저자: BrettWinton · 외국인
요약 보기
ARK의 Brett Winton은 AI 학습비용 급감과 연산능력 폭증으로 AI 전환이 iPhone급 이상의 기술 도약이라고 주장. 구글 검색 광고 모델은 AI 답변 엔진으로 인해 붕괴/압박받을 것이며, 지식노동 생산성 향상으로 2030년 글로벌 GDP 170조달러 이상 전망. 인플레이션은 기술 디플레이션으로 장기 해소되며, 비트코인은 M2 대비 경쟁 자산으로 장기 성장 여지가 있다고 봄.
다듬은 글
Hello everyone, welcome back to another episode of the Meet Kevin show. We're back for the second time with Brett Winton from Arc Invest. Thank you so much for coming back. Well, thanks for coming here, but also for returning to the program. Last time we did it via Streamyard and now we're meeting in person in Santa Monica.
Thank you so much for inviting me here. No, thank you for coming. Thank you for having me. So, we wanted to talk about many things, but I think what really got us interested in talking again was GPT chat. This is what's driving everyone crazy right now . People are hacking into GPT chat and saying it has a left-wing bias.
You have Bing, which apparently suffers from emotional depression because its language learning algorithm is slightly different from that of GPT chat. What's your opinion on this? Is it just another fad like the metaverse, or is it truly a moment of transition? I think it's a real turning point in artificial intelligence.
Um, we think the cost of training artificial intelligence systems is falling 3 times per year and will continue to do so over the course of the decade. And then you combine that with Google, Microsoft, and many startups, and everyone realizing, along with many large companies, that: "Hey, we can invest money in this and generate remarkable capabilities." And, um, you get things like GPT chat proving to everyone that these systems are massively effective, and you end up with a massive surge in capabilities.
So, if you look from 2021 to 2022, we think the aggregate capacity of AI systems increased tenfold this year compared to 2022. So, GPT chat was launched at the end of 2022. Uh, we think there will be another sevenfold increase in the aggregate capacity of AI systems this year, in 2023. And that will continue, you know, throughout this business cycle every year, in our opinion.
And what that means is that by 2025, whatever the state of the world in AI is today, you'll have an improvement of thousands of times by 2025. And then, um, or 2026. And one way to think about that is, today's iPhone, the one you probably have in your pocket, is, on a CPU basis, a couple of hundred times better than the original iPhone.
On a GPU basis, a couple of thousand times. So imagine going from the original iPhone to today's iPhone, but doing it in two and a half or three years . Oh, well, because that's 2007, that's 16 years ago. And you're saying that 16 years of technological advancement could happen in 3 years? Exact. With AI.
If you think about the original iPhone and how efficient it was, that's ChatGPT. Now, if you skip ahead 3 years, the equivalent advancement would be going from that original iPhone, which is ChatGPT, to an iPhone with a magic camera that takes perfect photos, has an App Store, and everything else. The iPhone 1 is strange because it's already 16 years old.
Many people don't even know that the iPhone 1 didn't have an App Store. TRUE? That didn't happen until the iPhone 3G, which was the second one. So, are you saying that an iPhone without the App Store is basically ChatGPT 1 and that in 3 years we'll be on the equivalent of the iPhone 14 Pro Max? Yes.
Exactly. Exactly. I think I use that comparison because it's very difficult to conceptualize what a thousand-times improvement feels like. I believe it will manifest itself in the fact that every company will have to adapt to AI or die. My God. Oh. Today we learned that Google's CEO resigned. We have seen their advertising revenue decline.
YouTube, the CEO of YouTube. Ah, yes. Correct. Sorry, not the CEO of Google. Thank you for correcting me. Then, the CEO of YouTube resigned. In fact, we have seen that advertising on YouTube has slowed down. Google search traffic also slowed, although it remains positive year over year. Meanwhile, YouTube's numbers dropped.
Apparently, some say it could be because FTX wasn't paying some of its advertising bills. We'll see. Which is quite remarkable . Yes. What do you think of the argument that if ChatGPT and this search AI work this well, the number of searches will plummet and Google's advertising business will implode or suffer substantial deflation?
I think it needs to be viewed in four stages of realization that will occur. First, there are people who are still in the "this doesn't affect Google" stage. I think that's a mistake. And I think that could be a profound mistake. The next stage is: well, Google has a great AI capability, which they do have.
Well, they can run a language model, just like Microsoft does, and use it to answer search queries. And that's fine, except that Google gets one and a half cents of revenue per search, and if, uh, you allocate the costs , they get maybe half a cent of profit per search, and the cost of inferring ChatGPT, the cost of running that AI system.
Like seven, right? Well, it depends on how you make the estimates, but it's on the order of, you know, a penny, maybe half a penny, whatever. If Google had to run an extensive language model on every search result, that would wipe out all its profits and then some . In other words, it would massively destroy the margins of...
I see. Alright. So, at Google's scale, they might be able to reduce the cost to half a cent if you basically include AI in every search, but that's all they make anyway right now , that 's the concern. Exact. And so, this is what people are starting to consider, that if Microsoft puts competitive pressure on them by saying, "Hey, look, Bingbot does searches really well," and Google has to respond to maintain its position by doing the same, that's all very well for Google, except that its costs skyrocket relative to its revenue.
The third stage of understanding, which I don't think people have really grasped yet, is that the way a search engine becomes an extended language model means it is no longer a search engine. Instead, it's an answer engine. If I'm getting an answer, I don't click anywhere else to get the material. And hey, you know, their entire advertising model is based on clicks on search ads.
Therefore, there is also a reasonable argument that if you switch to extensive language models, you not only increase the cost per search, but you also harm search revenue. So you affect gross income. And the next stage is, wow, not only is it more expensive to deliver these results, but the entire business model on which you base this quasi-monopoly of, you know, $160 billion a year, is deeply broken.
And then, the fourth stage of realization, and I think this is the actual way it will happen, is , uh, the whole process by which we go from doing something to saying, " I'm going to step out of this experience as a consumer, business worker, or whoever," to go to a separate portal that will then take me somewhere else in search of information.
And that's what the search is. It's basically like a transit station that sends people to other places. I think all of that is going to disappear. I do a lot of work in Excel, right? And my typical workflow is: I need some data. I'm going to Google. I'm looking for the data. I go to the primary sources.
I get the data. Perhaps sometimes it's through an academic site that tries to charge me for a subscription. So there's a sponsored link that I click on and that generates revenue for Google. In contrast, requesting that data will be directly integrated into Excel ; It will be like: "I need the CPI from 1882 to 2017 or whatever" and I will simply ask him and he will complete it.
And then , I'll never have to go to that through station . That's why I believe the final stage is that the text indicator, the extended language model, is not separate from everything else you do, but directly in the final application you're working on. And that, ultimately, will impact Google's traffic, and, well, do we really need a separate search engine?
Exactly. Particularly because the response that the language model can provide will be much more powerful, since it will have the context of what you are working on. Correct. Correct. And therefore, it's not clear to me that the way you plan a trip requires you to jump into a search engine. Instead, it could be directly in your calendar app or your maps app .
In your Tesla or in your Tesla, right? So, while you're on your way here, you can sort out your restaurant reservation, hotel booking, or any other tasks you need to find out directly from the Tesla interface. I suppose, in theory, we don't want a Google, we want an answer. Exactly. Interesting. So, if we're looking for an answer on YouTube, in theory, we don't want a video.
