Sharp Tech with Ben Thompson - Your Regularly Scheduled Nvidia Hysteria, Nvidia and the Metaverse, Questions on Sony and Adyen

Episode Date: August 29, 2023

Another round of monster earnings for Nvidia, the challenges and opportunities that will define the next decade for Nvidia and other AI players, and questions on Sony's success and Adyen's hiring practices. At the end: More on nut picker-uppers!

Hello and welcome back to another episode of Sharp Tech. I'm Andrew Sharp and on the other line, Ben Thompson. Ben, how you doing? Not so well. This is our second take because I referenced the fact that Memorial Day was next week. I'm just losing it over here. Yeah, well, you were just complaining about the pressure of having to answer how you doing on every single podcast we record. And ladies and gentlemen, Ben cracked under the pressure tonight.

So this is take two and a half here. It is good to be back at it. Sailing the Sea of Takes here in late August, you know, surveying the landscape. I have two programming notes to get us started here. Ben, are you ready? I am ready. Okay, so programming note number one, this is a free episode of Sharp Tech. So if you're not subscribed to Stratory, now is the time to get on board.

Start receiving daily content from me and Ben and the rest of the Stratory. Techery Universe. And if you're already subscribed to Stratory, stay subscribe to Stratory. If you're in line, stay in line. And programming note number two, Ben, we are recording a second Sharp Tech episode Wednesday night. So if anyone out there has questions for that show, you can send them to email at sharptech. We'd love to hear from you. And Ben, one thing before we get into it today, you famously do not give stock tips either in writing or on podcasts, but I did have one idea I'd like to run by you.

Can you help me out? Sure. Maybe. I'm not actually, no, I'm not committing to anything. Have you ever heard of a company called Nvidia? They're a little bit under the radar. I do think there might be some upside there. Are you familiar with them?

I kind of regret not writing like a definitive can point to, look, this is obviously going to be a great thing. I mean, I wrote an article. It's crazy. It was only 11 months ago called Nvidia in the Valley, basically saying, look, long term, this company is in great shape. They have all these different areas from gaming to AI and something else like, you know, they're going to get through this. And the reason that article needed to be written is they had just gone through this massive crash in their stock. And it was, you know, like a third of the value or something along those lines. And it is wild that that was by one and only opportunity to sort of really explicitly make the case of you should buy in now, you know, sort of like meta style when they were when they were sort of in their doldrums.

And yes. So again, my apologies for not giving a tip implicitly, not explicitly, but yes, it's pretty crazy to see. Yeah, I mean, they were a sensation during the pandemic, then crashed in the valley. And now, I mean, we last talked about Nvidia on June 1st when they made headlines as they were flirting with a trillion dollar market cap. I believe at the time we recorded that episode, they had dipped back below the one trillion mark. Three months later, they are now safely above the trillion threshold, 1.16 trillion as of this recording. and on the heels of their earnings report last week, I'll just set the scene here with some facts from Forbes

and the Wall Street Journal and Reuters. Number one, NVIDIA's stock, already the top performer in the S&P 500 this year, rose 7.5% following last week's earning results, which would be about $87 billion in market value, so not a bad afternoon there. NVIDIA's $13.5 billion in quarterly sales and $2.45 earnings per share smashed consensus analyst estimates of $11.2 billion and $2.8, according to a fact set survey. Thanks to a more than 300% surge in Vindia's share price, its market cap has exploded from less than $300 billion to nearly $1.2 trillion in 10 months.

and Vitya's shares traded at 45 times 12-month earnings estimates as of Wednesday compared with about 19 times for the overall S&P 500. Yeah, I mean, that's to be clear. That's not a super fair comparison point, though. I mean, like the numbers that you just listed make the case for why 45 times is not crazy because you get a higher premium if you have growth. and we are witnessing some of the most insane growth we've ever seen, period. There's a reason why everyone is like casting back to the dot-com era to even find something that's remotely analogous. So this, like, you could definitely make the case that 45 times is relatively cheap. You know, and again, all these things about beating analyst expectations, those are analyst expectations that were dramatically increased and blown.

up the last earnings where where you know so so this is basically two major sort of outperforms in a row and there's a good reason to expect it's going to happen for the next several quarters one of the natures of making these chips they're not just chips they're they're like wholesale computers is there's a long lead time and so invidia invidia is bound right now by supply by their ability to make enough and if your revenue is defined by your supply, you kind of have a pretty good idea that you're going to fulfill it. So they've talked about they have demand that's basically running out into next year. And there's an extent to which supply is increasing.

I think it increased faster than Nvidia expected. TSM right now, everyone else that makes high performance chips doesn't really need the space. So like this is kind of a got sent a TSMC in a certain respect as well. But yeah, I mean, this is something that's been unwrapping for a few quarters now. I mean, with this quarter, is like, are they going to actually meet these sky expectations and then to sort of exceed it again was remarkable? That said, there is a bit where this, you know, this expectation will get priced in pretty soon. And, you know, no company can grow at this rate forever to say the least. Yes, and we'll get there.

One specific question, though, as far as pricing in expectations, Nvidia announced a $25 billion stock buyback in conjunction with the quarterly earnings last week. my understanding is that companies typically do buybacks either when they think their stock is undervalued or when they're making so much money that they don't know what else to do with it, but return it to shareholders. So which is it in this case? Do you have a sense? Yeah.

I mean, I think there's just have so much cash that like what do you actually do with all that. Like there's a bit where you have too much more than cash than you can sort of efficiently deploy. And the reason to do a buyback instead of like a dividend is what number one, it's more tax efficient. and number two, you're not sort of locked into it sort of forever, right? There's a big sort of penalty in unwinding a dividend. And but I think I believe if I recall this authorization was sort of long term, like they don't have to buy all the back stock back with like within a year.

