use machine learning per se, we plan to make use of machine learning to improve our algorithms and to try and collect more sources of data to more accurately predict and let you know where you are. That's really where the future is headed for all of these applications.

TJ: That's amazing. That's fantastic. All right. Now, what are some challenges of integrated circuit engineering?

Ambroise: I think one of the major challenges is due to the nature of chips, which is hardware. And hardware, it's difficult to do. There's a high barrier entry. You can't improvise yourself a hardware company. And I'm not saying software is easy, but the barrier to entry is less for software. And one of the great features of software is if there is a problem, within a few hours, you can fix it and you can deploy to the field and fix it. And even one thing I find fascinating is... I was fortunate when I lived in the U.S. to drive a Tesla, Tesla keeps on updating their software all the times, nothing is more hardware than a car, if you think about it, they are still able to upgrade the functionality to the car overnight so quickly.

So coming back to your question. The problem is chip design takes a long time. The design cycle of a chip is long. So the major challenge for me is when you start thinking about, "I'm gonna do a chip to do that," you need, let's say, maybe two years. If you're super fast, maybe 18 months. But you've got anticipate what the market need is gonna be by that time. And nowadays, I mean, you guys know about more than I do. The technology moves so fast. Consumers, their needs move so fast is the big challenge for a chip company is you've got to place a bet today what the market will need tomorrow. And you can't compress it. Even though there's a lot of innovation in designing chips, new tools and make it faster, new ways of doing things. Ultimately, as you said in the introduction, it's a wafer and there's less and less companies around the world that know how to build chip, the biggest one and most well known is TSMC out of Taiwan. And it's billions of dollars to build the factory to do that. And the wafer fabrication itself is a minimum of several months.

TJ: Oh, yeah. I mean, they have to grow the silicon. They have to literally...

Yatri: They have to grow the crystals, the wafers, and then...

Ambroise: I'm talking about... So that, for me, is really a challenge. So people are trying to solve that problem by making chips themselves a little bit more flexible. Let me explain what I mean. In the old days, you would have, like, completely a chip like from Intel or AMD, that's a processor. It's configurable by nature, meaning people program software on top of it. But then if you would do a cell phone receiver chip from Texas Instruments or Qualcomm, in the old days, it would be very non-flexible, meaning you would implement the algorithms to do a GSM receiver or a Wi-Fi modem, but nothing else. And nowadays, people make those communication chips a little more flexible with more and more processesors in sight, so that they can adapt over time. So that if a new standard comes out, they can try and evolve it. So you have to find the right balance. Because if you make it too configurable, then usually it consumes more power, as opposed to the chip that does just what it's supposed to do. It does it perfectly. So that's really... there's a fine line.

So I'm sure you guys, if you had the podcast with phone conferences, they will tell you how they've made their chips much more configurable, so they can go out and, you know, 5G comes and they can make the 5G chip much faster.

TJ: Yeah, it's almost trying to integrate some modularity into it. So it's like, you know...well, and it's interesting, but now, they've gone away from it, I think, in the current version, but AMD, that's why they did the chiplet design. So it was a seven-nanometre chip, but it would be multiple of them on a 14-nanometer die, so that they could do the different configurations.

Ambroise: Exactly. So they're trying to solve this equation I was mentioning, which is to say, your chip is not flexible, so you have to design it ahead of time, they can still meet the needs tomorrow. So chiplet is one idea. And trying to make it more modular than that people who just do assemble a lot of chips within a package. And individually, the chip doesn't change that much. But when you're talking about new standards like 5G, Wi-Fi. I've worked on both. Wi-Fi, there's Wi-Fi 6. Now people need Wi-Fi, there's new bands at six gigahertz. Now, 5G evolves all the time. There's release 17, then release 18, release 19.

So it's really tough to do chip. That's why there's few companies, there's a lot of consolidation. There's only a few companies that are really good at it. But those that are good at it, they can really make a lot of money out of it. Because, as you point out, everybody needs chips. You need them everywhere.

