AI Reality Check: Why Apple & Tech Prices Are Skyrocketing | Utilizing AI Episode 34

The impact of AI is coming home, impacting businesses, manufacturers, and consumers in the form of higher prices for everything. This episode of "Utilizing AI" focuses on the impact of the hardware build-out, featuring Frederic Van Haren of Hyfence, John Swartz of Techstrong, and myself, Steven Foskett. Welcome to "Utilizing AI," the podcast focused on practical applications of artificial intelligence from the Futurum Group.

Every Wednesday, we explore news and use cases of the ways in which AI is transforming enterprise IT and the industries it serves. I'm your host, Steven Foskett, President of the Tech Field Day business unit here at the Futurum Group. Before we dive into this discussion, let's meet who's on the panel today.

Well, it's great to be here. I'm Frederic Van Haren. I'm the founder and CTO of Hyfence, and we provide HPC and AI consulting services.

Hello, I'm John Swartz. I'm in Silicon Valley. I'm a senior content writer for Techstrong Group, which is part of the Futurum Group.

Which is all this big, happy family that we're all under the same umbrella. So it's... Happy to be here, and I think we have a great topic today, Steven.

Absolutely. And John and I sort of kicked this off a little bit on the Techstrong gang, which you can catch every weekday as well. So, I'm going to kick this off with a story.

Let me tell you all a story. Last week, Apple raised prices on MacBooks and iPads and many other devices. " Mine is five years old, and the new OS won't support all the Apple Intelligence features natively on the original M1.

So I need a new MacBook. I better go get one now. So I ran up to Micro Center.

Now, this is not a Micro Center ad, but let me tell you, Micro Center rocks. I ran up to Micro Center. They have MacBooks in stock.

They actually had discounted ones. I was very excited. I picked one out.

I got there, and it had already been bought out from under me. " Nope, bought out from under me. Instead, I ended up picking a 48 gig M5 Pro, which is still a pretty solid machine, and Micro Center had it discounted $450 off list, which made it a pretty affordable machine, too.

I think it was $2,300. So I said, "Well, it's for work. " And how they do it there is they go get it from the warehouse, they bring it to the register, then you have to wait in line at the register and pay for it.

So I'm standing in line at the register and I hear the lady that sold me the Mac come running up to the register and saying, "Oh my God, Apple just raised prices in our system too, and canceled all of our sales. " And I'm like, "Hey, that's me. " So instead of $2,300, it's $3,300 in between saying I'm going to buy that, literally while I was standing in the register.

And that's a lot, even for a business to spend. And so, luckily, they did honor the price. I did get the thing, but I'm literally the last person to get the old pricing.

And it occurs to me, holy cow, Apple just raised the prices by $500, which is a huge amount, because of RAM and SSD costs. And of course, we know why that was. Apple raised the prices because RAM and SSD costs are incredibly high.

They released a statement where they literally spelled it out. " Now, AOC and Bernie Sanders say that's nonsense, but it's true that eventually the demand for chips, for RAM, for flash, it reaches a point where people can't afford to continue to subsidize these things, and suddenly, we have just unsustainable prices. I'm going to say that I would not have spent $3,300 on this MacBook.

Cory and I were just talking, his needs service, and if we had to buy a new one, that would be more like a $6,000 laptop. Well, that's just not possible. And I think that this is what's happening now with AI, is that the bill is coming home finally.

The cost of this thing is finally coming down to consumers, whether it's the cost of a new MacBook or a new iPad, or the cost of your next phone, or the cost of your next car or television. Or maybe it's just the cost of tokens, the cost of applications, the cost of energy. There's so many ways in which AI basically has been subsidized and is no longer going to be subsidized.

So, John, I want to throw this to you first. You're the Silicon Valley reporter. What do you think about the cost of AI coming home to roost with consumers?

Well, it's going to hit hard with consumers. So first of all, the consumer is intimidated, in fear of their job security. They don't quite know what AI is.

And so they, for the most part, a lot of them, even out here, have tried to avoid it. Now you can't, because as Tim Cook had pointed out, I think a week before they raised their prices, he warned Wall Street. He said that this global shortage was, he called it a hundred-year flood in the supply chain.

