Hey 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. actually me coming now here, but coming back on the show. last time we did this over stream yard and now we're in person in Santa Monica Thank you so much for inviting me out here! Ah, thanks for coming, absolutely for having me! so.

we wanted to talk about a lot of things, but I think the big thing that really got us interested in catching up again was chat GPT This is what everyone is going crazy over right now. People are jailbreaking chat GPT and saying it has a leftist biased you've got Bing having emotional depression apparently because it's language learning algorithm is slightly different than that from chat GPT What's your take on this and is this just another metaverse fad? or or is this actually a transitionary moment I Think it's a true inflection point in artificial intelligence. We think that artificial intelligence systems the cost to train AI systems is falling 3x per year and will continue to go through. Do so through the course of the decade and then you combine that with Google and Microsoft and lots of startups and everybody realizing also lots of large Enterprises Hey, we can throw money at this and and generate remarkable capabilities and you get the likes of chat GPT Demonstrating to everybody these systems are massively performative and you end up in in a massive capability boom.

So if you look 2021 to 2022, we think the aggregate capability of AI systems increase tenfold this year versus 2022. So Chachi PT was released at the end of 2022. We think there's going to be another seven-fold increase in the aggregate capability of AI systems this year and 23. N23 and that'll continue.

You know over the course of this business cycle every year in our view. And so what that means is by 2025, like whatever the state of the world and AI is today, you will have a multi-thousand fold Improvement by 2025, and so Um or 2026.. And so one way to think about that is is the Um. today's iPhone the one that's probably in your pocket is Uh is on a CPU basis.

It's 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 over two and a half or three years. Oh that, because that's 2007. That's 16 years and you're saying 16 years of technological advancement potentially coming in three years now exactly with AI And so if if you like, think of the original iPhone for you and how performant that was, that's chat GPT And now you're gonna three years forward the equivalent capability Advance would be upgrading from your original iPhone which is chat GPT to the you know iPhone with the magic camera that takes perfect pictures and it you know has the App Store and has everything built on top the iPhone one.

It's weird because now it's been 16 years. a lot of people don't even know the iPhone one. Didn't have an app store right? That didn't come out until the iPhone 3G which was the second one. And so you're saying an iPhone without an app store is basically chat Gpt1 and in three years we'll be at the iPhone What are we on now 14 Pro: Max Yeah, Exactly exactly.

And so I think it's hard to I use that as a comparison because it's really hard to conceptualize what is a a multi 100 folder or thousand fold like. yeah. and I Think the way it's going to manifest is that every single business Enterprise is either going to adapt to AI or they're going to die. Oh my Gosh.

Wow. So um, we just heard today that the CEO of Google stepped down. We've seen their advertising revenues decline YouTube the CEO of YouTube Uh, right. Correct.

I'm sorry, not the CEO of Google. Thank you for correcting me. Yeah, so the CEO of YouTube step down. We've actually seen YouTube advertising slow down.

We've also seen Google search traffic slow down, but still positive year over year while YouTube's down. Apparently some folks are saying that might be because FTX wasn't paying some of its advertising bills. We'll see. it's kind of remarkable.

Yeah, uh, but uh, what's your take about the argument that chat GPT if it works so well and this Ai and search works so well, the amount of search queries that will happen will plummet and Google's advertising business will implode or suffer substantial deflation. I I Think that the way to think about it is there. There's kind of four stages of realization that are going to occur. First, some people are still in the uh, this doesn't matter to Google stage I Think that's wrong Yes, uh and I think it may be is profoundly wrong.

Uh, the next stage is okay. Well, Google has great AI capability, which they do. so they can run a language model just like Microsoft is running and use that to answer search queries. And that's good.

except that Google makes maybe a penny and a half of Revenue per search. And and you, uh, allocate costs and they get maybe a half cent of profit per search and the cost to infer chat. GPT The cost to run that AI System seven-ish right? Well, it depends on how you run the estimates, but it's on the order of, you know, a penny, maybe a half penny. Whatever.

If Google had to run a large language model against every search result, that would take all of its profit away. Yeah, yeah, and then something like it it would, it would massively destroy the margins. I See, Okay, so on the scale of Google, they might be able to get it down to a half penny of cost for if you put AI basically in every search. but that's all they're making anyway.

Right now is the concern exactly. And so so the first. This is like the thing that people are being beginning to try to think about is that so if Microsoft puts competitive pressure on them by saying hey, look, Bingbot can do it searches really great and then Google has to respond to maintain share by doing the same. That's all well and good for Google except then their their costs Skyrocket relative to their revenues.

The the third stage of realization which I don't think people have really wrapped their heads around yet is that the way in which a search engine converts to a a large language model is it's no longer a search engine. instead it's an answer engine. If I'm getting an answer, I'm not clicking through to some other site to get the material. and hey, you know their entire advertising model is based on click-throughs to to search ads and so, uh, there's also a reasonable argument to be made that if you switch to large language models, you not only raise the cost per search, but you also hurt Revenue per service.