Yatri: I've seen this a lot in... so I do a little bit of hobbyist electronics, right? And in the IoT space, a lot of basically all the early IoT stuff was based on one board, like the ESP8266. More recently, we've had the ESP32 come out, and it's interesting because as security has become much more important for deployment, having a device that has a cryptography core onboard just helps a little bit in terms of being able to get TLS certs right on the device. And then you have an added bridge of security right on the silicon that you don't have to worry about and it's low power.

But you see a lot of people who are doing interesting things with that in the hobby space just because it's there, and they're cheap, and they're accessible. And so you get that feedback loop of like, "Oh, well, we could do something with this," and that solves this niche. And then you know, someone will hopefully specialize to bring prices down once it becomes more available to masses.

TJ: Well, and that's an interesting thing too like, this is something that we didn't discuss in the intro, but I think it's interesting to bring up. So like Intel, and AMD being the main sort of desktop CPU manufacturers, so they've got different ranges, right? So they've got their high-end chips all the way down to their low-end chips. And usually, those are the same thing. It's just some of it didn't turn out well in the manufacturing process so they turn a core off. So for instance, if you go buy a 12-core ryzen processor and a 6-core, they're arguably the exact same manufacturing process. It's just on the 6-core, six of the cores were bad, so they shut them down.

Yatri: Right. Like the Intel K series.

TJ: Yeah, those are just really binned chips that are the manufacturing process went well, and they...

Yatri: That's the cream of the crop. And they have a premium.

TJ: And it commands a premium price, because it's gonna get premium performance, which I what you're saying about the manufacturing process is so interesting, because it's a really, really difficult process. And that's what TSMC is, I mean, they manufacture for Intel and AMD. And I don't know if they manufacturer ARM or Qualcomm. I don't wanna say that because I'm not sure. But...

Ambroise: Oh, no, they definitely manufacturer of Qualcomm.

TJ: Yeah, I mean, and so the fact that...I mean, I know for a while AMD tried to have their own foundries, and it just...

Ambroise: AMD had the foundries and they spun them off to create global foundries. And one of the know that there's a new CEO of Intel, ist' announced because activist shareholders want Intel to get rid of their internal manufacturing, because it's become so complex, so expensive, that they believe it should be handled by dedicated companies, and they should focus on the design, and the marketing of the chips.

TJ: And I think it's exactly what Ambroise was talking about earlier is it allows them to focus on one aspect, which is a big broad aspect, the idea of a flexible chip, at the end of the day, they can focus on that if they don't have to worry about the actual manufacture of said chip, which is, I think...I don't wanna hate on Intel or anything, but that might be why they've been stuck on 14-nanometer for seven generations, or however long it's been.

Yatri: I will say it's nice to see AMD be more competitive again.

TJ: Oh, no. I've been an AMD...I've been such a big AMD fan, I used a terrible desktop processor on purpose for many years.

Ambroise: I hope you're also an AMD shareholder, because if you invested at the right time.

TJ: Luckily, I am when I built my new PC, I bought in at a very low price, and it's gone up quite well. So, yeah, I feel good about that.

Ambroise: Good for you.

TJ: But yeah, it's really interesting. And it's kind of speaks, I think a little bit, to what we're talking about is AMD doesn't manufacture anymore. And consequently, once they stopped that, they were able to get back to focusing on what can the chip do in the real world, rather than the manufacture process and are not worried about building the node. And so they ended up having a really successful couple of processor generations. And so hopefully, that kind of dictates where the market goes. All right, fantastic. Now, what do you say we dive into the lightning round? These are fast, quick questions, we'll all answer it. And then we'll kind of talk. First question, Ambroise, technology development innovation that you find the most interesting.

Ambroise: To me, it's really what I mentioned, like this edge machine learning processing that you could add to so many devices. And I view that location would be one of those data that they're processing. And one thing I didn't mention, but I wanted to add it is there's a technology called energy harvesting. And that refers to the ability to use energy from, let's say, solar panels, or what temperature or even radio signals, you know, in the air to power those devices, meaning you don't have to recharge the battery. If you can make your edge chip low power enough, and then you can just power it from the surrounding environment. And that just removes the ultimate barrier.