So I would imagine this will be deployed as a sort of prop or, you know, or what the stock does sort of go back down. That's when it will actually sort of be leveraged. But yeah, but there's a bit where, you know, they're almost getting. more cash than they can do anything with it. Again, because they're making physical products. They have to actually have space with TSMC. They have to actually secure all these components that go into these sort of chips.

So there is some sort of gate to how much they can do in sort of any one particular quarter. And, you know, so, yeah, I mean, there's a bit, there's sort of like cycles on cycles because the semiconductor industry generally is sort of very cyclical where you get these big buildouts. You have these big lead times. Again, we know this. This has happened in video literally a year ago where they had this huge overhang. They had all this inventory, you know, Ethereum, which was primarily sort of when it was proof of work, you used Nvidia GPUs for the miners did.

That moved a proof of stake, which basically removed that requirement, flooding the market with all this sort of used hardware. They had this PC sort of explosion and this huge run up in demand during COVID. that sort of hit a cliff and a wall. No one was buying sort of new cards and they had a new card coming out. And so, you know, it's really been a roller coaster just sort of in one year. I mean, six months ago, they weren't just writing off inventory. They were writing off future purchase orders with TSM.

And I wrote multiple articles or brought it up several times sort of, you know, seven, eight months ago. Like, what is going on here? We have it seems like chat GPT, which is really the turning point here, is driving all this demand. So why is Nvidia writing off sort of future purchase orders? And it turns out that they probably weren't. They wrote them off. And then now they sold it anyway. Like they actually raised the price of their previous generation chip, not just the H100, the current one, but the A100 because there's so much demand for it.

And yeah, needless to say they are using the allocation that they bought from TSM and more. Well, it's been wild to live through the take cycle with these guys because a year ago, there were a bunch of gamers upset at Nvidia's most recent products. And there were worries about how that would affect the short-term future of the business and everything else. And now it's like, oh, okay, so these guys are going to be the linchpin of the entire digital economy. moving forward because of how important they are to AI ambitions, at least in the short term. And I'll read from Stretecory. You wrote about them on Monday, and you wrote, the biggest challenge facing Nvidia is the one that is ultimately out of their control.

What does the final market look like? Go back to the dot-com era and the era that preceded it. The advent of computing, first in the form of mainframes and then the PC, digitized information, making it endless, duplicable. Then came the internet, which made the marginal cost of distributing that content go to zero, parentheses with the caveat that most people had very low bandwidth. This was an obvious business opportunity that plenty of startups jumped all over, even as telecon companies took on the bandwidth problem. Cisco was the beneficiary of both. The missing element, though, was demand. Consistent consumer demand for internet applications only started to arrive with the advent of

broadband connections in the 2000s, thanks in part to a buildout that bankrupted said telecom companies. And then it exploded with smartphones a decade later, which made the internet accessible anytime, anywhere. It was demand that made the router business as big as dot com investors thought it might be, although by then Cisco had a host of competitors, including large cloud providers who built and open-sourced their own. So if I'm reading you correctly, Nvidia's depends on whether the timing is right for AI products to actually transform all of society in the short to medium term while they still have these massive advantages over potential competitors. Is that accurate? Yeah, I mean, because this question of is Nvidia Cisco is an interesting

one. So I think there's lots of reasons to say that they're not. They actually do have a more defensible business. There's certain natures of the way the market works today that is better. I mean, for example, when Nvidia sells a GPU to a cloud company, which is over 50% of their sales, that GPU is put to work immediately, right? The cloud company is making money from that GPU from the moment they turn it on. And that wasn't necessarily the case if you had like a startup buying, you know, Sun servers and Cisco routers. Like they had to actually then build the product and get people to use it. And you had all these companies that were buying all this equipment that never ever built viable businesses.

The clouds have viable businesses. At the same time, the people that are using those GPUs from the clouds, they are the ones trying to build quality businesses. So we sort of just move the challenge sort of a bit down the road. And we're in this, you know, so the real question is in the long run. Right now, Chad GPT inspired just this absolute explosion and of this sort of foamow amongst like companies and startups and VCs that like this. We got it. This is clearly the future. We've been in this moment of Stasi as we've kind of the end of the beginning I wrote,

where we have these dominant mobile cloud platforms. We have these dominant sort of cloud providers. Where's the opportunity is going forward? And AI comes along. I was like, there it is. There it is. There is the opportunity. And I think that they are right.

That is the long-term opportunity. There will be very large companies built around this. But there's a question of timing. This goes back to the opening of the podcast, right? You had to get the timing on Nvidia right. And you know, and that you have to get the timing right with investing sort of in general. Very famously, you go back through the most famous dot com blowups. All of those companies exist today.

But they don't exist as those companies. They exist as new companies that were started like the 2010s that were actually delivering pet food, right? That were doing all these sorts of, you know, e-commerce startups, or whatever they might be, that were doing all these crazy ideas that, that in the 90s, people had the right idea. The internet was going to be transformative. But the part that's easy to forget, particularly as technologists where you're sort of living on the cutting edge, as investors where you're looking for the next big thing, is that, you know,

and this ties back into our recent podcast where we talked about the U.S. versus China and this idea of how important demand is. And that's actually the most sustainable aspect of an economy if you want to sort of you know, run the world from the sort of economic perspective is if you have demand. And demand is ultimately what makes or breaks companies, which makes or breaks sort of markets in the long run. And demand is one of those things where it takes a lot longer, I think, to materialize than you might think when you see this sort of amazing technology. And, you know, Chad GPT is kind of an example where on one hand, it's incredible the extent to which it blew up and became a big thing. On the other hand, as we've discussed, GPT3 had been out for two, two and a half years by the time chat GPT watched, right?

