Futurum Research: The State of AI PCs

I'm Ron Westfall, research Director of Communication Networks at the Futureum Group. And so, yeah, I'm naturally delighted to talk about our primary research. And today we'll be looking at A IPC specifically.

And to kick it off, I'm going to provide an overview of the Future Room Group. Uh, Tom did a great job of talking about, you know, some of our, uh, different, uh, groups and we already know Tech Field Day, here we are, and all the benefits that flow from it. We're certainly delighted about how, you know, the presenters, the delegates, the community, all benefit from Tech Field Day, such as Mobility Field Day 13 here today.

Um, Tom touched on Six Five Media, and I will elaborate that we're actually having a six five summit in mid-June, so please tune into that. That's where we're getting some great insights from folks like Michael Dell looking at, you know, why AI and so important to the entire industry Techstrong. That definitely is, uh, deserves a shout out.

Our, uh, segment today is being streamed on Techstrong as well. And what Techstrong, uh, brings is the ability to target specific audiences, certainly technical audiences, but they are, uh, definitely a capital a amplification capability. So that helps assure that when we deliver content, that it's really getting out there, that the audiences are receiving it.

And Signal 65 is our lab. We're excited about that. We're doing lots of great testing with, you know, a an entire segments of the industry, including the processor, uh, guys, the switch guys, the storage guys, et cetera.

You name it. And what I'm going to drill down on is future research and intelligence. That's where our primary research is housed.

And so I will cut over to yes, uh, what we do with our research. It's basically fueling our research and advisory, uh, mode. That is, we use primary research across 11 practice areas, AI devices being one, but we also have AI software and tools.

We also have C-E-O-C-I-O insights. We also do enterprise applications, uh, cybersecurity. So these are naturally the practice areas that are aligning with, you know, what our customers are asking for.

And so today it's very helpful to look at, you know, what we're doing with AI devices. We take our primary research and we combine it with our C-suite advisory capabilities, and that has outputs such as, you know, working on go to market strategies, developing product development objectives, uh, how to optimize a marketing campaign and so forth. Alright, so let's dive right in.

And this is, uh, our data from the A IPC study that we, uh, completed recently. And naturally, what's important to, you know, provide a level set is, you know, what are a IPCs? And fundamentally, they are PCs that are integrating dedicated AI accelerators such as neural processing units or NPUs, and they are really tasked to handling things such as gen AI and machine learning, deep learning and so forth.

And, uh, I'll drill down more on that, on, you know, what are, you know, some of the other requirements and, you know, examples of a I PCs and, uh, that I think, uh, links to, okay, where are we getting this data from? I think Tom did a great job of pointing out that we are doing our own research. In this case, it's an online research that reached out to 852 enterprise decision makers and influencers.

So we have that critical mass. You know, this is, I think, data that we could be very confident about representing what's going on out there. 3 billion or 68% of the worldwide A IPC revenues.

And that stands to reason, you know, most folks purchase laptops both on the consumer side as well as the enterprise side, but that doesn't leave out the fact that desktops and workstations are definitely categories that are generating their own important revenue streams. And I think what's important to note here is that when you look at the workstation segment, which is 17% of the revenues out there, it's really adhering to the demand out there for high end computing for, you know, AI, workload optimization, AI inferencing, as well as a specific category such as scientific research and professional content creation such as animation. Ron, uh, are these numbers, are they encompassing of all operating systems shipped, uh, Mac Windows Analytics?

Yes, that's an excellent question. And, and some, yes. Uh, and specifically the manufacturers that are included in this, uh, study include Apple.

It includes, uh, Dell, it includes hp, Lenovo, ACEs, really, you know, the, the, um, uh, usual candidates, but it does exclude is, uh, China. Uh, so to be clear about that, it's worldwide except for China and Russia. All right.

Now on the next snapshot here, we're looking at how the market's breaking out by pricing. And I think we understand, you know, why the low end segment, which is defined as below a thousand dollars, has traction. And that is it's budget friendly.