We just want the answer. Correct. You're bankrupting me with these ideas, sir. Actually, I think there is an interesting, likely byproduct of this, which is that the information that exists and with which these engines are trained will, in fact, be amplified. And then, you know, maybe you're not famous enough for it to happen, but for, you know, um, for, um, the Harry Potter universe, for example, my kids use ChatGPT to put themselves in scenes with Harry Potter and get little short stories of them dueling with him at Hogwarts, and it's really good at that.
It can generate a scene that has dialogues between my six-year-old daughter and Harry Potter, and she saves the day by curing him of some spell that has been cast on him. And it's entertaining, and it does so within the context of who Harry Potter is; it somehow understands how that dialogue would work within that world.
And that's because there's a very deep text database of all that information. And that means that existing intellectual property owners, people who already have intellectual property that has built a world around it, I think that will be amplified. We're never going to get out of Harry Potter because JK Rowling doesn't really need to produce more material for more material to be produced based on that .
Clear. Clear. Take this, learn these books, and manipulate the stories however anyone would prefer them to be manipulated. Inject several children or whatever . Exact. And in the future that could theoretically also include video generation. Yes. Either video game generation or virtual reality. Yes.
It's like you might imagine, you know, these stories aren't as good as the original books, but they're actually pretty, pretty good. Like when I read a short story to my daughter from there. I think: this is ingenious. He kind of invents spells that are funny. You know, he does or, perhaps saying "invents" is the incorrect verb.
But, you know, it's captivating enough for her at least, at least as good as those old made- for-TV movie sequels, you know, those bad ones Disney used to produce. Now, the amazing thing about this is that most people—and this might be in your early stages of thinking about AI—say, "Oh, the way to protect yourself from AI is simply to be creative," but what you've just done is destroy even that.
Well, maybe. Yes. Unless you need to create new bases. Yes. Yes. I think, I think that people, you know, for a long time thought, "Oh my God! Artificial intelligence and automation are going to kill the blue-collar industry." Exact. Exact. Well, it seems that what it's actually doing is affecting creatives, software developers, and, you know, jobs that require a lot of technical skill and are very well paid.
Now, our belief is that the way it will manifest is by making creators who are truly creative more powerful , you know. Basically, turn them into JK Rowlings. Clear. Clear. It's like a massive force amplification. See what's going to happen in the software space with GitHub Copilot; They published a study suggesting that they could make software developers twice as productive just by using Copilot as it exists today.
Clear. And each of these systems is improving. So, if you look at the entire knowledge work space and the amount of productivity advancement that we believe AI software is going to provide to knowledge workers, it could be the greatest macroeconomic productivity leap of all time. That is, we literally spend, I think it's on the order of 50 billion dollars a year on knowledge worker salaries.
So imagine you give all those knowledge workers software that makes them twice as productive. That would be worth, you know, 50 billion dollars in production. So, even if you pay a discount for that, we think you'll end up spending over 10 billion on AI software . That's incredible, simply because of the speed at which money is being spent.
You're talking about 5 or 6 times more for every dollar spent. If people can be much more productive and earn more money, they can spend more money. GDP is growing substantially faster than currently estimated. I believe you have GDP estimates for 2030 of over $100 trillion for the United States, something like that .
Well, today there is approximately 100 trillion in global GDP. If you look at the consensus forecasts. For the United States, it ranged from about 20 to 21 billion to 100 billion more or less. Yes. So the global consensus forecast is $130 trillion by 2030. We believe that, based solely on robotaxis and what will happen with robots alone, we think they will reach over $170 trillion in GDP by 2030.
So the difference is that the consensus for average real GDP per capita is $15,000, and we believe it is likely to exceed $20,000. So everyone on Earth, on average, will get a 30% increase over what was expected thanks to the technologies we are focusing on. That leaves out the AI software aspect, because it's less clear to me how that will be reflected in real GDP, such as, for example, that AI software will enable robotaxis.
That will be very well measured with traditional macroeconomic statistics. The fact that we will have much better software in a couple of years, something I am convinced of, is not so clear as to give credit to GDP for it. Because you at ARK also believe that working hours will continue to decrease.
So you have more time for family or anything else . Clear. And you can see, you know, I think there are many ways in which consumer surplus is not very well measured in GDP. Think of it this way. It's like deciding to stop paying for cable television and instead opting for a number of streaming services; So, in terms of what you pay for entertainment, you could go from over $100 to about $50 or $60 .
TRUE? That's a terrible event for GDP. You've basically reduced GDP by $40 just by making that decision. But you're probably seeing the same amount of entertainment, and it's probably better. Yes. So, from your experience in the economic world...it's better. You're getting cheaper entertainment of the same quality.
In fact, probably of higher quality. You don't have to watch ads. But macroeconomic statistics would measure that as a decrease in consumption. That's fascinating. Now, when you mention that by 2030 everyone's per capita income could increase by 30%. Today, obviously, in an inflationary world, people don't feel very good about that kind of potential increase.
But I think the world you describe for 2030 is, in fact, substantially deflationary. So a 30% increase in revenue could be worth much more than that. How do you plan to address the current inflation in the short term ? What are your thoughts? I mean, today we had a pretty stagnant PPI report. We had a slightly strong CPI report for January .
We had a jobs report that was intense. Many people say that January is the month of seasonal adjustments. What is your opinion? Yes, I mean, I think technology circumvents inflation and will continue to do so. Think, for example, about the electric vehicle space with your Tesla. You know, if someone buys a Model 3 instead of a Toyota Camry, they pay maybe one and a half times more for that car upfront.
Clear. TRUE. And then, its ongoing annual expense is 0.7 times what it would be with the Toyota Camry. So the total cost of ownership is actually lower. They're getting, they're getting a base of one dollar per mile. They travel from one place to another in a less expensive way. And the way it manifests itself is by spending in advance in exchange for spending less later.
And that's it, really, you know, that reading of the inflation statistics is a bit mixed. As you know , they try to adjust for the fact that you 're buying the more expensive vehicle, but there's no adjustment anticipated for the lower maintenance and fuel costs you'll pay over time. Uh, and one example, I think there is, uh, you know, in a way, inflation is a function of people still having money they can afford to spend.
If you pay less for entertainment, if I cancel cable, yes. True. And I pay less for entertainment, well then I have extra money that I'm going to spend on, you know, steak instead of rice, right? And then, you know, there are some relative price adjustments and ultimately I think the deflationary forces of technology are going to pull in the cost adjustments.
So, it's clear that inflation should disappear in the long run due to the deflationary aspects of technology, that's what you're saying. But in the short term, hey, people still have plenty of money to spend. I think Bank of America just published a report. They said that the average person who before the pandemic had $2,500 to $5,000 in their bank account now has $12,800, and that has only decreased by 4.4% in the last 12 months.
Yes. So, don't you see that excessive spending disappearing? People feel richer. They feel they can continue spending. I think American Express was talking about how people are spending despite the recession or whatever it is we 're going through. Is that your opinion that yes, in the short term you can have that inflationary impulse here , but technology is clearly going to lead us to a deflationary trend in the longer term.
Yes. And I also think there are some strange mixed changes happening, like, imagine within the housing sector, which clearly, given what's going on with interest rates , should be in serious trouble. It does n't seem to be. I think it's partly because, you know, people who are still working from home need more space, and what that means is that, in the meantime, the commercial real estate sector is in quite a bit of trouble.
There were two large buildings in downtown Los Angeles that were returned to the bank; They had just finished them and said they wouldn't be able to rent them out. We are forgoing the loan and returning it to you. That's how it is. He appeared on Bloomberg saying: we're returning the keys to the bank.