" And so it was passed on. Those costs have been passed on to the consumer. And in an era of affordability where we stress out over gas prices, meat, eggs, what have you, now we have this component shortage which is leading to higher Apple prices, which really hits the consumer more so than any other product.

It's going to go across the board, not just with Apple, but with others. So it feeds into this affordability issue. So it kind of comes home and establishes AI as this unavoidable behemoth, this boogeyman that is not only threatening everything we do in life, but also hitting our pocketbook.

And again, this is not an anti-AI screed. This is just a reality check, and it's something that just happened and is going to accelerate over the next several months. Well, first of all, lucky you, Steven, getting the deal of the year.

I wouldn't say the deal of the century, but I would say the deal of the year. The supply chain is an interesting fact because it's really like a domino game, right? Early on it was the GPUs, then it was the DIMs and VME drives.

Today, even a hard drive is part of the supply chain problem. And I think it has to do with the fact that people, and when I say people, it's the AI providers, instead of providing the highest and the best technology they can, are switching over to a business model. More to your point, where it's not about subsidizing, it's more about coming up with a business where profitability is maybe not something for today, but at least something for the foreseeable future.

Certainly, if all of them are going to do IPOs, just like SpaceX, there has to be some kind of a profitability in the future. And I think also AI today is more of a consumption article than it was a few years ago. A few years ago, it was more people kind of developing high level, specialized applications.

While today it's all over the place, right? To John's point, not everybody who's using AI today understands AI in such a detail that they can have a conversation about the underlying layers, but nevertheless, it's all over the place, and consumption kind of drives the whole effort. Plays also into this, Frederic, I'm glad you mentioned SpaceX because that was a really big story.

More so because it created the first paper trillionaire with Elon Musk, and there was all this hype around it. But the reality is that aside from maybe establishing data centers in space in the long term, there are a lot of questions about SpaceX and its ability to quantify or to rationalize its market valuation, which is absolutely insane. So the stock has consequently gone down.

So there's this thought, oh, okay, there's a land rush. OpenAI is trying, interestingly enough, OpenAI wants to go public. They wanted to go public this year.

They're not probably in all likelihood until next year because of some financial issues that were leaked through the information and other outlets. So there's this kind of equation. All these companies kept telling us a year ago, "Oh, this is going to improve our ROI.

We're going to be more profitable than ever. " And yet we see these token costs and how these companies have burned through that budget. We're looking at, and especially OpenAI, the revenue reality versus where they would like to be, which underscores the fact the IPO isn't probably going to be as hot as people initially thought it would be.

So there's kind of a retinkering, and then we see these Apple price increases. In other words, it's costing us more so, and even I'll add another one, data centers. We can go down that road in terms of the impact on consumers and this whole debate over what it means for our energy bills if we are to subsidize the infrastructure of AI.

Yeah, let me take up the SpaceX IPO, I think, because that is actually a really illustrative point, isn't it? Because I think if you ask mom and pop America what SpaceX is, they'll talk rockets, right? That's going to be the thing that SpaceX is to them.

But those of us in the industry know that SpaceX, the rocket business, is actually not the major contributor to the valuation of SpaceX. 3 trillion by 2030. So essentially, and I'm sorry, that would be the SpaceX share of the TAM.

In other words, SpaceX figures that they could be earning literally trillions from AI in just a few years. They do not claim that they're going to earn trillions from space launch and orbital services. In fact, they estimated that's about $100 billion of revenue contribution.

And actually, the Starlink satellites is actually a bigger contribution, $150 billion toward their revenue picture, according to industry analysts. So, in other words, the SpaceX IPO is not a rocket IPO. It's not even a communication services IPO.

It's an AI IPO. And interestingly, SpaceX, as John and I have talked about on "Techstrong Gang," is not actually using AI the same way. They're not actually a direct competitor for Anthropic and OpenAI and the others.

In fact, they actually resemble a Neo cloud more than they do ... a foundation model developer in that they have built these colossus data centers in Memphis, and I think in other places, but certainly in Memphis. And those are actually being rented out just like a company like CoreWeave might, and that's a big revenue contributor for SpaceX.