Like, it took that long to even get sort of a minimum viable product that surfaced this technology in a way that was compelling to end users. And while we have like ideas and scenarios, we have things like the Microsoft co-pilot things that we've talked about, you have all these enterprises investing in it. When will this actually provide a measurable return to the money that's being spent on it? And in order to do that, do these products need to cross over to normies of the world and transform the way normies work and live? Is that ultimately what we're talking about here? I think there's a multi-step process. So I think that this will happen in enterprise first. Like, because you have an enterprise, you have the real driver of potentially significant cost savings, which is going to motivate sort of action.

And you're going to figure out a way to do customer support via this. Or you're going to figure out a way to sort of surface content in a way that makes your employees more productive. And you're going to measure it. You're going to sort of see what is the ROI on sort of these investments. And all of this is a significant opportunity. It's one that I sort of analogized to just the general build out of the web in the late 90s and 2000s. A company here, what was a big winner of that era? IBM actually was actually a big winner of that era,

where a lot of what they were doing was just providing middleware. They're going to these companies saying, look, you're a big enterprise, you have successful business, you know that the internet is probably important. Let us come in and build out, you know, sort of your website and all these sorts of things. And it's so sort of rudimentary in retrospect, but that was sort of a necessary step that needed to go through. And I think that's going to be the first markets that do make sense for ChachapT. That's why I've been pretty optimistic about Microsoft's positioning, for example. And that, yeah, there's why Open AI just today, as we're recording,

announced their sort of enterprise chat GPG product, which I think makes a lot of sense and is sort of an obvious market opportunity. That said, the enterprise still pales in comparison to the consumer market. Like, that's where the real money is made. That's the real explosion. And that's the real sort of feedback loop where you have things like smartphones that are at such volumes that have such an impact that fill. back into other products and into other opportunities.

And you have an opportunity to create entirely new kinds of companies that exist in this sort of new paradigm with a new device. And that bit definitely takes longer. If you go back and you start the count of the start of the internet in 1991, 993 or whatever you might want to choose, you have the dot com bubble in 1993. You have sort of the emergence of social media in the mid-2000s. It wasn't until 2007 that the iPhone. No, that's why I loved what you wrote because I've been online for basically my entire life,

like 1994, 1995 is when I started to mess around and check like basketball box scores and everything else. But it didn't feel like we were living in the future and the entire economy was being upended until like 2007, 2008. And that's when all these businesses just sort took over the world. And I had never really connected those dots in my mind. And when you think about where we are with AI, we are sort of in those early stages where it's clear that this is going to be something, but what's going to be the thing that brings it all together and creates like a entirely different paradigm that anchors the economy moving forward? And I think that remains an open question. Well, to an extent, though, I mean, I think there's a good chance that a lot of the use cases being put for today for AI will be the use cases that AI is actually good at, right?

We've talked about like the legal thing, for example, right? Also true of early internet. There are all these failed businesses. Exactly. Yeah. Exactly. Like, and sort of the point of this article was like there is just an inherent timing of this of aspect where even if it's super obvious to you and I or to the investor and the startup that look, This is obviously better.

It's the next thing. Yeah. It takes a lot longer than you realize to actually meaningfully change consumer behavior. And sometimes it does take something like a device shift or a paradigm shift to really accelerate that. I mean, even if the 2000s felt like a lot happened, the 2010 is what was really insane as far as growth happens, you know, growth in tech or whatever. I think someone, I saw something on Twitter the other day about basically the free cash flow of the big top. five internet companies in the 2010s was equal to or surpassed their collective valuation at the beginning of the 2010s, which just goes to show that the valuations were drastically too

low, sort of way back then, right? I mean, honestly, that's always been the real strategy alpha is I was so big on the big internet companies and saying, look, you don't understand how essential and important and insanely profitable these companies are going to be sort of in the early days of trajectory. And the problem is I had no money to invest back then. I had to sort of making the first place. But if you really wanted to make money, if you just went, you know, back then in the early 2010s, if you just bought the big five internet companies, you're probably retired today. Well, listen, Ben, I'm telling you, I heard about this company called Nvidia. It could be a smart play for all of us here. You've got more resources today.

Well, and that's, you know, and again, Nvidia is also different. They have real existing businesses like the gaming business and like the auto business and all those sorts of things. They do have real moats in terms of just the quality of their, chip set, the software ecosystem, which didn't really exist for routers. At the end of the day, a router is a router. It's not to say it's not hard to build. It's not complicated, but there's a reason why Cisco got a lot of competitors. I think like Jupiter Network sort of came along, political networks, and then you had companies like Facebook that, I mean, Google was the first sort of build all their own and sort of have software-defined sort of router. So there's just using

generic PC parts and then building their own software. Facebook did that, but open-sourced it all. And then everyone sort of standardized and sort of that sort of thing. And, And the reason to bring that up is that is the shape of competition for Nvidia, sort of in the long run. That benefit Nvidia has of primarily selling to cloud providers who are putting these GPUs to work immediately and thus have a high incentive to keep buying more and try to secure market share. That is a long run a bit of a detriment because those cloud providers are heavily incentivized to find an alternative to Nvidia, to find something that's cheaper, to find, you know, Amazon is working. on their own chips. That'll cut into the margins. Right.

And the real, you know, this is nothing new, but like the real long term market here is for inference, the actual running of the finished models, right? So even if Nvidia dominates training, and that's, sir, going to be a very lucrative business, the big money. When this stuff is used widely, will be an inference. There's going to be a heavy motivation to solve that. There'll be a heavy motivation to build up a software ecosystem that's independent of Kuda. And that will be sort of a challenge for them in the long run.

for sure. Yeah, well, a couple hosting notes here. Number one, you unleashed a groundbreaking pronunciation of the word stasis earlier, as you were describing the tech industry a couple of years ago. Stasi or something. Yeah, it was like a silent S, perhaps a nod to the Stasi of East Germany years ago. Also, we're not going to dive into the Cisco parallel because we discussed that a lot on the June first episode, but you wrote about that on Stratory. We'll link the June 1st discussion as well. Cisco is relevant because they were another company that was providing essentially infrastructure to the entire digital economy for a while there.