I don't think it's that easy, but oh well. But the reason they can't rent them is because companies have partly passed on the cost of office real estate to their workers. And what does that mean? This means that it no longer falls on the business's cost line. So the business seems to have a higher profit margin .
But the worker is there thinking: well, now I have to pay for an extra room. So they go out and look for an extra room. Then they go back to the company and say: well, you have to pay me more because I'm covering this. So, in reality, this shifts the cost to residential real estate, and the consumer recovers some of that through wages.
But isn't that inflationary? It's really just a change in where the spending occurs, whether on the corporate balance sheet or the consumer's. And I think we're still processing some of those changes in behavior. Um, I mean , you've actually argued here that some of the biggest wage increases we've seen might not be real inflation at all .
It's simply shifting the expense from the company to the individual. Yes. It's incredible. Yes. Yes. But traditional data surveys can't measure that, because this hasn't happened before. Technology and the ability to work from home after the pandemic. Correct. Correct. So, I think there's a lot of this, certainly as a response to the pandemic, but also on the technological side.
As technology accelerates, it becomes more difficult for macroeconomic statisticians to capture it in GDP in a way that accurately reflects reality. For example, even with smartphones , it took them a decade to realize they needed to deflate specifically because of improvements in smartphone performance , as in What do you mean?
So, imagine you spend $100 on a phone every year and suddenly you go from buying, you know , a Nokia phone to an iPhone and then you end up spending 4 times as much for the phone. For a decade that was counted as, um, inflationary pressure. Ah. True, because the way it manifests itself is as if the cost of phones were going down.
Clear. All the time the cost just stopped going down so much. It kind of leveled out. It leveled off, not because costs stopped going down, but because people were buying a much more complete product at a higher price . Yes. Now they buy a video camera , a phone, and more. Exactly. How do you make an adjustment for that?
Yes. And then there wasn't a real attempt to make an adjustment because of that until the iPhone 10 came out, which was much more expensive, and they said, "Oh, dear, we really need to adjust it ." So, that's right. Because AT&T, T-Mobile, or whoever, were subsidizing many of these with the monthly plan.
And then initially, as you said, you went from $100 to $400, but now you're paying $1,000 to $1,500 for a phone, right? TRUE. And it's a phone that is, you know, massively more efficient. It's certainly more efficient than the rotary phone you had, so for a decade of smartphone use, they used that same series of declining costs that applied to the rotary phone.
So it seems inflationary when the reality, with all the technological improvements you get, is that it is substantially deflationary. Exactly. Exactly. And then, going back to AI, it's like, I 'm not sure, you know, it'll probably take a decade before macroeconomic statisticians come back and say, oh, we really need to reassess what's going on here, how big the economy is.
Um, but I do think that one of the things we've focused on over the last year is how technologies are converging and how you essentially have an acceleration. One technology is enabling an acceleration in another. And I'm sure that the acceleration in AI that's happening now won't be limited to just a chatbot that, you know, argues with you on the web, right?
Medical care, vehicles, whatever . Yes. It's like, look, you know, Tesla's breakthrough in full self-driving during AI Day, the most recent AI Day . Its ability to navigate lane intersections is based on the transformer architecture that was originally introduced only to improve language translation, as a breakthrough to translate, I don't know, Dutch to German in general.
It turns out that this advancement applies directly to how a car can understand an intersection it has never seen before. That's extraordinary. What is your opinion about Tesla? Obviously, with Elon Musk's acquisition of Twitter, there have been many rumors that he might take Tesla's AI technology and bring it to Twitter, and perhaps even create a competitor.
Some people have been spreading rumors about ChatGPT. Do you see that as an ambition of Musk's to get more involved in that neural AI that people then link to the future Tesla Bot? I mean, I think it's clear that Twitter has a unique dataset and should be able to develop its own language models that should be quite powerful.
Um, one of the people who created the character limit on tweets may have actually inadvertently made their dataset more powerful than it would otherwise be . Oh really? So, shorter is better. You're saying because it forces people to basically compartmentalize, to quantify their ideas, like if I send you a block of text and you 're a machine and you need to break it down into unique ideas, that might be difficult, and because people can amplify individual ideas, it means they have a unique data set in that way.
TRUE. TRUE. Because if you have a thread, a computer can now see which part of that thread is amplified the most, instead of receiving the entire block. You figure out what the highlights are, computer. Yes. Exactly. Exactly. And then, the fact that there is metadata at the idea level. Metadata referring to retweets and likes.
And that essentially the human work of breaking down larger ideas not just into sentences, but into whatever the maximum of 280 characters is, call it the minimum viable idea, means that they have a really interesting data set . And then, but I don't think , you know, it's not like it's an advancement that's in Tesla that's going to go inside Twitter.
On the other hand, I think Twitter will try to use all the language advancements that are being published and that companies like OpenAI are using. I believe that, in the field of robotics, it is fascinating how much the use of these language models can make robots perform better. And now Tesla has bought its own robotics facilities in Germany.
Their technology was from a German company in 2016, I believe. Obviously we know that they have autonomy, they have robotics. What advantages do you see in them since our last conversation, compared to Ford, for example, where Mr. Farley, the CEO, mentioned that they had an extra mile of wiring in the car and everyone was saying that Ford is the best at manufacturing?
They've been doing it for 100 years. Meanwhile , have the Lightning manufacturing lines been closed for 3 or 4 days? They are expected to remain inactive for another week due to battery issues. What is your opinion on the ability of traditional manufacturers to compete in this electric vehicle space?
I do n't think it's clear that they can. I mean , I think Tesla just has an advantage when it comes to powertrain efficiency. I think it's a three or four year advantage. And then, that... that's just in manufacturing. Yes. It's basically about how efficient a powertrain can be per dollar, including the battery system.
And I think the challenge for traditional manufacturers is that they still seem to follow a design-to-production innovation cycle of about four or five years. They are stuck in a half- decade routine, planning and launching well in advance. And it seems that, once they've been on the road for a year, they're already conditioned to where they're going to end up.
So, the problem is, and we saw it with the Model 3... Tesla launched the Model 3 in 2017. It took traditional manufacturers five years to launch models this year that perform similarly to the 2017 Model 3 , right? And, on March 1st, I believe Tesla will announce a vehicle that, according to them, they will be able to produce at half the cost.
Do you think they'll announce a car? Oh, yes. Yes. Yes. Yes. I mean, they've definitely indicated that it's very likely they will. Well, I think it will be a new model that, uh, you know , will be, uh, very economical and designed to be a robotaxi, probably. And that, you know, will just throw the whole industry off balance in terms of costs again, because now the industry is trying to figure out "how are we going to scale?" And, you know, I just think they're designing with a manufacturing cost that won't allow them to compete profitably at Tesla's level.
But Charlie Munger says BYD is better. I mean , I think, for starters, the Chinese market is a very different beast than the American market. And me... The average number of vehicles there is about 34,000. Our average new car price is 47,000, give or take . Yes. Well, I don't know the Chinese average, but there's a whole class of vehicles that are more like neighborhood electric vehicles.
So it depends on how many of them you include in how you calculate the average. But overall, I think that in China Tesla is more likely to be positioned at a premium level and not be competitive against the high- volume manufacturers we will see in the Chinese market . But I think it's very clear to us that almost the entire automotive market will be based on electric propulsion systems .
We believe that in 5 years it will be more than 90% of the automotive market. We also think the automotive market will be smaller because robotaxis will cover more kilometers than traditional vehicles. So there will be a compression in overall sales, but you'll end up with around 60 million electric vehicles sold in 2027.