They're actually making a billion dollars a month, I think, on renting out their AI infrastructure. So essentially, SpaceX is not a space story, it's an AI story, and it is another example of the fact that this revenue has to go somewhere, has to come from somewhere, that there is a real flow of money happening here, and that that money is not necessarily going in the places that we think it is. Right, Frederick?

Right. Exactly. Right after their IPO a few weeks ago, they made their intention clear to acquire Cursor, which is a company that develops coding tools, which has nothing to do with space as we know.

But I think it's all coming down to making it more of a business under the umbrella of claiming to be an AI company. AI is such a broad term and also a relative term because as we talk about AI today, it's a quite different conversation from AI conversations we had maybe a year or two ago. And who knows what kind of a conversation it's going to be in two years.

But you have to give it a name, I guess. So it's SpaceX for now, but it's AI and all driven around business. So it's interesting to see what's going to happen.

And Steven, as you mentioned, they are now also kind of renting out data center space through their colossus data center. So it's all AI, I guess. I was going to mention, Steven, we talk about this kind of AI and how it's going to hit our pocketbook.

The timing for the industry couldn't be worse than what's going on because we're coming up to these midterm elections and a lot of contention, but contentious talk about affordability. And this plays into that narrative between the haves and the have-nots. So out here, there's no question that these tech bros or the tech billionaires have the ear of the administration and have been incurring favors and been granted special status.

They know where their bread's buttered. So that's going to be playing, we're going to hear a lot more about this. It's not going to be just Erin Brockovich talking about data centers or Bernie Sanders talking about affordability or AOC.

I think it's going to become a bipartisan issue and it's going to be an effective one to run on or run against. And it's going to be interesting to see how the industry, in a sense, kind of navigates between lobbying to make their point and not even further establishing themselves as the black hats. But I think again, with Apple and the price increase, and again, you were very fortunate because on average, the increases were at least 250 bucks on average for the hardware.

We're just going to hear more about this. It's going to deepen, too, because there's no immediate solution in sight in terms of the supply chain issue. No, it's not going to go away, and I think that it's going to hit people.

So Apple did not raise the prices of iPhones, but I think that the smart money is they will raise the prices of the next model of iPhones simply because they have to. Right now, they're obviously subsidizing the costs relative to an iPad, which is basically the same device in a different form factor. I also expect that the same issue is going to hit all the Android phones across the market.

Yes. Mm-hmm. I think that we're going to see things with lower specs, and this is going to cause...

But it's a domino effect because if the phone doesn't have as much memory, then that phone is going to offload AI application processing into the cloud, and somebody's got to pay for that. And I think that it's one of those things where we're going to rapidly see the cost of all of these features that are being promoted coming down on consumers, on businesses, on suppliers, on manufacturers. And everyone is going to have to reckon with the fact that we simply don't have enough supply to meet demand, which means that costs are going to continue to go up, even as geopolitical situations have caused the cost of transportation and manufacturing to go up for other reasons, too.

And so essentially, I think we're looking at a global, I don't want to say recession or depression, but at least a global inflation shock as AI hits everyone. And to your point on politics, I am shocked that we don't yet have a national or even global political leader who is out there anti-AIing left and right in order to get attention and to draw votes from people who are frustrated by all this. Yeah.

So if this had come from any other person besides Tim Cook, I wouldn't give it as much weight. But what he said and what he outlined what's going on carries a lot of influence because this guy arguably was the greatest COO ever, especially his great- And specifically a manufacturing and supply chain COO. Right.

Exactly. And so he's leaving in September, so his timing is good, but I'm wondering, he's leaving around the time they announce their new iPhones, and they really want to play-- They're not playing catch-up with AI, but they really want to establish the iPhone as some sort of AI device. And I wonder how this impacts them in terms of the price increase, whether that undercuts what they're trying to do.

So they're hit As hard as anybody, actually. Yeah, I think also that what I'm seeing on the enterprise side is that because of the price increases, people buy less, right? So they're kind of postponing, which is not necessarily a good answer if the supply chain issue will take a long time.