And the stock exploded. And then there was a crash. And Cisco was never really the same. Right. I mean, say that with Microsoft and Intel. I mean, but, you know, Microsoft, it took them, I think, 17 years to reach their dot com heights, like 2017 or whatever. when Nadella took over and, you know,

or a couple of years into that, this is why investing is hard, right? Like all that whole time, Cisco and Microsoft were all growing revenue. They were all very profitable and their stock was totally flat. Because like there's that sort of the challenge. But there's also, there's a real benefit to this.

I mentioned in passing those sort of telecom companies that went bankrupt, right? Building out all that fiber, it was called dark fiber. So it was like all this stuff that was laid out and then no one used was super important for the internet's development broadly. Like Google famously bought up a lot of it and could provide this sort of service, which was a huge jump start to the internet generally when you have this aggregator that could help you find anything on the internet.

And their service was better in part because they got all this fiber at super cheap rates that went to other data centers together and provided a better product. Right. And broadband just generally being built out was super important to this. And this is a feature of these technological cycles. I mean, you know, Colorado Perez has written a lot about this where you have these times of frenzy when all this stuff is sort of being built out and you overbuilt. And with GPUs, you can imagine a future where what's the what's the limitation I talked about? Infference is expensive.

Actually running applications costs money. Well, a good way to run applications more cheaply is when you have a bunch of GPUs that were already paid for and all the companies that bought them went bankrupt and someone's going to use them. Right. Now, it's not quite the same because there is a real pay. out from faster GPUs and more performance, right? We're by no means at the limit like we are, arguably with processors, where does your computer, for most people, really need a faster processor? Maybe not.

Like, right now, we could definitely use faster GPUs. But there's a bit where this frenzy is a good thing, even if sometime in the process, there's probably going to be a lot of people that do get burned. Right. Yeah. Well, and as far as the frenzy is concerned, you mentioned. in chat GPT earlier as the perfect example of what we're talking about with questions about timing and changing consumer behavior. I thought you were going to say it's the perfect example

because it's this amazing product that most people don't have a reason to use on a regular basis. And when you think back to the transformation in 2007 and in the five to 10 years that followed as the smartphone took over the world, like the real transformation there was suddenly my mom was doing all her shopping online. And normal people who didn't really care about the revolution that was happening in the digital economy, everything became so convenient so quickly that the whole world just sort of shifted their behavior. And I am really curious to see if that's possible in the next five to 10 years with AI or whether it will take longer for regular consumers of the world to, look around at these products. Like, people keep talking about the enterprise and how they're going to

use these large language models. I can imagine enterprise is sinking money into large language models and then having just like the internal chatbot sit there mostly unused for the first several years because nobody really cares and it slightly improves efficiency. But the Microsoft $30 receipt billing will work fine. They'll be cashed those checks, absolutely. But it's just in terms of the paradigm shift that's required or would be necessary for some of these more bombastic predictions about the future to actually come to fruition, it may take longer. This is where it's a big opportunity to be a technologist and to be willing to explore this sort of stuff. I mean, the one area where generative AI is definitely having a big impact is in software development, where you have things like, GitHub co-pilot that it is just helping you.

Like there is a lot of sort of scaffolding and busy work that goes into writing code. If that can be sort of automated away now, that doesn't remove the sort of need to have a broader architectural vision and understand what you're trying to do and have the logic worked on your head. But a big part of it is you build this structure in your mind, then you have to put it on to the screen in terms of text. And so it's actually a good example of how, of how. how AI probably makes sense, where it's an augmentation, taking care of busy stuff and making

you sort of more, more productive. But you go back to the pre-in-net eras, all the geeks and engineers were online early too, and they were in forums asking how would I do this sort of thing and sharing information and sort of like getting a leg up, relatively speaking, by virtue of that. And so, you know, if you're, there's an aspect, if you're sort of a young person, it is a big opportunity. And as much as we are annoyed by and hate the stupid AI Xers that are like putting these unbelievably irritating threads, which by the way, seem to have gotten away a bit. I think everyone's complaint sort of broke through and they've been sort of devaluing the algorithm. Yeah. First of all, I can't believe you just called them Xers. I was confused

what you were referring to. Now I understand heavy former Twitter users, now X users. the algorithm is serving me these confident predictions about how chat GPT is going to transform every single sector of society and all this crazy shit and my eyes can't help but roll. I'm a disagreeable personality.

But yes, I'm open to the possibility that ultimately they will be right. What interested me about what you wrote on Monday for Sertechri is it seemed like you and Jensen Huang had sort of divergent, ideas about how this could work well for

Nvidia in the short to medium term. He was talking about accelerated computing and you were talking about the Metaverse. Can you flesh out both sides of that spectrum? I want to be careful because I don't want to mischaracterize Jensen's point. I mean, obviously he's got a phenomenal track record and is much smarter than me, so I don't want to...

Seems like he's doing okay lately. Yeah, I think he's managing fine. But he talks a lot about this. idea of accelerated computing. And it's sort of general thesis is that, you know, we've sort of hit a wall as far as traditional computing. He talks about the end of Moore's law all the time. And there, you know, we have hit the end of Moore's law, particularly economically, where it used to be the point that getting a smaller, faster transistor was also cheaper. And that hasn't been the case for a while now. It's more expensive. And now it's getting harder to even get smaller and faster. And so

if we're not at the end of Moore's law, sort of technically, we are there economically. And, And we can see the end state from a sort of technical perspective. And so his argument is, look, we make these insanely expensive servers. Like what they're selling costs tens of thousands of dollars, which they're selling as much as they can make of, right? But his argument is the amount of performance you get per dollar you spend. And not just dollar you spend, but the amount of space in your data center, the amount of energy consume. Your performance per what is so much bigger with our computing that, It's inevitable that this computing sort of displaces traditional, you know, data center spend,

which goes to regular GPUs, you know, from Intel, from AMD, increasingly from arm-based GPUs that are made by TSM. And there's this, and he talks about this sort of replacement effect. So basically, you're just spending more efficiently in that scenario. Right. And now, maybe he's talking about future spend, but there is an aspect. I'm a little more skeptical of this idea of there being. this general accelerated computing opportunity that is replacing traditional computing.