And we think that, you know, Tesla is going to... Globally, of course. Worldwide. Yes. Tesla is going to dominate a significant part of that. You know, about 20 million vehicles or so. And, uh... I find it hard to see how at least one traditional car manufacturer can scale up and compete successfully.
And then there are startups; I think designing an electric vehicle from scratch is a better way to approach the problem. There's a reason Ford ended up with too much wiring in their vehicle, as you said, and it's because, you know, they didn't ... They probably didn't take a first-principles approach to design.
And, uh... Yes, I just think there are huge problems with traditional car manufacturers. Um, even putting aside the pressure they'll face from robotaxis and, well , the fact that they have a lot of auto loans on their books that probably won't look too good in the coming years. Yes. I imagine you saw it, but I found it so ironic—or perhaps not ironic really—that a month ago Mr.
Toyota, the CEO of Toyota, was at a conference talking about how he thinks he's part of the silent majority, that people want hybrids and that's why they're going to bet everything on hybrids, and then, in 30 days, we find out that he's going to be fired. A new CEO is coming who will drastically transform Toyota.
So here we are in 2023 and they will only have a new CEO in April who will only now begin their purely electric division. What's your answer to that? Is it just an expected transition? I mean, our perspective has always been that electric vehicles simply make more sense. They perform better. There is less complexity.
I mean, with a hybrid, you really have to build almost two simultaneous powertrains. And then, um, and I'm sympathetic to the idea. I think the reason people find hybrids appealing is because they say, "Listen, the average person only drives 30 miles a day anyway, right? So why do they need a 250- mile battery if they can drive 30 miles using electricity and then anything extra they can use the gasoline engine?" But then you introduce much more engineering complexity.
Also, for example, with plug-in hybrids, people don't plug them in. Because it's like, "Well, I'm going to drive it anyway." And then you never really stay in electric mode as long as people assume. Um, and you add a lot more weight to the vehicle and sacrifice performance. And finally, um, also, you end up with a vehicle that isn't as safe because you still have a large internal combustion engine at the front of the car that can be pushed towards you in a head-on collision.
And so, uh, you end up with a less safe vehicle, with less performance, um, and, uh, and, you know, more expensive. And I simply don't think a consumer is going to optimize for that . I find it difficult to see how that could be the dominant mode. Um, I've had the impression that we could probably reduce it, and I 've mentioned it several times on YouTube, and there are always people who say, "No way." It's never going to happen .
But I want to know your opinion on this . My idea is that we probably only need a 20 kWh battery for most people. That reduces the vehicle's weight because you're carrying a battery that weighs a third as much. Certainly, a third of the cost, maybe even less. And that would give it a range of 80 to 100 miles.
That's more than enough for your daily trip to the supermarket, as you mentioned. Probably. I mean , I think that was it... Reach anxiety is not rational; I think even the first electric vehicles like the GM Bolt had that idea: "Hey, we'll deliver a vehicle that...you're not going to take a long trip in it, but it will be your electric car for the city." I think for the American buyer in particular, when someone buys a car, they think, "What's the most exciting thing I'm going to do with this car this year?" And it needs to satisfy that need.
So, and they never do anything. Yes. And it might even be in four years. It's like, why do people buy pickup trucks? It's not because they transport things every day. It's because one day they'll feel great when they need to throw a mattress in the back and can do it. So I think when you make a big purchase like that, you're buying it with that unlikely event in mind.
And that is simply the reality for the consumer. USA. That could be it...I 'm focusing on the American buyer because I'm more familiar with that. Yes. And I think the neighborhood electric vehicle segment is very interesting and hasn't been penetrated with a product as compelling as it could be. Interesting.
In other words, they all look like golf carts right now, and they're expensive golf carts . But I always thought, I never... understood why Apple didn't go after that segment. If they were going to make a car, why not make a $15,000 Apple car that was for the city, or even $20,000? And you'd have a group of buyers who would probably say, "Oh, that's it ." "If it looked cool, you'd massively expand your design and engineering space because you're no longer designing something to go 70 miles per hour on the highway, right?
And then it could be a really interesting product. That's interesting. Now, as for the robotaxi, you know, there, because it's carrying a lot of people around, you still need a relatively large battery because they'll probably be going 200 or 300 miles, maybe even in a day, right? And so, in the end, you do n't really end up with that small battery size.
That's interesting. So, your idea is that the next vehicle Tesla might announce wouldn't necessarily be something with a smaller battery. The robotaxi is the direction, I think. So, and I think, I mean, if I had to guess, I think there's an optimization they could make by offering a robotaxi that at least has two or maybe four separate passenger compartments that are independent of each other, like in...
Oh, wow. Because part of the problem with the robotaxi, in our opinion, is that you'll massively increase the number of miles the People drive on the highways because it's much less expensive and much more convenient to travel by car. Right. And so, for example, I don't go to downtown Los Angeles because it's an hour and fifteen minutes ride and it's expensive in an Uber, and driving myself is tiring.
But what if instead of costing $3 per mile, it cost 30 cents per mile? Yes. And I just sit in the back seat. Not only will I do it, everyone will . And so, traffic will get worse. Because the pain of being stuck in traffic isn't as high, and it's not as expensive to get from one place to another. Right.
Okay . And then, you know, Uber had this great idea. They said, we'll do Uber Pool. People will save money by riding with other people along the route. And... but that's weird. People, yeah. People didn't use it because then you're sitting next to a stranger, right? You know, and then, you definitely wouldn't want to do that in a robotaxi because in a robotaxi, either you're alone, you don't even have the driver...
Uber, right? You're either alone or with a stranger. There's no way you'd choose the stranger to save like 5 cents a mile, right? But if you could create a robotaxi that has separate compartments that open up and is like your own little pod. And since it seems like you don't really care how long it takes because you're just, you know, watching Netflix anyway, then you could have something that would allow that kind of "pull" concept to work.
That's interesting. They could do it. Because they also have some FCC patents that, in my opinion, are starting to suggest that maybe they're trying to bring Starlink to cars. That could potentially align with this idea. Who knows? Sure. I mean, I think Starlink is everywhere—I mean , you know, they already have Starlink for RVs.
I think, and this is Sam's work, who covers reusable rockets for us, but we think there's a tens of billions of dollars opportunity in connecting moving Starlinks— our constellation of satellites in low Earth orbit—to moving vehicles, cars, airplanes, RVs, boats, you name it. Yeah, maybe. Yeah, I think there is, you're weighing the cost of the antenna against the utility you get from having it on the vehicle.
But it seems like, if they're doing something like enabling connectivity for people on their commute, it makes a lot of sense. So that's really interesting. So, initially, since you guys have a very strong bullish thesis on autonomous robotaxis, I think your projections for Tesla almost exclusively, and your bullish scenario, give most of the weight to robotaxis.
It's like two-thirds of Tesla's value in five years, in our view, driven by robotaxis, versus one-third electric vehicles. Um. And I suspect the product they'll launch will be an inexpensive vehicle platform to manufacture, optimized to fit the robotaxi, but that we can also sell to people who are the average car buyer—not just the average in the US—which is priced lower.
So that would be the car in about $25,000, plus maybe some inflation since then. Exactly. Correct. And yes, I don't know what the price would be, but think about half the production cost of the Model 3. So, if the average Model 3 sells for 50, yes, maybe 25 or 30. And why do you think they would announce that now?
Why? Because I think they're getting ready to manufacture it because they're doing it even before the Cybertruck and the Semi. Well, no. I think they will announce that they're doing it. They'll reveal it just like they did with the Cybertruck. They'll reveal what it looks like and then go through the engineering for manufacturing .