It's, uh, basically, you know, what many, uh, consumers and businesses need in order to, you know, just get on, uh, to the internet. But really the Goldilocks range is in that medium range. And that's, you know, the part, uh, that is, uh, you know, dominating the market.

5% of all revenues is coming within that range. Again, it's across the board. It's, uh, both, uh, laptops, workstations, and, uh, desktops.

And yes, the high end, uh, is clearly correlating to what I noted about that they handle the most demanding, uh, performance, uh, uh, thresholds and are dedicated to, you know, ni uh, niche categories, uh, such as, uh, design teams and so forth. And, uh, so that I think is helpful to understand, you know, what is going on, uh, with, uh, the pricing of a I PCs. Okay, next, we're looking at revenue by performance.

And again, this is something that we found is an high demand. You know, how is this breaking out? And I think one thing that's important to note here, when you're on our Futureum intelligence dashboard, you can flexibly correlate, okay?

I wanna know how, uh, the performance of this category is, uh, connected to, say, the pricing category, the geographic category, and so forth. So it's just an intuitive tool to allow you to use the data the way that you want it. And, uh, what we discovered here is that the 20 to 30 tops category or, um, terra operations per, uh, second is, uh, the primary one.

88%. So clearly that's, uh, the highest demand one. But what we're anticipating is that the 40 to six tops category, the highest performing ones, the ones that actually meet the threshold that was defined by the launch of the Microsoft Co-pilot framework, and you, uh, 40 tops or more, that I think is going to grow dramatically.

And, uh, also correlating to, you know, low cost, uh, uh, PCs is the fact like, uh, the lowest performing ones that're available in the market below 20 tops. Uh, but that's something also we're gonna anticipate seeing more of as more models, uh, hit the market and are geared toward, uh, cost affordability. Ron, I've got a quick question.

Looking at the graph that you put up before where the primary, um, source of revenue in the laptop market was from the one to 2K market, which a lot of people would probably consider to be more like the budget or affordable option, is it that manufacturers are putting, are over specing the NPUs that they're putting in these machines in the hopes of being able to unlock future revenue from AI software by saying, I'm not gonna put the cheapest thing in, I'm gonna put maybe the next step up because then I can, I can kind of maybe charge for like, I don't know, a subscription revenue in the future? Yeah, excellent question, Tom. And I would say, uh, there certainly the latter is in play.

That is okay, we're se uh, setting up the, uh, ability to, you know, uh, upsell, uh, more AI software, uh, capabilities that, you know, again, enable that AI workload optimization, particularly when it comes, uh, comes to ai inferencing, comes to, uh, ai, uh, edge workloads and so forth in, uh, essence, avoiding having to use the cloud, you know, for anything that can be done locally. And this is, you know, that fundamental principle that's we are seeing the era of hybrid ai. It's really not a cloud only proposition or is obviously an on-prem, on-premise only proposition.

It's the blending of the two, but we're seeing increasingly more and more of the workloads being processed at the local and edge level. Also, uh, and I'll touch on this, uh, Qualcomm basically fired up the category, and so that injected a, a competitive new dynamic that forced, I believe the pricing, uh, down in terms of, you know, let's get a I PCs kicked off and, uh, but we'll see, uh, what I touched on that, uh, more diversification of the price ranges as the market matures naturally. And just to clarify for everybody at home, this is only devices that have a dedicated AI offload chip, like an NPU, this doesn't count correct?

Anything that uses GPUs to do AI work that is not integrated into the system. Yeah. Uh, for this study, that is the correlation.

That is how we are looking at the A IPC market. And so again, it's, uh, like the, uh, the co-pilot framework, there are some, you know, specific requirements in order to qualify as an IPC. However, as we all know, it's not unique to the, uh, PC market.

They're marketing games out there that make it seem like, okay, this is AI PC ready. So like when we saw, you know, uh, 4G capable phones being marketed as 5G uh, phones on say the at t network, however, you know, the, uh, 5G standalone implementation had not actually been, uh, delivered. So, uh, that's, you know, uh, something to be aware of as we're looking at how this market is evolving.