My view is there are big GPU opportunities with generative AI being the biggest one right now. And there are things like machine learning algorithms and recommendation engines that are additive to the computing we do today. But how much is sort of replacing? Now, there's a little bit of A, a little bit of B. There is an aspect of where a lot of sort of machine learning algorithms and ranking algorithms were run on CPUs. and particularly the inference sort of bit,

and in part because it was cheaper and it's sort of more efficient to do that. And I do think he has an argument in some areas where running this on GPUs, which costs more, will be worth it because the performance gains will be so astronomical. But I am more skeptical of there being

a significant substitute effect over what works, in part because humans are lazy. They don't want to rebuild all this stuff. And developing for GPUs is still much more difficult than developing for a CPU, With a CPU, it's highly abstracted. You sort of write that super structure in your head, and then the compiler and the computer takes care of the rest.

That's the point of it being a sort of a general purpose computer. These GPUs, you have to be much more cognizant of the sort of the structure of your program because you have to paralyze it. You have to do lots of things at the same time to realize these sort of effects. There are some jobs that lend itself to that, but there are other jobs where you could do it faster, but you're going to have to really think about it and work. it out. And I'm just a little more skeptical of that being a big deal. And this matters for the question of will they sort of hit a cliff at some point. Now, to be clear, for the next, they can see

the demand build out. It will take a while for this whole AI wave to figure out what works, what doesn't. They're going to make a lot of money probably for the next at least two years or so. But at some point, if he's right and disaccelerated computing is so much better, it starts replacing a huge number of traditional CPU workloads, that maybe there is no cliff. Maybe it's invidia to the moon forever. I'm a little more skeptical of that. I think there will be a cliff

sort of at some point, in part because the key to this deployment and leveraging these GPUs will be new use cases. And as we just discussed, it's the new use cases that take longer to emerge and more like,

more importantly, take longer to spread insufficient volume that it makes sort of a big difference. Okay, so new use cases, the Metaverse. Are we talking VR headsets as being sort of the iPhone, if we keep the analogy going from the early 2000s here?

Well, to be good, I didn't say VR headsets, although I do think that is probably the most likely- That's what I read into your article there. Yeah, and it's hard to sort of talk about. But I think there's this trend that we've talked about in the context of Instagram, for example, and TikTok, where you started with this idea of just sort of, you know, getting content from your friends and family.

And then it was sort of more sorted and it was algorithmic in nature. And that changed the nature of the sort of content that succeeded and was popular. You had the rise of the influencer on Instagram, for example. And then you have TikTok come along. And it's like, why are we limiting ourselves to content from your network? If we can actually have the best, most compelling, most addictive content in the world, if we truly harvest millions and millions of people uploading videos across our entire network. And the step beyond that is if content is compelling, why do humans have to make it?

Why can't AI be making it? So there's an aspect where this is where the metaverse definition gets fuzzy, but a future sort of TikTok-like app that is just surfacing content that's compelling to you, which may or may not be made by humans, it's gotten so good that you're not even sure if it's an AI or you can't tell. And that's kind of like a metaverse reality, right? You're in your own virtual reality that does not exist for anyone else. It's completely customized to you. And of course, that could be a headset. It's truly immersive. It could just be on your phone.

And, you know, this sounds very sort of dystopian and sort of distasteful for sure. Elscape where Nvidia actually does pay off in the long run. But it's, if you look at the general track of everything of video games, of social networks of all these sort of trends, that definitely seems to be where we're going. And it's kind of fascinating because if the internet ended up being the, you know, culminating with all of us being in the same place, getting mad at each other on Twitter, the end state of this sort of AI revolution may be this retreat to all these completely customized virtual experiences. And, you know, again, it's easy to see the downside of that. On the other hand,

maybe like, we will be the ones who will be prescient as far as our touch grass movement. That is the stock that will truly be going to the moon. Oh, there you go. Well, I think that's a nice natural endpoint for the conversation. I'm enjoying our quarterly check-ins as the entire world freaks out at each Nvidia earnings report here. But I thought that was an interesting sort of divergence between you and Huang. And I also am hoping that the touchgrass movement can continue to thrive in the decades to come

as this simulated reality as we move from like echo chambers on Twitter to every human on earth has their own little version of the world around them based on whatever AI is generating for them. Yeah, it's not great. I mean, you definitely see the problems with it. But it's funny because, I mean, I'm reminded of the, the sort of biblical story of the Tower of Babel, which I think there's versions of this and like lots of sort of traditions and religions and things like that.

but where you had this sort of everyone uniting, because, you know, to build this, this monument that was like an affront to God or whatever might be. And so he struck it down and made them all speak different languages. And this is like, whatever. Like, but there's, it's fascinating to think about the internet sort of following that path, where we, we had this sort of common, everyone's on Facebook and everyone's on Twitter. And you had this every, this idea that was dystopian in its own. Right. We've talked about this bit where we're not really good at humans of functioning in a world where we have exposure to everyone and everything. There was actually a lot of good stuff that came from having sort of day-to-day friction and just being focused. Like there was this shift that happened with TV even earlier where it used to be the focus of your life was your local environment and the people around you. And occasionally the broader world would break into that consciousness because there was a world war or whatever it might be. Right. And we've shifted.

to this world where everyone every day is obsessed with and aware of what's happening nationally and internationally. And they're following whoever is the president or whatever these might be. And then what's happening is they've lost sight of what's around them. What actually has a meaningful impact on their day-to-day happiness. Sometimes I see people who are really stressed out about the state of the world and I want to grab them by the shoulders and shake them and say, pay attention less. Your life's pretty good. Less news. Everything is okay. This is sort of the touch grass sort of idea, right?