So that will take a year or two. Yes. Yes. Yes. Wow. Wow. Okay. So, everything at Tesla except when Jerome Powell gets angry, huh? Well, I mean, listen, in general, innovation stocks have underperformed year over year. There's no question about that. I think what's really fascinating to me is that the rate of technological improvement in the technological platforms that we focus on has never been higher Fast.
At least in my career analyzing technology and markets. It's, you know, in the AI space, I've gone from having trouble keeping up with academic papers to having trouble keeping up with product launches. Yes. Right. And so, the rate—and it's not that academic papers have decreased. You know, if anything, having these tools to play with allows people to do more experimentation and discovery, and that directly feeds into all the other technologies we focus on.
What do you think about that? Because I don't think you guys invest a lot in chip manufacturing; you're more software-driven. What's your take on that? Are you intentionally avoiding anything that's labor-intensive and commodity-driven, or is there a different reason? Perhaps, for example, even an equipment manufacturer like ASML, or Intel, you know.
Yes, I mean, I think there are structural issues with Intel in that the CPU is no longer the critical path for performance in ... any kind of computing application, as it's really your AI accelerator. And so, if the CPU is Just as a glorified router instead of something that actually drives performance, why bother spending on Intel margins when you can use an ARM CPU that's just as efficient, maybe even more so, for less cost?
It seems like they're now moving decisively into manufacturing because I feel like they're starting to see the reality. Right. Right. And then, there's the foundry side. Is chip manufacturing an interesting business? It is. And well, we think TSMC is a very interesting company, but it's a business that requires a lot of capital and a lot of investment to get to the next node and deliver the next level of performance, which usually translates to lower margins than we're interested in.
But I think the case for companies like TSMC is that, as AI will be in a wide variety of devices, there will be some kind of chip in that camera that requires more computing power than exists today. So, that means TSMC will have a wider range of buyers for its chips. Therefore, it should have more pricing power in that situation.
But, you know, we tend to focus on— I think we do invest in companies that are the enabling tools for the technologies we're focused on—but we really try, as much as possible, to focus on the platform itself. The end product, essentially. In other words, going back to what we started saying at the beginning, we don't want google.com, we want the answer.
What the chip does for us, ultimately, is what I want to deliver a good return to my end customers. But I think when we analyze companies and even entire sectors, we look at where in the value chain we think the cash flow will accumulate and how those companies are valued. I think with AI, the analogy with the early days is quite apt.
Back in '92, or maybe '91, the company to invest in was Cisco . You know, it wasn't... Amazon didn't exist yet, did it? I mean, even AOL hadn't started operating, right ? And that's because providing the tools that connected all these computers together was the right place to be. to invest. And I think we're still in the early days of connecting the world through AI.
That could mean that, with Nvidia on the other hand, we have massively aggressive expectations for investment in AI hardware. We think that by 2030, spending on AI computing infrastructure will be on the order of $2 trillion a year, so a portion of that will flow to Nvidia. But, you know, it's still valued very highly, even with that expectation.
Right. To some extent, you know, the challenge of capital markets is that what everyone expects to happen is already priced in. Yes. Exactly. Yes. Absolutely. So, well, I always want to bring this up as an opportunity for its defense. A lot of people in the mainstream media like to say, " Oh, Kathy and the Ark team, it's unprofitable tech." But then you look at some of the companies, and they actually have free cash flow.
There's just a lot of stock-based compensation that might show negative net income in some quarters. What do you say in What's the answer to that conventional argument? Well, ultimately, companies are paid based on cash flow—not the cash flow they're generating right now , but the free cash flow they generate over the life of the investment.
Right? Or the market's anticipation of that free cash flow once the market starts anticipating it, right? And so , what we pay a lot of attention to is the sensitivity of companies' balance sheets. So, you know, how much money do they need to grow at the rate we think they're capable of growing, and can they generate those needs internally, and if not, how are they going to raise the cash to do so?
But regardless of whether a company is investing heavily in a strategic platform that we think is very large or not, we would actually prefer that they invest heavily, not try to distribute cash to shareholders, pay dividends, or buy back shares. I think generating a lot of excess cash means they haven't identified a good place to deploy that excess cash.
And the reason we invest in the technologies we invest in is because we believe there are a huge value accumulation push behind them. And so, companies, to the extent they can spend money to increase their probability and strategic footprint in these fundamental technologies. That's in the best interest of shareholders and their investors.
That's an interesting idea: that if companies are doing buybacks, they've lost sight of what we should do with this money. Let's just buy back the shares because they've dropped a little. Now, I think you all have big ambitions, or should I say big projections, for connected TV. You also have a big investment in Roku and not in Trade Desk.
Trade Desk just announced big buybacks, and Roku just increased its net loss. What do you think of Trade Desk? Is that an example of Trade Desk doing buybacks because they don't know where to invest the money? Let's continue with Roku. I mean , I think I'll say that we can make a very clear case for Roku and the value appreciation that we believe will happen with Roku.
I mean, if you look, they just reported a quarter where all their engagement metrics exceeded even the case Based on our published valuation model. One way to think about Roku is that they're currently generating about 3 cents per hour streamed in net revenue for the company. Right. And think about traditional cable TV; when you pay for a cable package, you're paying around a dollar per hour, right?
And even if you pay for Netflix, given their streaming trends, it could be 20 to 25 cents per hour. Right . So Roku shouldn't demand that, because they're serving as a generation of potential customers for Netflix, right? But we believe that as people leave traditional TV and switch to Roku, they'll create an ecosystem that should generate 5 cents per hour.
That's factored into our expectation for the company. And the engagement and audience trends they 're seeing are right in line with our base forecast. So, if you combine that 5 cents per hour in net revenue and the continuation of the engagement trends, you arrive at our target price, which is about $600 per share.
Exactly, and then... I think it's a really interesting exposure Because, to a large extent, ad dollars haven't yet followed people into the streaming ecosystem. Correct. True. When and if they do, Trade Desk could also be an interesting exposition on this basis . We personally think, or at ARK we think, that having the operating system for the TV platform is the best place to accumulate that lead generation toward ad dollars, but that there should be an entity that sits between all those ad dollars and distributes them to viewers, and Roku fills that position.
You know, I think something that really resonated with me from what you mentioned was this balance sheet sensitivity . I very briefly analyzed Roku and Shopify's earnings this morning, and while they're both losing money—substantial losses in terms of net income, much of which is stock-based compensation—their balance sheets are incredibly strong.
I think Shopify has over $5 billion in available cash. I think Roku is somewhere close to $1.6 billion, if I recall correctly. So I think you're right. They have plenty of time to essentially make those things happen. Visions. Is that one of the big priorities when you at ARK do any kind of analysis, to see if you have the balance sheet to back up that vision?
Well, for us, the first way we refer to our decision-making flow is top-down: Is this company positioned against a technology that we believe in and that we believe is going to... that basically has a structural tailwind in a sector where we think the cash flow will build extraordinarily? And then, does our view differ from the rest of the market in terms of that top-down approach?
That's a consideration from the very beginning: Do we have a different opinion than everyone else? Because if it's something expected, it's already priced in. Right, okay. True. But it's also an easy way. I mean, I think, you know, we look at technologies like virtual reality as an example where we've said, "We do our job, somebody else does theirs; We don't look at each other's work before we finish, compare, and say, 'Oh, we're even,' or 'We're below where everyone else is.' So, it won't do us any good to keep trying to bet on that strength.