TJ: Yeah, because battery is such... oh. Yeah.

Ambroise: To adopting IoT chips.

Yatri: Absolutely. Yeah, you can put it in places where you don't have to worry so much. I mean, it's like ideal for things like scientific sensors, things like that, where you have natural motion or other forms of energy.

Ambroise: When I was a kid, I remember I had a calculator with a little solar panel to make it work. And that was in the '80s. That was very advanced.

TJ: And if you covered it up, it would die literally within like five seconds, if you covered up the panel. I remember having this.

Ambroise: Exactly. And now the idea is if you can put a tracking device that works for months on your shipment, your cargo shipment going from China, across the Pacific, to San Francisco, and by truck to wherever, and you have only powered by an all small solar panel and giving you a position every hour, that's fantastic, I mean, for logistics, for whatever, such improvement. And that same device, if you can do smart processing of the data, not just location, but the temperature, whatever, and tell you that shipment has been under the right condition all of that time. I think that's the most exciting. And that's where I think a lot of innovation is gonna go.

TJ: That's huge. Yatri, what's yours?

Yatri: I find it interesting right now how...I mean, it's basically along these lines, but I find it interesting how people are putting I guess the process overall that we're learning so much about how to put together some of these existing circuits. So to break it down a little bit, just the reason Intel and AMD chipsets are as fast as they are now, even though they're butting up against Moore's laws, because they can sort of identify instructions coming in and predictively say it's going to probably go in this direction, and kind of push it down the right side a streamlined series of gates, right. I find it interesting how people are deciding and designing these things.

TJ: That's super cool. I'm not as smart as Ambroise or Yatri. So mine is at CES, they showed off some transparent TVs yesterday, which looks pretty cool. I want a TV that I can mount on my wall that looks like a piece of art. And then when I turn on the TV, it looks like a TV.

Yatri: You got one in front of the regular TV so you can watch TV while you watch TV.

TJ: No, I was just gonna put it in front of my face at all times. It's gonna be great. All right, now, next lightning round question. Ambroise, name your favorite TV or movie robot.

Ambroise: Sorry, the name escapes me. I was trying to find one. I'll settle with R2-D2.

TJ: R2-D2. Oh yeah.

Yatri: I mean, classic. Absolutely, you can't go wrong with R2-D2. TJ, what about you?

TJ: Maybe, I'm going with robot from "Lost in Space."

Yatri: Oh, throwback classic.

TJ: Well, no, no. It could be throwback classic, I was more so talking about the current Netflix.

Yatri: Oh, oh, that's fair. That's fair.

TJ: Who is basically the destroyer from before, but in "Lost in Space." Yatri?

Yatri: I'm gonna with the late '80s, I'm gonna go with Vicki.

TJ: Oh, yeah, yeah. Oh, yeah, that's a win. That's a win. All right. Now, this one is my favorite question of the day. Again. I'm less intelligent than these two gentlemen with me here today. Ambroise, favorite flavor of potato chips?

Ambroise: I like the sweet potatoes, sweet potato chips.

TJ: Oh, yeah. Good win, Yatri?

Yatri: That's a really good one. I will have to say spicy barbecue.

TJ: Spicy barbecue. Good win. I'm gonna go jalapeño, jalapeño chips.

Yatri: Classic there too.

TJ: Little bit of spice. Little bit of spice. All right, guys, it has been our great pleasure having you here. Thank you so much for listening to "Tec Trek." My name is TJ and again, as always with...

Yatri: Yatri. Thank you again for listening to "Tec Trek." I wanna thank Ambroise again. Thank you so much for joining us.

Ambroise: Yeah, thank you to both of you. It was really great to be with you guys and so on.

TJ: Yeah, great pleasure having you. Got to hear more podcasts, subscribe on iTunes, Spotify, Stitcher, or wherever you get your podcast. And to find out more, visit us at Thanks so much, guys.


Episode Links