Alright? And yes, this is having a direct impact on the decision making at least of, you know, enterprises. And that, uh, is the fact that we see that 57% of organizations strongly believe that AI PCs are the next evolution of PCs.

And as a result, they are looking at over the next two to five years to really replace their existing PC fleets with I-A-I-P-C fleets. And so, uh, that doesn't deny, uh, if you look at, you know, the lower bars that there's skepticism out there, this is not, you know, across the board, it's not an overnight sensation. You know, uh, there are many folks who are registering their skepticism about having to have an A IPC.

However, if you're, uh, uh, are a a I PC manufacturer or somebody who's supplying processors, Jimmy Crack, Korn, you don't care because 57% is a pretty remarkable critical mass target. And as a result, we see, uh, that this is going to continue driving, uh, investment decisions to adopt a IPCs, but certainly, uh, the PC manufacturers are gonna bet more heavily on a IPCs growing more and more. And, uh, to, uh, Tom's question about, okay, what is, uh, distinct about A IPC?

Well, one thing that is important is that it's optimizing power conception, a consumption technique such as target wait time, which aligns with wifi seven's energy efficient protocol. So this is tying in, uh, you know, why, why there's there impact on the wifi wireless market due to A IPC, uh, evolution. And I'll, uh, drill down on that more in, in the next, uh, few slides.

But also, uh, it's reinforcing the fact that, uh, to the ability to more, uh, effectively implement WiFi's WPA A three and, uh, o encryption capabilities. And this is something that we anticipate this is going to, uh, help, uh, trigger implementations of those capabilities with those, uh, you know, uh, encryption techniques embedded within the A I PCs that are coming onto the market more and more. Alright?

And this is important. It's, uh, I think organizations believe, okay, if we're gonna have a successful AI journey and that we can, you know, deliver proof points that our AI strategy is meeting the bottom line, improving business outcomes, we have to have a IPCs as part of our overall strategy. If we're not already adopting them, then we certainly need to plan on when to adopt them.

And so this, I think, is, uh, good data as to explaining, you know, why it's, the perception is key. It's, uh, a IPCs, yes, there's the marketing bus surrounding it, but also there's the reality on the ground that folks are making this a high priority in terms of their ability to, you know, support, uh, you know, critical applications. And, you know, we've seen, you know, for example, ar vr, uh, having been over hypes before, but we expect that those capabilities with a IPCs will become, uh, more mainstream already been talking with not only organizations out there but also, uh, the makers.

And they're talking about scenarios like a construction sites where AR is now critical to their ability to, you know, make construction, uh, processes more efficient, uh, improve again, uh, the business outcomes and so forth. And likewise with manufacturing facilities, smart factories, and so forth. And I think it's also important to note here, we're doing more research on this as well, that a I PCs correlate to improved workforce experience, and we know that workforce experience is, uh, important.

Uh, it will be interesting to see just how strongly it correlates with improving customer experience, which is always important to, you know, the organizations out there. Um, but I think, uh, what is making the difference is many of us have already personally have used, you know, gen AI capabilities on our existing laptops, but it's, uh, critical for streamlining content creation, uh, for improving data analysis and so forth, in essence boosting productivity. And that's, you know, part of the, uh, answering the question why AI PCs, okay.

And this is something that, uh, we definitely had demand to have a better understanding. It's the, you know, recycle, uh, calendars out there. And as a result, how can you know the makers out there better target this, you know, take a better advantage?

And, uh, one, I think important takeaway is that 90% of the folks we talked to or surveyed is that they have accelerated, uh, their PC refresh cycles to take advantage of A IPC capabilities. And I think it's embedded right there clearly on, uh, the chart here. And what's also, I think, uh, fortunate is that it's aligning with the fact that Windows 10 is being sunset, uh, later this year.

And so that is an added impetus. It's not only about, okay, uh, we need to take advantage of AI PCs, but hey, goody goody, this is aligning with the fact that we want to upgrade at Windows 11, and hey, with Windows 11, we can definitely, uh, support, you know, uh, these AI PC capabilities across the organization. And again, it's, uh, a heavy focus on being able to optimize AI workloads at the local level.