This idea that, look, being with people is actually pretty great. And anyone that's listening to this podcast who has the luxury, I guess this is a free version of the podcast, but by and large, you have the luxury of paying money to read content online. All things consider in the grand sort of scope of human history or life's pretty great. And it's kind of a shame that the way that our media environment has evolved, and the internet sort of impact is to have people obsessed over everything that's terrible when the reality is there's always been terrible stuff in the world just wasn't in your face every day. And,

and you know, there's a bit where it's easy to say right, you know, you always think right now is good in the future sort of dystopian or that's a general sort of thing. But this idea of everyone in their virtual role, of course, that's bad, that's terrible. And yes, it's easy to imagine how that manifests sort of terribly. But there is a bit where, you know, seeking out human connections. which is a natural sort of human way to operate was the internet provided this pale imitation of connection. That was not real and actually manifested the worst aspects of humanity

in a way that was incredibly addicting and just made everyone miserable when everything was pretty great. And it's just fascinating that this whole AI movement, to the extent I'm right, where the end state is increased virtual experiences, where you have these things generated on command, it's kind of dissolving this commonality of the internet. And while that sounds bad at first blush,

it's not quite clear that commonality of the internet was a total win. It hasn't been great. Yeah, well, one final thought. A couple weeks ago, I had my first cigar night in like two months. I texted you afterwards and said it was great to get back. The cigar group is growing our satellite cigar night here in Washington, D.C. We have to start like the new like,

Wyans Club or Kwanis or whatever it is. It's like the touch grass sort of movement. To be clear, our cigar night, not everyone smokes cigars. You don't need to smoke cigars. You don't have to. That's true. It's the concept that matters, right? It's the organizing.

A good qualifier. Well, I was touching grass. I came home that night. It was a Saturday night and I discovered that my wife had been watching Instagram reels for, according to her, an hour and a half straight. And she was cackling to herself and having a great time.

That's the irony of like the TikTok Instagram thing. You come out of it with certainly guilt. But there's an aspect of you actually had a really great time, right? Exactly. So this isn't prescriptive. Whatever works. Live your best life. No judgment here from either host on Sharp Tech.

What was she not doing? She was not on Twitter in the middle of political arguments getting angry. That's true. Yeah, exactly. Yeah. Her algorithm vastly superior to mine on all social. media networks. All right, well, let's run through a couple different mailbag questions here. Chris says, Sony has been mentioned several times in different episodes covering different themes,

but never really had its own segment. In music and film, it has avoided the streaming wars. In imaging, it leads in mirrorless and smartphone cameras and now supplies the Apple Vision Pro, and the PS5 has won the console war. Is there a bigger story behind how this once-struggling Japanese conglomerate has managed to avoid the strategic mistakes of its competitors. And then Aaron says, Ben always praises Sony for simply doing the smart thing and selling its content to the highest bidder. However, is it time to stop ignoring the $20 billion elephant in the room? That's right. Anime. By all accounts, the Sony-owned streaming platform Crunchyroll has been doing gangbusters and is poised to be responsible for 36% of Sony pictures profits in a few years.

So, Ben, we don't have to go into too much detail here, but we received these emails in succession a few weeks ago. And I'm just curious, do you have any big picture thoughts on Sony? I do feel like they're sort of overlooked in the conversation about the most successful companies on earth. I could just be projecting on that front. I had never heard of Crunchy Roll until this email from Aaron. So what do you think of the big picture there? I had heard of Crunchy Roll. I have not sort of an anime watcher. And I was not aware of its sort of financial success, but there is an aspect where it makes sense. What is it? It's sort of a niche product that if you care, you're definitely going to pay. Yeah. Yeah. And that's the way to success sort of

on the internet. If you're not going to be a big scale sort of aggregator, that if you're going to succeed by going direct to consumers, you have to have a super well-defined value proposition. You have to have cost that align with that and all those sorts of things. And it sounds like that they have it. Again, with the caveat that it's not a business I've dug into sort of deeply, it is in line with a lot of discipline that Sony has shown across their businesses broadly. What they've done with Sony pictures of basically they're the most, you know, they're the most attractive partner at the ball because everyone else is trying to build their own streaming service, which is broad-based, this is general entertainment, is not defined like Crunchy Roll.

And so Sony's there like, hey, we got great content. We came out on a movie. We had a TV show and it's up for sale. And what they, you know, they have made money from that. More importantly, they have not wasted astronomical amounts of money. I tried to sort of do something different. And that's made a lot of sense. And they've done well in that regard.

PlayStation, dominant in its category. right? Like, like, and it has an ecosystem component where they don't just make money from, you know, they make money from the games, anyone who sort of contributes to it. And they were very early, as I've written about in the context of consoles, building up sort of understanding how the nature of the competition was changing. They sort of really tripped up with the PlayStation 3 by making it too complicated. Because back then, you won by having, in some respects, hardware differentiation was important. It quickly realized it was not important. what happened was developers wrote the same game for the PS3 and the Xbox 360, and they dumped it down, relatively speaking, on the PS3 because cross-compatibility or cross-platform capability was more important because what was happening in gaming was the cost were shifting to asset creation. That was outweighing the cost of game engine development, which, by the way, you would just buy from Epic or whatever might be sort of anyway.

And so Sony realized, look, the way to differentiate or constantly going forward, you can't differentiate via hardware, the way you will differentiate is through exclusive content, but the incentives of content makers are to go, is to sell to both sides. So we have to own the content makers that will drive the business generally and we'll make up the opportunity cost in not selling on Xbox by gaining license fees from companies like Activision that are still on both, but more of them will be on PS5. super smart strategy, very duplicitous in their objection

to the Activision acquisition where they're spinning up all these sort of scenarios when the reality is there's, you know, there's a reason like they don't want Microsoft go on the same path they did, and they don't want to build a subscription service because that's going to kill sort of their, the golden egg in that regard.