But the other thing I'm also seeing is that a lot of the vendors who provide infrastructure are talking more and more about optimizing. And so looking at NVIDIA, for example, one of their main inference cards is an RTX card. So last year they released the RTX 6000, and then everybody was expecting a better, faster card this year.

And then they came up with a smaller card, the RTX 4500, which was less powerful but more power efficient, and generated less heat, and so on. So I think also the industry is kind of trying to adapt to the idea that doubling the need for power and memory every year or every 18 months is not something they can keep on doing. But I agree.

At some point, the whole thing is going to fall apart, right? Or it's with new innovative technology or some kind of efforts to kind of reduce the need for higher performance and higher consumption area. One other thing that, maybe I'm going a little bit off the beaten track, but I always think about also the government's influence and things like Anthropic and this rush, this model race, which has turned into an infrastructure race.

And then the government's actions against Anthropic and Mythos in particular, I know they lifted some restrictions only for vetted US companies or enterprises for the use of Mythos, but there's still the ban for consumers of Sable. And I'm wondering what impact that has going forward on revenue. And again, Sam Altman, of all people, came to the defense of Anthropic because he sees what's happening to Anthropic and is happening to OpenAI.

Like if the government gets more and more involved in policy for national security reasons, that impacts the scalability and also the profitability of these companies that want to go public. And I wonder what influence this has, again, on raising costs for them and for the consumer and for enterprises. It all kind of plays, fills into with one another.

These stories to me are as much about geopolitics as they are about basic economics, and unfortunately, these companies have to navigate through that, so it's even more treacherous or a little trickier than they thought it would be. And this goes all the way down to their customers, the consumer, to everyone, this whole price affordability thing, and this race to build out this infrastructure. $650 billion, I think, among the biggest guys are being spent on those data centers this year.

Yeah, it is pretty wild because essentially we're in a situation where there's a race to the bottom in terms of token costs. So the tokenomics story is one of absolutely trimming the cost. At the same time, all of these companies have to figure out how to make money.

And I was listening to an analyst talk about this same question the other day, and talking about how ridiculous it is that we're in a situation where the big American frontier labs are trying to compete on a per-token cost with the Chinese companies, which is just a fool's errand. What they ought to be doing is competing in terms of quality and efficiency in order to try to build better models. Instead of trying to build ever bigger, ever cheaper models, or ever bigger, ever better, ever more expensive models, they should basically be trying to figure out how to optimize the economics here.

And I think that some of them are. We did just see announcements of ASICs to run inferencing. Certainly, that's something that we've heard a lot about from Google and Amazon.

And I think that, was it OpenAI that just talked about their first ASIC as well? Is that right? But they overall are still basically treating tokens as sort of an undifferentiated fungible thing, and that's not necessarily true.

I'm wondering, Frederik, if you have any insight here in terms of how enterprises are looking at this, and how they're going to handle this. Right. So when we talk about token economics, there are actually two sides to it, right?

So you can look at the consumer side, which is basically the amount of tokens that go in the model, and then the amount of tokens that go out of the model. And so there is a cost associated with each token going in, there's a cost associated with each token going out. However, and I think, Steven, you pointed out, the model generates a significant amount of tokens internally, which is not exposed to the consumer and is, to a certain degree, not really exposed financially to the customer.

In other words, if you build a model that does five times, or generates five times more tokens than a lower-end model, you as a consumer, you don't see it. The only price difference you see is by selecting the model. And I think a lot of the efficiencies Will have to come from those models because it's almost like using a sledgehammer to hit a nail, right?

Today, a lot of those large language models do a lot more than what we ask from those models. And I think there has to be some kind of a way where we use a regular hammer as opposed to a sledgehammer. I think maybe that's where we talk about the smaller large language models or maybe better fine-tuned.

And I think economically, the Anthropic and OpenAI will have to go there, right? Delivering those mammoth models, in the hope that it covers all the use cases for people. Those days are over, I think.

We need to have something that is more specific, such that we don't have to keep growing. But token economics, it's kind of weird we use tokens, because it doesn't show the whole picture. But I guess it's the best metric we have today to explain to people and to charge people for their consumption.