So you hurt the top line And so like, The next stage is, oh gosh, it's actually not only more costly to serve these results, but then the whole business model that you're basing this you know, 160 billion dollar per year near Monopoly is profoundly broken. and then the fourth stage of realization and I think this is the real way it'll play out Is that, uh, the the whole process by which we go from doing something to saying I'm going to move out of this experience as a consumer or an Enterprise worker or whoever go to a separate portal that will then deliver me to another place for information and that's what search is. It's basically like a Way Station delivering people on I Think that's all going to break if I I work in Excel A lot, right? And the my typical like workflow is hey, I need a piece of data I go to Google I search for the piece of data I go through to the primary sources I get the data. Maybe sometimes it's through like an academic site that's trying to charge me for a subscription.

so there's a sponsored link and I'm clicking through and that's shifts Revenue to Google Instead that asking for data is going to sit right inside of Excel. as in it's like I need CPI through 2017 from 1882 or whatever and I'll just ask it and it'll fill it in and so then I will never go to that way station. And so I think the final stage is like the text prompt. The large language model does not sit separate from everything else you're doing, but it's directly in the end application that you're working on.

And so then The Um. then that ultimately will feed through into Google's traffic. And and and you know, do we actually need in a separate search engine at all? Exactly. Particularly because the the answer that the language model will be able to provide will be um, much more powerful because it will have the context of the thing that you're working on right right? And so um, it's like not clear to me that the way in which you plan travel will require you to bounce into a search engine and instead it might be right in your calendar app.

Or it might be right in your Maps application, in your Tesla or in your Tesla right? And so it's like as you're coming down here um, you know it can directly from the Tesla interfaces where you figure out the res, the restaurant reservation, the hotel reservation. you know whatever kind of like set of tasks you're needing to find out. I Suppose in theory, we don't want a Google We want an answer. Exactly interesting.

So if we're searching for an answer on YouTube in theory, we don't want a video, we just want the answer right. You're bankrupting me with these ideas, sir. Actually, actually. so I think there's an interesting a Um probable byproduct of this is that kind of like the the information that's out there that these engines are trained on will actually get Amplified And so you know, maybe you're not famous enough for it to happen.

but for like you know, um for um, the Harry Potter Universe for example, like my kids use chat GPT to put themselves into scenes with Harry Potter and to like get little short stories of them dueling with him in Hogwarts and it's really good at that. It can generate like a scene that has dialogue between my six-year-old daughter and Harry Potter and her saving the day by curing him from some spell that's been cast on him. and it's entertaining and it and it does it within the context of who Harry Potter is and all of the kind of like it understands kind of the way in which that dialogue would work within that world and it's because there is a very um depthy text database of all of that information and so it means that existing IP owners, people who have intellectual property that already have like built up a world around it I think that's going to be Amplified We're never going to get out of Harry Potter because JK Rowling doesn't actually need to produce more material for more material to be produced. with that as the foundation sure sure take here, learn these books and manipulate stories.

However, anybody would prefer them to be manipulated, inject multiple children or whatever exactly and in the future that could be in theory video generation as well. Yeah, or video game generation VR Yeah, it's like you could imagine that the um, you know in these stories they're not as good as the actual books. but they're actually pretty pretty good. as in I read a short story to my daughter from it I'm like this is clever like it invents spells that are funny it you know it does or and maybe saying in Vince is the wrong like verb but but but but the you know it's compelling enough at least for her.

Yeah that it's um, it's new, at least as good as like uh the old school like bad you know made for TV sequels to movies that Disney used to produce. Now what's incredible about that is most of and this might be in your earlier stages of thought for for AI a lot of folks say oh, the way to protect yourself from AI is just be a creative but what you've just done is destroyed even that Well maybe yes unless you have to use a New Foundation yeah yeah I think I think I think that Certainly people you know for a long time were like oh gosh, artificial intelligence and automation is going to kill Blue Collar Industry right? right? Well it takes time. It seems like the thing that it's doing instead is is it's it's you know, impacting creatives and and software developers and and you know jobs that were uh, required a lot of technical skill and are very well compensated. Now our belief is the way it's going to manifest is it's going to make you know creators that are really creative, like more powerful The JK Rowlings? Basically, Sure, right? Sure, it's It's like a massive Force amplification.

Like look at what's going to happen in the software space if um, with GitHub Co-pilot you know they released a study that suggested they could make software developers 2x more productive just by using Copilot as it exists today. Sure, and every one of these systems is improving. so it's it's um, if you look across the entire knowledge workspace and the amount of productivity advanced, you think we think that AI software is going to deliver to knowledge workers. it could be the largest macroeconomic productivity advance of all time.