And I think it's also important to note that it will, uh, accelerate integration of, uh, mu, uh, MIO capabilities as well as, uh, having on device hyper-personalized ag agentic ai, uh, user experiences now becoming just that more mainstream. So we're really at the front of this. Uh, and so I think it's exciting that we're gonna see, you know, how IPCs can deliver innovations that, you know, have, uh, previously been stalled out because of the limitations of previous, uh, pc, uh, form factors.

And yes, Do you, it's not a wifi question, it's more like a 5G question, but do you know if they're also planning on integrating like 5G natively on those new air AI laptops? Uh, the short answer is yes. And, uh, yeah, we've had conversations, and that could be specific with folks like, uh, T-Mobile, uh, Intel.

And, uh, the fundamental reason is that connected laptops are becoming more important for organizations that do have top security requirements. And that is some organizations, uh, perceive wifi as not being the ideal, uh, security implementation for their connectivity. So there have become new requirements that you have connected laptop capabilities, 5G uh, specifically, and that includes government agencies, but also contractors, uh, with, uh, government agencies.

Uh, we've seen, you know, some serious breaches arise not because of the agency itself, which you know, can certainly happen, but because of the contractor loophole and where there's, you know, uh, I guess you can say, um, more opportunity for, uh, enabling a security breach. And, but the, the other factors, uh, a standard reason is just having 5G mobility capability where there's no wifi wouldn't be great just to have a laptop that you can continue to use using a 5G network. And, uh, that is, you know, the drivers, that market has also had its hype cycle, but we're anticipating the A I PCs are gonna help lift, uh, connected laptops as a, uh, specific market segment.

Ron, I've got a quick question about that because one of the questions that kind of came up is how much of the, um, how much of the AI focused things are actually hitting the NPU on the laptop versus being offloaded to a cloud service, you know, cluster type thing. And I know that Microsoft copilot specifically requires the use of that Qualcomm snap grabbing NPU to do some things, but they also hinted that they were maybe gonna release a patch for a MD and Intel machines late last year. Do you see companies like a MD and Intel being able to kind of muscle their way past this desire to put an NPU in everything?

Or do you feel like that the coming wave of PC refreshes is basically gonna leave them in the dust and they're gonna say, we're not gonna back port this because you're gonna need to buy a new machine anyway? You might as well wait, I think you saw my slide deck. Excellent question.

I'm going to talk about, uh, processor preferences at like the next slide or two. Okay. And so I can, uh, drill down that on, uh, more, but the short answer is, uh, yes, the A and Ds, the Intels are gonna have enduring plays in this market and that they will diversify not just the processor choices, but also the A IPC models and capabilities out there as a result.

Alright, I won't spend too much time on this one. It's, uh, basically reinforcing what we're talking about. Not only are the plans out there ramping up, but also the testing of it.

As we can see, 83% are actively testing IPCs. And I would say, uh, virtually almost all of them are gonna be planning to do that. And I think a, a couple of important takeaways is that we're seeing at the front end, this are the manufacturing logistics, uh, also healthcare segments, uh, for, you know, the fundamental reason that downtime or errors in those particular, uh, areas, uh, can have harsh consequences and thus, you know, uh, the, uh, emphasis on, on, uh, adopting A IPC sooner.

Also, another important factor here is that with a I PCs, uh, it's better at data protection than what Tom was talking about. Okay. How much, uh, is the cloud gonna be involved?

And, uh, to address that here, this is an primary example of where it has to be processed locally for things like HIPAA requirements and of course, uh, GDPR and so forth. And we're seeing, uh, more requirements at the sovereign level that you have to run your cloud within our territory, within our, our country, but also, uh, that we're seeing the capabilities such as sovereign sass e being able to support, uh, these capabilities so that, uh, it's also lifting really, uh, the security, um, component there in terms of sovereign sass e becoming a more widely implemented with a I PCs. So yeah, this is all exciting in terms of, okay, it's creating, uh, opportunities not just for the hardware software guys, but certainly for the security folks and I, I would like to address now, okay, cloud, uh, versus the edge, yes, uh, the GPU training, uh, or the training of, uh, AI on GPU clusters happens at the cloud that has to be implemented there.