Yes. And then, given our expectation for the technology and the company's position , what might their revenue look like, and what margin structure would the company need to be able to build to get to that number? And if that's interesting, then we re- evaluate: Okay, here's where they are from a balance sheet perspective .
This is what their cash burn looks like if they're still burning through cash, and then what do they need to get to that tipping point where... you know, the point where the market is likely to start noticing and recognizing the business's cash flow buildup potential. Ah, that's fascinating. Yes, I love the idea that if it's something that's already a popular belief, you're going to look elsewhere.
I think that's fantastic. Now, with virtual reality, you didn't sound very optimistic about it . Yes, I want to. I mean, I think it goes totally against the trend in technology and user behavior with respect to technology. It's difficult, uh, at least with the current format, for us to see this actually work.
I think, well, we've done a cost reduction on VR headsets, and we think that by, I think by 2032, for between $1,000 and $3,000, you'll have a headset that you put on and it will be visually indistinguishable from reality. So I do n't know if that's really significant. Conceptually, there should be something interesting before that happens, but, but, but in terms of...
thinking about it, it's still, it's still very expensive for what it is for the end consumer, and, you know, there are rumors that Apple's headset will cost $3,000. Yeah. That's not a product that's going to work at that price. Interesting. And so, you know, the reality of these headsets seems to be that , for the most part, they're given as Christmas presents and then they just sit on a shelf.
People don't He uses them. So Nick, who covers the consumer internet space for us, he has one. And it's like, "Nick, you have to let me know when you actually use one of these things." You know, because there's a real chicken-and-egg problem between if you only have a small audience buying them, then no game developer is going to dedicate a big budget to creating a game specifically for that experience.
Without a gaming experience, like a game specifically for that experience, it's going to be hard to get people to buy them. And so, I think for me, a subjective indicator I look for is: when am I going to read a news story about some poor teenager who starved to death because they were so engrossed in their VR experience that they didn't go out to get food or something, right?
There have to be those kinds of dramatic stories about this being a terrible stain on humanity before we're anywhere near a product that's going to be mass-adopted. There have to be early adopters who get lost in it before it's something that somebody will spend, you know, $1,000 on. And then put on a helmet.
And that's not even close to happening, in other words. Right. Right. Right. So, making the extreme case, obviously we don't want anyone to die, but since we know that no one is taking it even close to the extreme limits , probably not... I feel really bad for the teenager. I'm just saying if that were happening today, it would be on the front page of the New York Times or something, because it's a very morbid story.
To be convincing to them, too. I want to prove that. It's not that there isn't, you know, a market for that kind of media, and since that media doesn't exist, it means these things aren't used as much. Yeah. Right. And and and uh, the and then, like, the trend isn't just having one screen that's the only thing I look at.
It's having like six screens, right? It's like instead I'm watching TV and on Twitter at the same time. So I just think if virtual reality takes off the way it's being presented, it's a huge shift from the direction people have been going. You get virtual reality in your robotaxi. Look, but I mean, you know, on one hand, you'll just have a screen in the robotaxi, but realistically, you probably won't want to put on someone else's headset.
Taxi headsets are going to be a little weird. Yeah. Uh, so you guys were in Matterport briefly. Yeah. Not at war. What happened? I mean , you know, in the process of the recession last year, there's a sorting in portfolios, as we say. You know, it's not that we don't like this or that company, but, uh, you know, we try to concentrate portfolios during recessions to, to really, essentially, put more strain on the spring for change.
Uh, and so, uh, it's not like there was, you know, some kind of morbid event or anything that, uh, it's more like a relative value perspective and where do we have more confidence? Yeah. Yeah. I mean, anyway, I agree with you. I love Matterport, don't get me wrong, but I also know that The number of users has increased like crazy selling the iPhone version of Matterport, and I'm not a big fan of that.
The larger cameras they use for, you know, purposes like scanning or real estate are fantastic, but we know what's going on in real estate right now , commercial real estate, right? And there's a certain sensitivity in the markets and the macro cycle. Absolutely. Something you mentioned made me think that you need to build a group of people, you need people who use the headsets to create the games, but you have to offer something compelling so that people want to buy those headsets.
It made me think of— and nobody likes to mention this company anymore—but Nikola had a very interesting idea about the hybrid hydrogen- electric battery vehicle, and that having people with a hydrogen vehicle would induce demand for refueling stations, but in the meantime, they could use the electric battery charging stations.
Is there any remote future for that to ever come back ? I mean, the work we 've done on hydrogen fuel cells suggests not, but you know, with every technology we look at, we periodically review it and We asked whether the cost decrease we calculated is still holding true or what has changed in the unit economics for buyers.
The problem with hydrogen passenger cars, and even in the transportation sector, is that... a hydrogen station is very expensive. Let's say a million dollars or a couple of million, whereas a supercharger costs about $100,000. It's that kind of difference in magnitude, right? So, from the consumer side, it's really difficult to complete all the necessary infrastructure to make it useful.
And then the cost of the fuel cell itself is still very high. The work we do suggests that it would take a passenger vehicle selling at levels similar to the Toyota Prius for a decade to make fuel cells cost-competitive with electric powertrains. Oh, wow. I see. I see. Now, obviously, that brings us to lithium.
We 've all seen lithium increase tenfold in the last few years. Everyone is thinking, how can I invest in and profit from lithium? But you don't invest in raw materials Commodities. We 've done it in the past. I mean, we've invested in Albemarle before. We're not doing that right now. And I think with commodities, the reason for not doing that is certainly the number of people talking about how to invest in lithium makes me think it's probably not a good time to do it.
Interesting. And you know, with commodity markets, the price signal is very powerful, and it's not like ... if you look at the whole raw materials chain for electric vehicles, cobalt was a bigger concern for us because cobalt is co-produced with, I think copper and tin maybe, anyway , it's co-produced.
So, if I give you a high price signal for cobalt, that won't necessarily drive the economics of bringing in more supply, because you could have a price signal without an underlying improvement, so you wouldn't increase production. Lithium, on the other hand, um, you know, a miner sees the price of lithium and it's like a big flashing dollar sign that says, "produce more lithium." Exactly, exactly.
Want to get rich? Go mine lithium. And if you look at the reserve-to- production ratios for lithium, they're, um, extremely high, which means there are tons of lithium out there ; you just need to get the mines up and running. If you look at our forecast for electric vehicles, which is more aggressive, I think , than anyone else's that I know of, um, the total capital expenditure needed to support that and bring in additional lithium supply is $30 billion .
Okay. So, relative, it's kind of like manufacturers have announced $600 billion in capital expenditure for electric vehicle programs . So, actually, it's not that big of a sum of money to get the lithium. And in the history of commodity markets, when people point out there's going to be a huge supply problem in five years, that generally doesn't end up being a supply problem.
We were supposed to run out of oil 10 years ago, and then... And so, if you look, I think you can review any commodity supercycle, And people always say, "Oh, we're going to run out of copper , we're going to run out of reserves regarding production." And copper has basically been like that for 40 years, even as we've continued to use it, right?
"So it's like you're always, you know, always increasing stockpiles and production, but also increasing stockpiles proportionally. So, um, it's just not, um, in commodities it's not, you know, where you have a clear price signal , the supply should appear. Are we going to war with China? I hope not.
Yeah. I mean, I think, I don't know. This is probably way out, actually. It's probably so outside my area of expertise. I'll just be like, very fair. Okay. I do n't know. So you also mostly steer clear of Russia and Ukraine. I mean, yeah. I think there's a reason why in our tech forecasts we don't rely on political developments, and that's because these are very idiosyncratic and strange, and they depend on who's in the room, and I think there are things that you just can't forecast, and it's better to just accept that that will be noise in the forecast rather than trying to anticipate it.