And then, yeah, components. They've been dominant components, particularly in camera sensors. Now, this is a tough business. When you're in components, it kind of goes back, and forth from profitable to not profitable, but they have stayed on sort of the technological leading edge in that regard. And all things considered compared to where Sony was 20 years ago, where, you know, he was looking pretty rough and they couldn't make a transition really to sort

of, you know, to digital. They've never been very good at software. They've actually done a pretty good job of still not necessarily being that good at software, but understanding the dynamics of the markets and positioning themselves well to sort of take advantage of it. So I think it's been, I think it's all around been, been a pretty impressive sort of 20-year run. Yeah. Well, and that's the aspect of it that I find most impressive is I think a lot of people, when I was growing up, Sony's influence appeared to be receding. And now looking up here at the various categories that they're dominating that Chris mentions, it's just really impressive that they've been able to adapt and thrive in so many different areas. And I feel like it's a story that's not told as much.

given how successful they are in so many different little worlds there. So good summary from you. Any final thoughts? I mean, I remember it is interesting because Sony was such a consumer sort of touchpoint and brand. And there's bits and pieces of that that do persist. But it is, you know, they're not really that anymore. They're a component maker.

And they're sort of, you know, if there's a big Spider-Man movie, you don't think about the fact that it's Sony. Right. If you think, you know, the PlayStation is in many respects its own brain. right? And it's been impressive the way they've managed that transition. Well, and Apple launches the Vision Pro and you don't think that, okay, so there's going to be Sony components that make this possible and Sony's going to be part of the success story. Every iPhone, every iPhone has a Sony camera sensor. Right.

It's a very good business. Yep. Okay. So Otto says there was a line in last Monday's strategy article that stood out to me. It was from Audien CEO Peter Venner. underdose, and he said, quote, we are a global company and we're hiring where our merchants are. So we're also hiring more in markets, which are often more expensive than the Netherlands. Thinking about Ben's past comments that remote work in the U.S. could quickly lead to cheaper

remote work internationally, isn't this the perfect opportunity for remote EU engineers to develop U.S. products? As you said, U.S. payments are simpler, so it's hard to see why EU employees employees couldn't solve the U.S. market and turn the higher gross margins into higher net margins. If there's a reason the U.S. market is more efficiently solved by U.S. employees, does that affect your view on remote employee tradeoffs for domestic versus international hires? What do you think, Ben? Yeah, there's a lot of debate on Twitter about this sort of line sort of in the article. You know, one of them being like, look, Stripe is VC funded.

They've been super extravagant. and, you know, a D& is, it's ownership, right? They actually have control. But there's actually, there are real, and that, there's a little bit of that that is true, but there are real structural differences in sort of employment. There's a bit, one bit is if you work for a Silicon Valley startup, you have an expectation of ownership in the company, right? And versus you're just going to be an employee.

And, you know, and that's one aspect that if you actually cost through the cost of employee, that's a pretty significant one. On the flip side, there's a reason why you give employees ownership of the company because of a certain, you know, what you're trying to inspire and activities you're trying to solve. You get better employees too. Right. Will this matter in payments? We'll see. I mean, there's a, I think there's a certain segment that views payments as inherently sort of commoditized. And in which case, Adian's real just a superior sort of attention. to costs and whatever structural advantages they have is a meaningful differentiator will give them sort of an advantage in the long run. That is a very legitimate point of view. The other sort of

point of view is that, look, because it's a commodity, if you're actually going to make money in the long run, you have to build differentiated services that companies are willing to pay for above and beyond. You're going to have to give away the payment processing, you know, basically for free or at very, very low margins. And creating entirely new products that companies are willing to pay for takes innovation, takes sort of employees that are invested, takes sort of a lot of R&D. And so in this case, a sort of striped sort of model where, yes, they have a payment provision, but they have a much broader range of products than a D& does, is actually going to win out in the long run.

It's unclear. It's going to be sort of very interesting to observe what matters in that. case and what doesn't. But there is a bit, you know, as far as the sort of international versus, I mean, there is some question, you know, engineers theoretically could be anywhere. There is an aspect of understanding the U.S. market. Certainly when it comes to sales, you have to sort of be, you have to be closer. There's one of the big things that Aden is focused on is being this sort of platform opportunity. Like an example here is Stripe undergurge Shopify payments. And, you know, and this idea is Shopify is this huge payment business that at the end of the day is built on

Stripe. And so Stripe, that's a very low margin business for Stripe, but they don't have to do any work to an extent, right? Shopify is the one acquiring the customers is doing the sort of go-to-market activity. And Stripe is sort of a long for the ride to certain extent. Now, that's overstating it. Of course, they do a lot of work to support it. But it's a different business than signing up merchants directly.

And to the extent, Adian wants to focus on those kind of businesses, which they've talked about is their big focus in the U.S. market. it's they have to be close to the companies that they are integrating with. They need people on the ground and sort of interacting. Yeah. And so that's something, you know, that's harder to do remotely. And it's harder to do in a different cultural context and a different time zone and maybe different work habits or whatever it might be.