But, uh, increasingly once that data is trained, it can then be inferenced at the local level and many of the, uh, data processing requirements, uh, will become increasingly localized. But, uh, at an overall level, it's gonna be hybrid in terms of, okay, how can gen AI be optimized? But when it comes to, you know, security conscious organizations and so forth, there will be increasing requirements to make all the inferencing done without using the cloud or minimizing the cloud.

Alright? And here we are, the preferred processor question. And, uh, so we could see Intel continues to have a strong foothold, you know, they're the incumbent, uh, so that stands to reason, but they have a battle in their hands.

And if you saw this same chart five years ago, it would've been dramatically different. And so this is good news, uh, for, you know, everybody, uh, consumers, uh, enterprises, this competitive mix is making those price points more competitive. And I think what we're seeing here is that, uh, Qualcomm, uh, basically infused, you know, with the Snapdragon X offering, uh, uh, really, uh, the starting of the A IPC category, uh, when it was certified for Copilot Plus, and therefore they were really, uh, from our perspective, the early market mover, and they gained a competitive advantage in that.

Now, uh, A and d and Intel have subsequently come out with their own, uh, rise in, uh, meteor Lake, uh, slash um, core ultra capabilities to address the A-A-I-P-C market. And, uh, but I think, uh, what's interesting here is that even though Intel has, uh, this advantage, it's becoming, uh, more, uh, diluted. In fact, uh, what we're seeing here in the breakout is that 19% will be, uh, going primarily, uh, with a MD Qualcomm 15% and Apple with 11%, uh, because of their distinct, you know, ecosystem.

And so, uh, this is, uh, I think something that we're going to be keeping a close eye on. Uh, we'll see how successful Intel will be at, uh, defending its, uh, foothold here, but we're seeing already that, uh, the diversification is well, uh, kicked off and, uh, we'll see how much of that will, uh, play out. Alright, tying into specifically wifi and, uh, wireless.

And as we know, a I PCs act as powerful edge devices, and that's, uh, not just about, uh, you know, enterprise environments, but also, uh, being able to, uh, optimize iot devices, uh, you know, enabling smart homes and so forth. We see the use cases in these areas as improving if you have an A IPC implementation as an integral part of it. Also, uh, a I PCs, uh, are built to enable AI algorithms to monitor and optimize, uh, wireless networks in real time.

Uh, this is something that I believe, uh, we're seeing already with the wireless manufacturers, uh, being able to take advantage of these capabilities, many of which were, uh, really assessed, uh, with great detail in our conversations over the last three days. And I think, uh, it's also important to note that IPCs, uh, fundamentally will improve, uh, uh, latency, again, because of the ability to, uh, integrate, uh, wifi seven capabilities on a, um, uh, a native basis. You can say.

Alright, I touched on the fact that when Microsoft copilot plus, uh, PC framework came out, that really was the genesis for the AI PC category to take off. And as we're now seeing is that ai, uh, powered wifi systems on a IPCs, uh, also deliver the ability to improve anomaly detection, uh, identify rogue access points, as well as automate, uh, threat responses, remediation and so forth. And so this is definitely, you know, in that top three of, okay, when we adopt IPCs, what is it going to help us solve?

Well, security issues, at least the ones that, uh, that we know of today. And also, uh, we're seeing that, uh, with IPCs, uh, can help propagate wifi system capabilities that take advantage of AI to improve dynamic channel management, application prioritization, efficient device handoffs and so forth. So this is something that I think is being welcomed by the organizations out there.

Also, I think it's important to note that it's having impact across the, uh, overall digital ecosystem that is Adobe and Zoom, for example, have optimized our applications already for on device ai. And so I think, uh, we're gonna increasingly see the benefits of that over, uh, not just the rest of this year, but certainly beyond. Alright.