So that's something that, obviously, isn't financial advice for the viewer, but it is something to..." As an investor, you have to consider that there's a lot of political noise, geopolitical noise, election cycles, the debt ceiling . These are things you don't even really pay attention to. You focus on innovation and whether it takes an extra year or not because of politics or this or that.
We look at the end result, which is that innovation , and that's where nobody else is looking. Would that be a correct conclusion? Yes, I mean, I think you can, in a very broad sense, not get caught up in the politics of things. For example, take lithium: part of the reason we think lithium supply isn't a big issue is because once you decide to start building a mine, you can do it in, you know, two years, but then it can take eight years to get the permits and approvals.
So the fact that politicians in many countries are really keen to allow electric vehicles to be built probably allows that process to happen more quickly, to shorten it. I mean, we're seeing it with Europe, China, and the United States investing hundreds of billions of dollars in battery storage and electric vehicles.
Exactly. And I would never want to make a forecast based on the style where people say, "Oh, there are so many subsidies for electric vehicles here, and this country has this target," because that's not how economics works. It's really the market that, you know, determines the equilibrium price for these vehicles, and there's a problem: if you subsidize an exponential technology, the political cost of that subsidy also grows exponentially.
And then, it becomes very politically costly to keep subsidizing it. That's why we simply try to avoid taking that into account. But I think it's incredibly challenging to do macropolitical trading because it's so volatile and unpredictable, and I'm sure there are people out there who do it well. It's just not what we focus on.
If you think, for example, about the role we play in the market, it's to, uh, demonstrate and support, uh, technology exposures that, uh, we believe are materially undervalued, and then expose clients to them. And it's not , you know, if clients want to try to prepare for political headwinds elsewhere.
That's something they need to do outside of what we do, you know what I mean? Yes, uh, um, battery technology. I know you guys are all big fans. I haven't seen NFACE emerge. Any interest at all in that space and in solar power? I mean, I actually asked you about that last time. Yes. I'm not sure. I mean, like I said, I think, or maybe I said last time.
It's, I think, and this is also true for Tesla's, um, solar division. Their solar division. Yes, their solar division. We haven't, um, incorporated that into our model in a deep way. And it's partly because an electron is an electron, so it's difficult, I think it's difficult, uh, to differentiate ourselves over reasonable time horizons and get a good margin out of it.
I think that's going to get compressed, in other words. So the margins could be very high on those home batteries today. Maybe, I mean, some of these margins on these home batteries are 50% today, but your Would the expectation be that that's going to disappear? Well, yes. I mean, I think, you know, how is it possible that, um, Tesla has a sales advantage in that they're already selling a vehicle to someone's home, and then that person has to deal with an electrician anyway if they want to install a charger for their car, so then they'll buy some kind of Tesla battery solar system.
In addition to that, I think Tesla would argue that, you know, we can sell more efficiently to utility companies because they need a lot of data that these systems work, and they want to, once, you know, they, the utility companies, are, um, less cost-sensitive than reputation-sensitive, and they know that mega-pack systems work, so Tesla will have a lasting advantage on that basis.
Um, and it's possible. It's just not an area that, um, that I'm confident will necessarily build up a lot of cash flow over time relative to, for Tesla, relative to electric vehicles and the robotaxi, where I think it's very easy to underwrite. Oh, and then there's the energy aspect. Residential solar is, um, you know, it's so sales-intensive; you know, convincing people to install solar systems is very, very hard work, and it's a very complicated sale, so...
Yeah. Yeah, yeah, it makes a lot of sense that it's quite sales-intensive. Oh my God . I think I see a Tesla or a solar panel. Get your solar energy for free. Click here. Right. Right. Right. Do they...? I mean, they used to literally go door to door trying to convince people to put solar panels on their roofs, and that's, well, I mean, it's hard, it's hard.
Yeah. Okay. So, software is obviously a big exposure for ARC. A lot of people have had the belief that, well, we're going through layoffs in the tech sector. Does n't that mean growth rates have now compressed? Is all that just more noise in the analysis of ARC? I think the market pressure that's been put on these companies has, in a way, made it easier for them to raise rounds of Layoffs they probably wanted to make anyway.
I mean, if you have a workforce that you've expanded a lot and some of those hires were mistakes, the market gives you an excuse to say, "Hey, let's review this and get rid of the bad hires." And that's not all, and I know I'm sure some of these employees are extremely good, but if you're heavily reliant on stock compensation and their stock has dropped a lot, that makes that expenditure much more expensive.
And so, you know, these companies have become cost-sensitive to a certain extent. I don't think it's really indicative of future earnings. I think, um, the idea that the technology was a passing fad that only happened because of COVID is fiction. And I think, you know, is it more or less likely that people will consume less oil in the future?
I think people are going to consume less oil. Are people going to buy more electric cars or fewer? I think they're going to buy more. And it's pretty obvious. It's not like there isn't that simple idea. Yes. Yes. No There is, there isn't , you know, it's like, um, you know, will AI perform better or worse next year?
It's going to perform better. There are going to be like multiple Bing bots arguing with each other on our behalf to try to resolve an insurance claim, right? Wow. I mean , I think part of what's going to happen with AI is that there are a lot of business and everyday life processes based on natural language generation , like written or spoken language, being a proof-of-work demonstration.
For example, let's say I want to cancel my cable bill, right? They might require me to call them to say so, right? Why would they do that? Because they want to say, "Hey, we'll give you a discount; "We'll give you a 15% discount." Sales, right? You hit cancel and go to sales. Now, what if I could press a button and it was an AI bot acting on my behalf as an annoyed customer who's going to cancel the service?
And if they still have a traditional customer service representative trying to offer me a discount, their systems will be overwhelmed with people asking for that discount because it's free or the cost is much lower for the person to do it. And so, they'll have to respond. Well, yes. So, they'll have to respond by making their customer-facing agents AI, too.
And then, the best AI agent will get the best discount. So, there will be an incentive to have a better, more persuasive AI agent who can credibly threaten to leave. And I think you'll end up with a lot of companies that... well, there's the case of the company that doesn't realize what's going on, and they'll just be totally overwhelmed and swamped, and their whole business model will end up destroyed because their costs will go up, and they'll end up giving all these discounts just to try to retain customers.
to the customers. Or, alternatively, both sides will invest in AI, and there will be a net shift of dollars toward AI software companies at the expense of the margins of the business that was counting on customers not being able to, you know, basically haggle with them. Well, yes, absolutely; I mean , right now someone who's an Adobe subscriber can log into their account, click cancel, and the next window is, "How about a 30% discount?" " Yes.
If I could just pull out my phone and press a button and say, 'Do that with every company I pay.' Right. Right. Just press a button, right? And not only that, you know, it's like some newspapers—literally, to unsubscribe, you have to call them, right? And so, basically, companies that rely on legacy systems as a competitive advantage in some way, or a mechanism to prevent customers from unsubscribing, I think they're going to find themselves inundated, because part of the power of artificial intelligence and these generally helpful AI agents is that they can connect to legacy infrastructures just like they 're human, right?
And so it means that a lot of companies and business models that have been immune to disruption may end up succumbing to it in this business cycle. Lemonade, when they went public, were very excited about their chatbot and their AI for assessing claims. Is Lemonade an interesting company for ARC? That's a good question.
I mean , we don't have it, or at least not in the core strategy. And, you know, I think the idea that you can Making better business underwriting or better financial business cost recovery using AI is likely, but I'm not sure that's what makes you win or lose the business, in the insurance sector and in the lending sector.