Hey, Frederic, can I ask you, I don't want to put you on the spot, and it's probably too early, but given Apple announced the price hikes on June 25th, has there been, and again, this is probably too quick, but has there been any market research that kind of looks at the impact on IT budgets and spending? I've seen some pre-Apple price hike information, but I'm wondering if that's accelerated or maybe even raised a concern, and maybe any type of research that indicates that we're going to see tighter budgets later this year? I think the challenge today, it's because the supply chain happened so quickly, is that the impact on the budgets is where budgets don't grow, but you have to do more with less, so to speak.

And considering that the prices are higher, you actually end up buying a lot less. So I don't think that people are increasing their budgets. I can tell you what the enterprise, or the hardware vendors are doing.

They come up with financial structures such that you can still buy the hardware you want, but then you end up with an OpEx model that is going to cost you a lot more. So in other words, hardware vendors are not trying to solve the problem, they're just trying to allow you to buy the same amount of hardware at a higher expense, which would tell me that the budgets actually will have to go higher next year. " They're more thinking about why don't we optimize and increase the efficiency?

" And then at some point, you have so much hardware that when you look at the efficiency of the hardware, you end up realizing that maybe there is a better way to do this than just buying more hardware. And I think that's also what we're seeing on the consumer side and more on the training and the fine-tuning side of the house, is where people are saying, "Let's optimize. " And look at Nvidia.

Nvidia bought Run:ai, which is a company that helps exactly with that. A lot of the markets today are not necessarily focused on hardware but are focused on optimizing and efficiency. Two weeks ago, I was at Techfield Day AI Infrastructure.

Almost all of them are talking about optimizing and efficiency. And it's an infrastructure session, right? It's people that supposedly should be pushing for hardware.

" Yeah. And let me hop in on that too, because the Futurum analysts have actually talked about these specific questions that you're discussing here. I think it's worth pointing out that the Futurum research side has specifically analyzed, for example, Micron's question of supply.

And as Frederic's pointing out, it's not a question of ramping up supply. They simply cannot ramp up supply. And I've had conversations with Daniel Newman and others about this exact point, that there's a historic challenge here, and that is that RAM and flash especially has been really in a boom and bust super cycle, where essentially every few years, RAM becomes incredibly sustained and there's a bunch of investment to build, to increase supply.

And then you end up with a bust problem where memory is literally selling below cost. Companies go out of business. Investment dries up.

We heard that from the CEO of Micron recently that they did not invest because they couldn't invest when there was a glut of supply. And so now they're facing a glut of demand, and they can't handle it. As Daniel said, there's a super cycle here.

Essentially, there's a multi-decade super cycle of supply and demand fluctuations, and the answer is not to build more, because the industry simply will not. The answer is, as Frederic is saying, to make better use of what we have. And I think ultimately that's going to be what's going to happen in 2027 and 2028, even as, yes, I'm sure that Micron and SK are going to try to build more memory factories to meet this demand, but they're not going to be able to bring any of this online in time.

And so instead, what we're going to have to do is make better use of what we have, which is actually Not that crazy considering that the utilization rate of a lot of this stuff is just incredibly low, especially owned corporate infrastructure as opposed to the neo clouds and the big AI model companies. They are actually making pretty good use of their hardware. But what I'm hearing from the analysts and as well as people like Frederic is that in many cases, there's a lot of equipment out there that has not been deployed or has not been deployed effectively.

Is there enough, Frederic, to meet demand, do you think, or is there enough maybe to stop the bleeding a little bit? I think personally, I think there's enough for reasonable demand. The problem is that we live in a very competitive market, which basically means if customer A, who might be a large bank, is doing something, then all of the competitors want to do the same thing.

So you have a tendency where a lot of capacity and a lot of infrastructure is deployed in order for the me-too game because you don't want to lose out. Right? I think the other problem is that for Micron and others, there is very little incentive to actually build new factories, even if they could, because they realize that at some point the peak will be over and then they're stuck with that, which is even worse because then Micron and others get into a financial problem.

So I think it's a problem we all together have to solve. How we do that is kind of challenging. But I think the reality is that efficiency and coming up with more reasonable AI applications as opposed to generate applications for every little thing we can come up with.

