- 15 hours ago
Category
🗞
NewsTranscript
00:00Why don't you tell us about Qualcomm, because I think it's really a behind-the-scenes company,
00:04because you don't sell a consumer product, but you're in all of them.
00:07Tell us about Qualcomm and exactly what you do there.
00:10Qualcomm has gone through, so we just celebrated our 40th year anniversary,
00:14and I've been through 28 of those 40 years, so every few years or so we have like a DNA mutation,
00:21and at this point in time, I think the way to think of Qualcomm is we provide all the solutions
00:29that power the devices that you carry around with yourself,
00:32going from doorbells to data centers now, as I'm talking of.
00:36So the way to capture all of this is intelligent connectivity,
00:40and the reason why we actually specifically mentioned that, it might be an often repeated phrase,
00:45but our origins were in wireless communication, so we maintain the connectivity,
00:49but increasingly we bring in high-performance but energy-efficient computing into devices
00:54that you and I can carry around with us.
00:56And that's kind of critical, because what we have seen is that the learnings that we have from devices
01:02in energy-efficient computing are equally relevant in data centers these days
01:06as we start paying attention to the overall, you know, the grid implications
01:11and the energy implications of that.
01:14So, and maybe just a word or two on the devices by itself,
01:18when we talk of energy-efficient computing,
01:21the computing can be broken down into three categories.
01:24There's general-purpose compute, that's the CPU, that actually runs all the efficient, you know,
01:31the processing that we do, but it has some applications to AI, but not necessarily all of it.
01:37And then we have the GPU, which is meant, it has a day job of graphics
01:40and all the other rendering that we do,
01:43but our processor for focus when it comes to generative AI is on the NPU.
01:49The NPU is designed from ground up specifically for generative AI workloads.
01:54And so based upon that, today, if you take a typical premium-tier smartphone,
02:00you can run a pretty large model on it, like a 10 billion parameter model on it.
02:05And just to put this in perspective, November of 2022, just short of three years back,
02:11is when the chat GPT moment happened.
02:15That's the time that all of us actually started paying attention.
02:17At least in the mainstream, more people paid attention to generative AI.
02:21But the 10 billion parameter model that you have today easily outperforms those models from back then.
02:28Bottom line, we are at a point where we can run pretty compelling generative AI use cases
02:33and very large models and devices.
02:36Today, whatever could run in the cloud last year, for instance,
02:39you are in a position to actually run that in devices now.
02:41So smartphones, PCs, glasses, they are glasses, and consumer and industrial IoT.
02:48Consumer IoT goes into wearables and automotive.
02:52These are like the broad category of devices that we talk of.
02:54So first things first.
02:56You know, today, a lot of us carry a phone in our pocket.
02:58Maybe we have a tablet.
03:00Maybe we've got a laptop.
03:01I'm wearing a smart ring, wearing a smart watch.
03:06What is the consumer, you know, not the, you know, early adopter consumer like me,
03:12but, you know, the mainstream, three years from now, what are you carrying?
03:17Okay.
03:18So there's a lot that's going on.
03:20But maybe I'll preface with it.
03:21What is it that we want in the devices around us?
03:23We call this as the universe of devices.
03:25By the way, as humans, we've kind of made our choices in some of these devices.
03:28As you correctly pointed out, we wear our glasses.
03:32We don't carry them with us.
03:33We wear our rings and wearables, which is watches.
03:38And then the notion is that, okay, and then we carry our phone with us.
03:41We don't wear our phone with us.
03:43It's a slightly different perspective.
03:45But we are in the business of connectivity and intelligence together in all the devices.
03:50Our watches are, by all means, connected directly to the network.
03:54We can actually have that.
03:55So if I were to actually pick one, just one concept,
04:00that would be glasses that you wear, which are directly connected to the network.
04:05So let's pause for a second.
04:07So I have these here.
04:08These are the meta Ray-Ban display glasses.
04:11These just launched a few days ago.
04:14If you want a pair, good luck finding one.
04:15But what's unique about these is they're one of the first mainstream smart glasses with a display.
04:21So I have this neural wristband here, and you can do hand movements and what have you to control these things.
04:27And they have a screen in it so you can see who's calling.
04:30You can do video conferencing notifications.
04:32And, you know, I look a little odd wearing them.
04:35I just actually took a picture of all of you.
04:38But they're kind of nifty.
04:40So, Jurgo, what's your role in these?
04:42So it's our chip that actually powers these glasses.
04:45So you're inside of these.
04:46They only work because of what you guys developed.
04:47That's right.
04:48And if you kind of think about it, by the way, I mean, I can go all pretty geeky into the tech over there.