Um, it 's hard to say you're doing better underwriting until you go through something like a credit event or the actual defaults in the future. These are AI-driven underwritings, and that could take years. Yes. Yes. Exactly. And so, you know, the problem from an investor's perspective is evaluating: Is this company doing good business or not?
And you can't know, I mean, you don't know yet. They grow very fast based on how much they're paying for customers. And maybe they're overpaying for customers because they have their underwriting models wrong, so you don't know for sure. Okay. Right. And which could be like OpenDoor, which underwrote a lot of real estate in the first quarter and completely lost sight of what was coming in 2022.
Right. Right. And, And I think that's the way it is. I suspect the way this optimizes customer acquisition is what good financial companies should be doing in the modern age. And we're interested in betting on digital wallets, in part because if you have a dashboard for your financial life on your phone , it serves as an incredible lead generation engine for many things that can be connected.
Absolutely. Last topic: blockchain, cryptocurrencies , are stablecoins dead for good? No, I think there's a very Western view of public blockchains, as if we think we already have a good enough financial infrastructure. That this is just something fun to play around with, and then FedNow comes along.
Yes. And then, who really needs these things? But if you look at emerging markets, the use of Tether, for example, is extremely useful for people who want exposure to the dollar and can't access it in foreign markets. Yes. They know it's not a great counterparty, and frankly, I do n't think Tether is a great counterparty either.
I mean, who knows what 's going on? Behind the scenes at Tether? Right? Where are the reserves? Exactly. Sure. I'm guessing they have some subset of reserves, and for someone in an emerging market, a synthetic exposure to the dollar with a bad counterparty could be a lot better than their local currency.
Sure. I mean, with the Turkish lira , for example, with 80% inflation, it's almost better to have Tether pegged to the dollar. Exactly. Exactly. And it's the same with public blockchains. I think , with Bitcoin's value proposition, this has solidified. It hasn't weakened with the FTX breakup. It's basically, the problem wasn't the decentralized protocol.
It was the centralized counterparty, right? It was kind of an affinity scam, where the idea of a public blockchain is that you don't need to worry about counterparty risk. So people were like, " Oh, I don't need to worry about counterparty risk." But it turned out that the fact that this entity was foreign meant you had to worry more about counterparty risk, not less, right?
And I think , somehow In that way, the value proposition of public blockchains has become clearer. Well, and actually, when we were putting together our Big Ideas report, there was a lot of debate about what we should do with our forecasts on cryptocurrencies and smart contracts. And you guys got a lot of press about that.
Well, but I mean, you don't have to look at it that way. Even with a million dollars per Bitcoin, we think that's a single-digit penetration of the total addressable market by 2030 relative to gold, of course. Well, relative. So, it's taking some of the gold, but basically what Bitcoin—and Ethereum to some extent—is competing with is M2, the total amount of the money supply.
And M2 is basically a function of GDP. And so, you know, we think GDP is actually higher than people expect it to be . And having a currency that isn't fragmented is an incredibly powerful economic force. I don't think people understand how big the economic shift would be if we all operated with the same currency.
Or if marginal economic activity started to to gravitate toward a single currency that can't be manipulated in favor of those already in power. Right. And then, you know, none of that has changed. And I think there were a lot of people who were very pro-Bitcoin, Ethereum, and blockchains in general who were wiped out by FTX in particular.
A lot of them were taken by surprise. Oh, yes. And my point is that they thought they had exposure to Bitcoin. It turned out they didn't, because it had been sold off without their knowledge. They were promised exposure to Bitcoin, and now they didn't have it. So, one, that's extremely painful for those who lost money, but two, it means there are people who are still believers and who have less exposure than they'd like.
Right. Right. Because there were a lot of losses. So it's not that they've stopped believing in it. It's just that they've been taken out of the game. So, those people...? I think it will be painful, but they'll be back at the table. In a way, they've demonstrated the utility of the technology in a profound and painful way, and they've brought a lot of people who were trying to take advantage of it back to square one.
From the outset, so they'll have to start all over again. I think there will be another adoption cycle in cryptocurrencies. Any concerns about how, for example, with Ethereum, the Tornado Cashes of the world might actually obscure the transparency that blockchain promises and lead to more regulatory crackdowns?
Well, I think actually—and this goes toward the question of stablecoins or central bank digital currency—I do think there is, um, a potentially profound danger in, uh, in, in basically, uh, preventing people from transacting privately. And I think the problem with a digital ledger currency is that it allows anyone to see any kind of exchange of information that takes place.
And that's what a transaction is. It's an exchange of information. And I think if there's somebody in power who can prevent that exchange of information from happening, then that's, you know, a decidedly bad thing for the world. I would argue that using the dollar as a mechanism for enforcing rules and conduct isn't clearly a good cost-benefit in terms of the economic damage it causes relative to what the US achieves with its strategy.
and geopolitical objectives. Wow. No, maybe I'm not sure there are people who vehemently disagree with that. But it seems to me that the way the dollar system is used to try to prevent behavior tends to hurt the less fortunate people in the sanctioned countries, while motivated cheaters can circumvent it through the Cayman Islands or wherever.
So what you 're really doing is punishing populations to try to get them to overthrow their governments. And frankly, I don't think that's right. That's why I think it's a very dangerous precedent to say that this particular software can't be used, and I don't think that will hold, in part because these technologies are openly available to anyone, anywhere, and that's the value proposition.
You can't stop someone from using Bitcoin because it's an open protocol that allows anyone to integrate. But is it a financial security? Well, Bitcoin isn't. They've already said that. But, well, I guess the verdict on Ethereum is still pending. They would also say it's a commodity. That's incredible.
This has been absolutely Incredible. Anything else you'd like to add? Uh, no. I mean, I think there's never been a more exciting time in technology than this. I think all sorts of things are going to happen, but I'll mention a few. One, Google will say, "You know what?" Actually , we didn't want that search traffic going to the big language models because it wasn't profitable anyway, and we don't understand how they're becoming profitable.
It's good that it's moving away from us because this traffic wasn't needed and it was causing reputational risks. But that will be a sign that they're really hooked by the innovators. They're becoming the legacy. Yes. And I think there will be... You know, in business model transitions like the one happening with AI, there will be new, huge companies that will accumulate monumental amounts of value.
And if you look at the history of technology, it's usually a much bigger leap. So we'll see a ten-trillion-dollar company. Or maybe even more, perhaps a twenty-trillion-dollar one, emerge because of AI. And when you still have exposures to innovation trading at a discount in the middle of a tech boom, the right thing to do is to get exposure to more innovation.
We believe that more than half of the global stock market capitalization will be exposed to the disruptive innovation platforms we're focusing on during this economic cycle. And it's naturally volatile. There will be a lot of ups and downs along the way. And I think I'm more confident in that assessment than ever, because the pace of change is mind-blowing right now.
I mean, right now, as I see how strange everything is. I'm just saying, look, when McDonald's beats innovation, it's probably time to buy innovation. Yeah. Well, it's like, you know, McDonald's is a big business. It's just the franchise. Brilliant. It's just a different business. And I think the cash flow that will be generated from the technology...
In the end, people will pay for that cash flow. Uh, and well, I think these markets, literally, at least as we model them, are measured in trillions of dollars. This has been fascinating. Thank you so much for your time. Sure. Thanks for having me. aryeninvest.com. Right. Yes. There's arc-invest.com to learn more.
Thank you so much for watching. If you liked it, please share the video, and I'll see you in the next one. Thanks again. Cheers. My pleasure.
Comments 0