It's strange to say, but I think we're in relatively good shape. I think this is, to a certain degree, it's nice to have the supply chain problem because it forces people to kind of stop buying more hardware and stop the crazy spending. Right?

Because in the end, all of that hardware will be old and deprecated in no time, and what are you going to do with all that gear in your data center? So I think right now it is a problem, but I think it forces people to think differently, and I don't think by itself that's a bad thing. I think it's to think reasonably about AI as opposed to let it grow wildly.

Yeah. I think we're getting a clearer picture of where things stand in a few weeks when the major companies start announcing their results. There was a guy, he's kind of this colorful character, Daniel Ives.

He's everywhere. He's always on CNBC. He's kind of an oracle of Wall Street, and he sent out a note a couple of days ago where he talked about efficiency and optimization in terms of enterprise spending, and where the companies, the big companies that are reporting that are public, are going to kind of give us an idea of where this is now, given the fluctuating and fast-moving parts that are going on in this whole supply chain issue.

But it will be interesting, and again, I'll beat a dead horse here, but it's going to play into this narrative of AI and what it means to you, what it means to your pocketbook beyond your fears or your hopes around the technology. And ultimately, I think that it's a good thing if the price of all this comes home to roost. We've been talking about this literally the whole time we've been talking, utilizing AI, and the industry at large has been talking about this.

I know the analysts have been talking about this. Essentially, all of this build-out has been hidden and disguised and subsidized and financed and leveraged and mortgaged by everybody up until this point. It is actually healthy for the true cost of this to be reflected, or at least to start to be reflected.

And if it costs Apple another $500 to manufacture a MacBook Pro, then maybe they ought to start charging that, because then maybe somebody's going to start asking themselves, "Do I actually need that machine? " It's the same question when it comes to these AI models, as I've heard both of you mention as well. If the real cost of the tokens that I'm burning to, I don't know, make a funny picture or write my book report or whatever, if that real cost was reflected, maybe I would make a different choice.

And I think that ultimately, that's going to benefit us. If the economics of AI start reflecting the reality and the real costs of AI, it's going to benefit us in society because maybe we will start saying it's not just throw AI at it and see what happens. We actually have to start making a case for this, and we have to make sure that what we're getting is a useful and valuable result.

Sorry, Steve. It's kind of refreshing to see this reality check around AI, because for so long it's been anything but a reality check. It's been hyperbole, superficial comments, this overhyped machine, this overhyped idea, and maybe now we're just kind of getting a better idea or definition of where it's going, and it's not going to be the solution to everything, and it also needs to be used with caution and with a certain amount of fiscal responsibility, and that is finally starting to hit home to the companies and the people involved.

You're right. That is a healthy thing. Absolutely.

Well, thank you both for joining us for this episode. We've hit our 30-minute mark here, so we've got to wrap. Before we go, I want to give both of you a chance to give us a little bit of a taste on how we can continue this conversation with you, since you all aren't on this particular podcast every week.

Frederic, where can we connect with you and continue? com. And a special call to our blogs around optimization and efficiency of AI clusters.

If you want to know more about how, why, find us on LinkedIn and read the articles. Yeah. So I'm on Techstrong gang more than usual now.

But I'm going to tell you something, I share something with you. I wrote a book that's coming out in October. The pre-orders have started.

Basically, the whole concept is working with robots, working with physical AI, or working with agentic AI, and how you're going to be able to navigate it or whether you can't, and what the threat is and what the promises is, what the upside is. That's coming out in October, so hopefully, and this is a plug for myself, but I'd like to talk about it on this podcast as it gets closer because I think it's going to hit home really hard in the next couple of months. Well, thank you both for joining us, and thank you everyone for listening to this episode of the "Utilizing AI" podcast.

If you enjoy this discussion, please do subscribe. You can find us on YouTube or in your favorite podcast application. And consider giving us a rating or a review.

We'd love to hear from you. This podcast was brought to you by the analysts and experts from the Futurum Group, where insights meet AI. ai, the "Utilizing AI" YouTube channel, or the Techstrong TV app.

Thanks for listening, and we'll see you next week.