04:52But just think about what we are talking about there.
04:54You just said you took a picture.
04:56There's an actual camera in there, and there's like a, you know, you can make voice calls.
05:01You can do lots of things with it, which are kind of, you know, just a few years back, it was not easily to foresee.
05:10And these are glasses that you basically walk into a store and buy.
05:14These are not pretty goofy glasses or anything of that sort.
05:17The form factor is really slick.
05:19And now if you think about once you add even the direct connectivity directly to the network as we talk through it,
05:26and this might be very well the beginnings of the 6G era in some sense.
05:30Every G has a way of defining, the defining device in that sense.
05:33It is possible that this could be one of those devices.
05:36At least that's what we are actually focused on.
05:38But that actually brings in a completely different perspective because what you can do now is run some of the AI processing in the glasses,
05:46and the rest of it can be run somewhere in the cloud, all the way into the data center or maybe somewhere at the edge of the network.
05:53It's distributed processing that occurs.
05:55And that's a very different kind of a world that we're going to live in.
05:58Given that you actually showed your device, I'm going to show one other one.
06:03Oh.
06:04This, this is, you know, last week we had our Snapdragon Summit.
06:10It's something that we do in terms of, and we talked about the latest generation of PCs.
06:14And it's in Hawaii.
06:15It's in Hawaii.
06:17It's a tough life, but, you know, someone's got to do it.
06:20So, it's very well attended, and we actually did it in two locations concurrently, both in Hawaii and in China at the same time.
06:28But we talked about the latest generation of Snapdragon processors and the platforms for PCs and smartphones.
06:36This, by the way, is a full-fledged PC.
06:39This is what, if you think of a Dell or an HP or a Lenovo or a Microsoft laptop, this has the complete processing power in that.
06:49Now, if you, you know, Microsoft is a very close partner of ours, just like Meta, and so are Dell, HP, and Lenovo.
06:53But as we start interacting with devices like this, one of the things that you might have noticed in the glasses, and I think it's far more stark in that, is every few decades, there's a change in the UI between a human and a processing capable, something that's a processor in front of you.
07:14I wasn't there in the 70s when we were talking about this, but the 70s was mainframe, command-line interface.
07:20That's what it was all about.
07:21Sometime in the 80s, that's when my first introduction to all of us, quite a few of us at least, to a mouse, something more graphical.
07:29That's a change in the UI.
07:31It's very different from just a command-line interface.
07:33A couple of decades later, we all got used to a device in which touch became a far more interesting interface,
07:41and then now, today, we all carry around devices with a clutter of apps in that.
07:45But you know what?
07:46You don't actually do that with it.
07:48You use your voice as the UI a lot more.
07:51So, in some sense, AI is, in fact, not only changing the UI, but is the new UI to the devices over here.
07:59I would like to talk to my laptop a lot more than actually search for things or do some things with it.
08:06And so, that's something that, as Qualcomm, we are investing in quite heavily.
08:10How do we make it easy for all our partners to build solutions on top of it?
08:14So, I think from an investor standpoint, people want to know what's going to happen to the phone.
08:20What's the impact going to be on these phone companies like Apple?
08:25What do you think?
08:26My personal prediction on that is the phones are here to stay.
08:29I think they are here to stay.
08:31It's the way that we interact with them that is going to gradually evolve.
08:33It's an evolution because most of the times what happens is that when you have a fairly mature device, unless you have, like, a full-fledged replication of that, there's not going to be, like, an abrupt transition.
08:45It's going to be a gradual transition.
08:47But the real innovation starts coming in from devices that didn't exist before, like the glasses that you just showed.
08:52That's where it's like a clean start.
08:53You can actually start with whatever and then you bring in something new.
08:56So, we think of it as an evolution in the smartphone and a far more revolutionary approach when it comes to new kinds of devices like that.
09:02And to your point, the phone, right, didn't kill the laptop.
09:06I mean, maybe there's some people who don't have a laptop now because they have a big phone.
09:10The tablet, everyone thought the iPad would come in and destroy the laptop, but that didn't happen.
09:15Laptops are as ubiquitous as ever, if not more so.
09:18What is Qualcomm's role in the laptop today?
09:20Because, really, you were known initially with, you know, cellular connectivity and then you have done phone chips, but more recently you're going big in PCs.
09:28That is correct.
09:29So, maybe I'll spend a few minutes on what we are doing in PCs.
09:33So, together with a close partner of ours, Microsoft, and all the other OEMs who are involved in this, last year you saw the birth of what was deemed as AI PCs.
09:44It's the first time that we started talking of AI integrated into the PCs, into laptops.