And here I will wrap it up. You know, our key, uh, key takeaways, uh, in some, uh, on device is like certainly generating more hyper personalized agen AI user experiences. And we touched on the use cases, uh, the segments and so forth that are taking advantage of it.

Also, uh, it's going to be a lift for improved security and data protection involving certainly wireless and wifi capabilities. Also, uh, AI will be a friend to power management and enhanced rf uh, connectivity. And I think one thing, uh, came out certainly is that when it comes to thermal management, that's vitally important.

And so AI PCs will play its role in terms of helping not just, you know, uh, the thermal management at the PC level, but organizations are increasingly looking at their, uh, network wide, uh, thermal requirements and how that will help them fulfill their sustainability goals. And yes, uh, to, uh, round out answering that question, you know, AI devices, uh, do not pull from the cloud nearly as much as previous devices. And as such, that is gonna be a major driver for the era of hybrid AI and all the benefits that can accrue to the wifi and wireless world as a result.

So with that, uh, thank you and naturally jump right in if they have any thoughts or questions. Just, I have a thought, um, with your last, your last statement, I was thinking, um, you know, we have a lot of wifi issues, uh, client issues in wifi, so I don't know if one day we'll be able to use that local AI to improve the way the wifi driver works and connect the network better connect to the network so we can have less wifi issues. I don't know if it's possible, but Yeah, it's an excellent question.

Well, the short answer is yes, at least you know, that is, uh, certainly a part of the product development prioritization, uh, with folks, uh, such as Qualcomm as well as, you know, the, the major, uh, PC manufacturer, certainly, uh, Intel. And that is, uh, using, uh, the ai, uh, EC capabilities, the ai, um, specifically, uh, to address, uh, the client, uh, issues on, on the device. And, uh, quite simply improve the visibility, improve the ability to troubleshoot of that, uh, you know, frequently troublesome area.

Kinda add on to that, is Qualcomm's also pushing their AI chips into the edge devices like today? Some consumer aps? It is.

Can you address more of the issues of not just the PCs, but Edge devices will have AI as well, and, and how are those gonna be used? Yeah, no, uh, they are prioritizing that. We see them, for example, uh, going after, uh, industrial, uh, innovation as one of their top priorities that is, uh, the IOT, uh, segment as well as the automotive segment as well as, you know, consumer, uh, AP devices.

So, uh, we are actually, uh, uh, I would say impressed with, uh, Qualcomm's, uh, recent quarterly results that reflect this ongoing diversification. So yes, they will continue to naturally be, uh, the primary supplier for premium smartphones and, uh, devices similar to the smartphones, but also, yes, they are making, I would say, an effective push into, uh, these additional areas, uh, to really, uh, take advantage of hybrid AI capabilities. And perhaps they might solve both mine and franwell's questions at the same time, the AI and the AP can talk to the AI and the client and figure out to make it work.

Yes. And I had kind of a similar question. Do you foresee, so kind of in the self, this idea of self-management of that PC device, instead of that AI processor being used for the applications running in the system, do you, is there a future where they're gonna help manage the functions of the system itself?

Right. It's, uh, aligns with, uh, autonomous, uh, capabilities. And the, uh, the short answer of from our discussions is that is something that is a goal that's being invested in.

Uh, we're not quite there today. Uh, however, uh, we just have to keep, you know, kicking the tires, doing the testing that, you know, we, uh, touched on here. And, um, I think, um, virtually everybody would be pleased if they have a desktop that could be more self-healing and autonomous and actually having it work, uh, the way it's, uh, uh, being presented.

You had mentioned FUTURUM 65 i I, one of the 60 fives, the the ones doing the, the lab work. Have you tested these PCs to see if they're hitting those tops numbers or is that just what comes off their spec sheets? Uh, Yes, uh, signals Signal For, uh, the testing.

I don't know about the, if we've done A IPC tops specifically, but I do know that we've done, uh, a testing of the, of the processors and we've done testing of, uh, you know, switches and uh, uh, storage, uh, offerings. And, uh, but, uh, we'll gladly follow up on, you know, here is the roster of things that we have tested and is out for, you know, public consumption.