And so this is going to be something I'm going to keep an eye on for sure. I think it's a very, very interesting sort of fight that is happening. You still have the big incumbent players like Chase that are, you know, that that have products that have evolved in response to competition. There are going to be businesses that just care about having the lowest cost provider possible. They don't want any of this other stuff. But yeah, but this question of, you know, the bigger question is initially U.S. versus European salaries. It's this idea of do you win in a commodity market by having the lowest cost structure,

or do you win in a commodity market by actually building in differentiation that lets you capture more margin than you would normally? That's kind of the big picture of debate. Right. And customizing for the big ticket customers would be the roadmap, right? Yeah. I mean, well, Deere had this big advantage in the U, which was, I mean, I don't want to say, criticize it. They seize this opportunity in the U, which is a bunch of different countries, a bunch of different currencies, a bunch of different regulations. And if you're a multinational like H&M wanting to have multiple stores in multiple countries with multiple rules, you don't, you can't afford to get the lowest cost option in. every single little country. Like your coordination costs are going to massively outweigh whatever percentage points

you gain. And so a D& could charge a little bit more, but sort of be one common API that you work around and build on, which is a huge sort of payoff. The U.S. is different because it's all dollars and it's all mostly the same thing. Yes, there are things like local sales tax and stuff along those lines, but the U.S. market is much simpler. The U.S.

we pay with credit cards, or that's pretty much it, right? In the European market, you have these bank schemes, you have debt. debit cards, you have credit cards, you have like just the, the, the complexity. Different data rules and stuff, I'm sure. Yeah, the, the complexity in the EU is a lot higher in a D and solved that complexity. I think the big question for them entering the U.S. market is there, it's actually a simple, the user references, it's actually a simpler market. And so their primary means of differentiation, the way in which they achieved higher margins

doesn't exist. And so either they need to build. new kinds of differentiation, which is expensive and they're behind on relative to, say, a stripe or something like that, or they need to just be a low-cost provider and say, look, we have a lot of infrastructure out there and we'll beat you on price. I actually think that's what they should do. I think they should be willing to sacrifice their margins and leverage, like, basically make their profits in the EU and make huge volume in the U.S.

by using the same common platform and just charging very little for it because those costs are already sort of leverageable and depreciated isn't the right word, but they've already been spent to build up the sort of European platform. But they kind of have it in their head that, no, we are always a relatively high margin provider. And I think that's going to be a tough road for them in the U.S. specifically. But again, there are very strong opinions about this space. People definitely disagree with me.

But it's going to be very interesting to watch and see how it plays out. Very interesting. There was more meat on that bone than I was expecting. And we got a little take at the end there on what a D-in should actually do. All right, well, we've come too far now. We're not going to get to the question about the Ben Thompson Cinematic Universe and what you learned from business school. But we will come back to that later in the week. That'll be a paid episode.

And then a couple follow-ups that I wanted to hit from last week's show. Chris says, big fan of the show. My college friend just moved to Minnesota. He's a lifelong New Yorker. And so has been feeding our group chat, his first follies into managing a lot of lawn. Anyways, two weeks ago, he complained, quote, I am completely outmatched by the number of acorns in this yard. And then last week, he sent us this photo of his new weasel nut gatherer to tackle his acorn problem. And this week, I hear about Ben's chestnut picker-upper. I didn't know nuts were such a

problem. And thank you to Chris for attaching a picture of his buddy with the weasel-nut gatherer. I put it in the rundown for you, Ben. Just an incredible look of satisfaction on this guy's face. That's what problem solving as a homeowner looks like. And it really delighted me last week. Yeah. And for the record, my issue is walnuts. The topic came up from Parker Thompson's tweet, which was chestnuts. But the nuts are a problem. That's basically, that's basically the big takeaway here. The point is that nuts are a problem throughout the American market. And the American market has responded, you know, consumers demanded nut picker uppers. And from coast to coast, there are nut picker uppers out there. Shocked me, I would share the picture of this satisfied homeowner.

I don't know if we have permission from this guy in Minnesota, but what a delight. Thank you, Chris, for writing in a lot of chestnut feedback. The second chestnut email that we have to read, Peter says, Dear Sharp Tech, by day, I'm a biomedical engineer, but my real passion in life is operating my permaculture farm where our two biggest tree crops are chestnuts and pawpaws. Needless to say, I was ecstatic to hear the multiple mentions of chestnuts and chestnut rollers on Monday's podcast. I delight in all things that get chestnuts out into the universe. I name all of our trees, usually after people. people in my life or various influences. Anyways, there are still couple of still nameless trees from this year's plantings,

and I've been a fan of the podcast for a while, so I figured why not? And named one of the seedlings after Sharp Tech. A picture is attached. This one we are going to share, so if you look at the podcast now, I think it'll show up for you. Look at your podcast player.

And Peter says, it is small now, 10 to 12 inches. They spend the first year mostly growing roots, but it will grow into a mighty tree that will bear nuts and help the local ecology for several centuries, literally. Ben, your thoughts? I mean, it sounds like we will be literally touching grass for several centuries. If that's not a real victory, what can I say? What an honor, a mighty tree bearing the Sharp Tech podcast name,

a 10 to 12-inch little shrub right now, but I can't wait to watch this tree just blossom through the years here. He gave us his location. It's pretty close by. It's pretty close to you. You might have to go get a fixture next to it. My mother-in-law lives in the same town and we make it out there three or four times per year. So I honestly, I may take you up on the offer to come visit and see the Sharp Tech chestnut tree in person, Peter.

Thank you so much. truly no greater honor than having actual infrastructure named after our digital product here. It's all part of our global takeover plan. Us and Nvidia just slowly is circling one another as we fight for the control of the future here. Nvidia nuts. That's what we're on about. Exactly. The whole world's nuts about Nvidia these days.

All right. Well, after I sort of mangled the attempt to tie everything. together at the end here. I think on that note, we should come back. Later in the week, we'll keep it rolling. Email at sharptech.fm. Anyone out there who wants to name some infrastructure after us, we're all, well, I don't know if I want to commit to naming certain.

Yeah, let's be careful. The chestnut tree was good. Let's not, let's be careful about taking this too far. Yeah, it's nice and harmless. But we would love to hear from anybody with chestnut takes, nut gatherer takes or anything in between. And Ben, I will talk to you in a couple days. Sounds good.

I'll talk to you later.