09:50In a way that you don't, you can run pretty large models and use cases, AI use cases, directly on the device with or without connectivity.
10:00And most of the laptops don't actually have cellular connectivity.
10:04So, this is even in instances where you're literally in an airplane mode and you're still getting some work done.
10:09And the AI PCs, starting from June of last year, they coincided with Microsoft pushing very strongly on the co-pilot features.
10:16So, these days, by the way, there's lots of nice features that are there where, you know, you enter a Teams meeting and then if you join in like a few minutes late, you actually get a summary.
10:25You can ask for a summary and get done with that.
10:28But there's a lot more.
10:29At the end of the day, the laptop is something that we, it's a bread and butter for, you know, code generation, software, content creation, and things of that sort.
10:38And AI is permeating all of those things.
10:42Like, generation of code these days in enterprises, a good fraction of it is, you know, there is dependence upon AI in one form or the other.
10:49At least that's emerging trends.
10:51And as Qualcomm, what we did is make sure that these PCs can run very energy efficient but pretty large models running directly on the platform itself.
11:02This is slightly different from what runs on the data center in the sense that when you have large models in the data center, you know, when you talk of a rack in a data center, that's like 150 kilowatts, right?
11:12And think of what actually powers your laptop and you want it to last for 24 hours.
11:17So, our stringent KPIs were extremely power efficient but high performance CPU.
11:25The ability for our NPU, for AI processing to run at a very low power but very large models.
11:31And there's always a balancing act there between the processing capability, the compute capability, and the actual memory footprint.
11:38There's always a balancing act.
11:39Sometimes you tend to be more memory bound than compute bound.
11:42But bottom line is we feel like we've kind of reignited.
11:47Like today, when we think of laptops now, it's in a different way.
11:51It's not what it was even five years back.
11:53A lot of things have changed quite dramatically in that space.
11:55So, we're in the business of providing all those computing solutions that go inside, not to mention the software stack that comes with it, to make it easy for developers to build use cases and applications on top.
12:07That's a software stack, making it more convenient and more easy.
12:09And, of course, working closely with the Windows team.
12:12So, I'm just thinking about this.
12:14Think about all the companies that we work with today or use their applications today.
12:20Companies like Uber or Instagram, Snapchat.
12:25These are multi-hundred-billion-dollar, trillion-dollar companies.
12:29AI is really going to change the need for these companies.
12:32You're going to see apps starting to go away.
12:34You're going to probably see the app economy be thrown in the air at some point.
12:38I think there's trillions of dollars at stake here because of AI.
12:41I'm curious what you think, how these companies are going to need to evolve.
12:45I think AI has emerged, whether we, you know, whatever you believe in it or not.
12:49I'm a big believer in AI.
12:51But it is here to stay.
12:53And it is certainly very disruptive in lots of things.
12:56The last time we saw something like this was probably at the birth of the smartphone.
12:59And before that, the birth of the wireline internet in the early 90s.
13:03It's one of those moments.
13:04So, clearly, there's lots of opportunities coming in.
13:07When it comes to the apps, as you mentioned, you know, we didn't speak a whole lot about agents.
13:12But there is a trend towards applications at the end of the day.
13:18They all live in their own individual silos.
13:20Like if I open one app and then I open a different app, they don't actually strictly speak and communicate with each other.
13:27And that is not how we as humans think.
13:30Sometimes we might want to do like some multitasking with three or four things put together.
13:33I want to buy something.
13:35But at the same time, I'm just checking whether I have enough in the bank balance or and then meanwhile go from there to a travel app.
13:44The idea of an agentic workflow is something that sits above apps, perhaps being the front end of the app.
13:51So that when we talked about AI as the new UI, your first interface is actually to an agent and not to the application.
13:57And what that means is that it's a pretty complex task that you explain in a very natural language.
14:03And it's the agent's job to figure out what to do from there onwards.
14:09Pick from a rich bouquet of AI models to figure out which one is more relevant for that task at that point in time.
14:15Interface with some sort of a personal knowledge graph that's residing either in the device or somewhere else, but tap into it to get a sense of your preferences and then use that information to see, okay, which apps do I need to process all of this and get back with that information.
14:31And so agents, the way we see it, are increasingly becoming more prominent and as the front end towards the users.
14:41What does it mean for apps?
14:42The apps still exist, and there might be a period of time of coexistence between the both, but increasingly there's a lot of interest in building agents which can actually – and by the way, there's not just one agent.
14:53You might have several agents.
14:54In your device, you might have like an enterprise agent, a medical agent, and things of that sort.
15:00And you can take it to one more extreme.
15:03Next time you open up – I'm just making this up – but next time you open up a specific spreadsheet and so on, you might actually open up a financial agent which then goes through all the financial things, but that's where the specialization is.
15:14So there's lots of opportunity over there.
15:16What it means for apps, I think it's to be seen.
15:18I don't have a good, clear answer on that.
15:21It's always hard to predict the future, in our industry at least.
15:23But what's clear is that agents are clearly taking over as a level above the apps at this point in time.
15:32But here's the big problem.
15:33We have the underlying hardware technologies from companies like Qualcomm that are building these things.
15:37We have the underlying AI technologies from companies like OpenAI, Anthropic, building these technologies and the ability to build these agents.
15:44But then we have this duopoly of operating systems that are a conduit to these devices across the world.
15:51You have Android and iOS.
15:53But neither Android or iOS – Android to more extent – are implementing this new user interface that feels inevitable.
16:01So how do we get there if these companies want to sort of control that moat and keep the app ecosystem alive when it's pretty clear this underlying technology is popping through from below?
16:11I mean, just to be very clear, you forgot to mention Windows, by the way.
16:15Windows?
16:15Who uses that?
16:17We, you know, as Qualcomm, Microsoft and Google are very close partners of ours.
16:22So we actually work closely with them in terms of going through – yes, it is true that if you were to actually unpack that a little bit further,
16:30these days when we think of AI and if you think of models, AI models are actually coming in from all over the place.
16:37And not necessarily just from – you picked some names like OpenAI is obviously GPT-5 and most recently they even announced the GPT-OS is 20 and 120 billion parameter models in open source out there.
16:51It's a new trend.
16:52But then – and then we have Anthropic.
16:55We have a few other model creators.
16:58But it's also important to understand that model creators exist all over the world.
17:01We actually work very closely with the model creators very recently, and you'll see us talking a little bit more about that.
17:07But, you know, both in the Middle East, in Japan, in India, and most certainly in China, there's a lot of other model creators who are in that language.
17:16But they also have a different way of actually looking at things.
17:19So model creators are coming up independently of, you know, everyone who has the existing OSs.
17:26And then there are certain devices where actually it's none of these OSs.
17:30In fact, Linux, open source Linux is actually used quite a bit.
17:33So it's very interesting times to see how this is going to evolve.
17:37Our take is that at the end of the day, all of our partners are very eager to bring in rich user experiences based upon what they have.
17:45And we are in an ecosystem where if you start seeing far more compelling use cases coming in from elsewhere, everyone tries to catch up over there.
17:54So personally, I'm not too worried about that one.
17:56Really looking forward to seeing all these use cases that I talked about come in.
18:01So if you think about going back to hardware, going back to the devices, right, like my arms are getting kind of filled up, my pockets are filled up.
18:09I mean, something's got to give, right?
18:11People don't want to charge seven, eight, nine things every night.
18:14Maybe you don't have enough outlets or desire to do that.
18:19What's going to win out?
18:21That's a tough one.
18:22The one thing that we definitely want to do is make sure that they all coexist and work in conjunction with each other.
18:29And does it mean that you have like a single vertical ecosystem?
18:35Not necessarily.
18:36There are protocols that are emerging right now in which devices can communicate with each other in the right way.
18:41You are right about one thing that we talked in the beginning, which is there's a lot of things that we wear with us, and those things are kind of taken.
18:52Now, maybe there is still some room for a pendant tomorrow.
18:56I don't know.
18:57Did you see this thing called the friend?
18:58Has anyone seen that?
18:59It's a necklace, and it's like an AI that you just wear on a necklace.
19:03It's kind of ridiculous.
19:03We've seen one instance of this like a couple of years back, but I like the fact that there's a lot of experimentation going in that space.
19:11I mean, we wouldn't be here with that sort of a device with the air glasses unless we keep experimenting.
19:15But those devices will continue to happen.
19:17Now, in terms of what else you can carry with us, well, of course, we have a smartphone, but it remains to be seen.
19:23I mean, what's the appetite for us to carry even more devices in addition to a phone, a laptop, and maybe another puck or something like that.
19:31But, yeah, there is probably a threshold over there.
19:34But we think of it as the universe of devices that we have chosen for ourselves.
19:38And just from us, we want to be the ones powering all of those platforms.
19:42And just from us, we want to be the ones powering all of those platforms.
Recommended
5:27
|
Up next
1:58
1:57
5:22
4:58
11:44
3:52
6:24
8:04
3:50
1:46
5:19
7:01
1:48
6:03
14:36
2:42
14:25
4:28
17:41
4:05
3:49
14:55
6:49
Be the first to comment