I Mean it's it's literally we spend I Think it's on the order of 50 trillion dollars a year on knowledge work salaries. So imagine you're providing all those knowledge workers with with software that makes them 2x more productive. That would be worth you know, 50 trillion dollars in output. So even if you're paying a discount for that, that we think people will, um, you know you end up with an AI software spend of 10 trillion dollars plus.

uh, that's incredible because of the just the velocity of money spent. You're talking five six hacks per dollar spent. If people can be so much more productive and can make more money, they can spend more money. GDP grows substantially faster than what's estimated today.

I Think you all actually have GDP estimates What? 2030 of in excess of uh of 100 trillion dollars for the United States Yeah, something like that. Well, so today globally there's roughly a hundred trillion dollars in GDP right? If you look at consensus for cast for United States was sort of me going through maybe 20 21 to 100.. Yeah, so consensus forecast globally is 130 trillion by 2030.. we think that um, just just based on um on Robo taxi And and what's going to happen with robots alone, We think you're going to get in excess of 170 trillion dollars in GDP by 2030..

So the difference is uh, real GDP per capita average of 15 000 is the consensus and we think north of twenty thousand dollars is likely. So everybody on Earth on average is getting a 30 raise relative to expectation because of the technologies that we're focused on. That leaves out the AI software side because it's less clear to me how that's going to reflect into actual GDP Okay, like AI software is going to enable Robo taxis that's going to be very well measured with traditional macroeconomics distribution. The fact that we're going to have a lot better software a couple years from now which I firmly believe it's not as clear that that that there will be GDP credit given for that because you also at ARC believe that ours worked will continue to decline so you have more time for family or whatever else.

Sure. and and you you can see you know it's I Think there's a lot of ways that consumer surplus is not very well measured in GDP Think about it this way. It's like if you if you decide you're no longer paying for cable TV and instead go with all the more than your average share of streaming services. then on a like um amount you're paying for entertainment basis, you might go from a hundred dollars plus to like 50 60.

Sure right? That's a terrible GDP event you've basically it's you've reduced GDP by forty dollars just by making that choice. Sure, but you're probably watching the same amount of entertainment. It's probably better. Yeah, So from a from a like what is your uh, experience of the economic world, it's better.

You're getting entertainment cheaper and it's just as high quality. In fact, probably more quality. You're not having to watch commercials. Sure, Um, but like the macro economic statistics would measure that as as a decline.

yeah in consumption. That's fascinating. Now when you mention somebody by 2030 their their income, everybody per capita, their race could be 30 today. Obviously, in the inflationary world, people don't feel that very good about potentially that kind of increase.

But I think the world that you're describing by 2030 is actually substantially deflationary. So a 30 increase in income can actually be worth substantially more than how do you address in the near term, the inflation that we're going through. What do you think I mean Today we had a sticky PPI report. We had a CPI report from January that was slightly strong.

We had a jobs report that was hot. A lot of folks saying that's January it's seasonal adjustment month. What's your take? Yeah, I mean I Think that you know technology routes around inflation and it'll continue to do so. Think about like even the electric vehicle space with your Tesla You know if someone buys a Model 3 relative to a Toyota Camry Yeah, they're paying maybe one and a half times more for that car up front, right? and then they're ongoing.

Spend every year is 0.7 times what it would be with the Toyota Camry So total cost of ownership is actually lower they're getting. They're getting on a dollar for mile basis. They're going from place to place, less expensively. Uh, in the way it manifests is pinned up front in exchange for Less spend later on.

and so that's it's it's. actually you know that read through into the inflation statistics is the Little Mix Like you know, they try to adjust for the fact that you're buying the higher price vehicle anticipated adjustment for for the lower maintenance you're going to pay for in the lower fuel costs over time. Uh, and so an example. and it's I Think that there's um.

You know, in some ways inflation is a function of The fact that people still have money that they can afford to spend if you're paying less for entertainment. if I cut the cord? Yeah, right. And and I'm paying less for entertainment, well, then I have additional money that I'm going to spend on, you know, steak instead of instead of rice. Yeah, right.

And so um, you know there's some relative price adjustments and ultimately I think the the deflationary forces of Technology are are going to carry down kind of the cost adjustments. So it's clear inflation should go away in the long term because of the deflationary aspects of of Uh technology is what you're saying. But in the short term hey, people still have a lot of money to spend is I think Bank of America actually just put out a piece. They said the average person before the pandemic who had two and a half to five thousand dollars in their bank account now has 12.8 000, right? And that's only declined 4.4 percent over the last 12 months.

Yeah, so you're not seeing that excess spend go. What people feel wealthier, they feel like they can keep spending. I think American Express was talking about people are spending through the recession or whatever we're going through is is is that sort of your view that yeah, hey, in the short term here you can have that inflationary impulse, but technology is very clearly going to push us into a longer term deflationary trend. Uh, and I I Also think there's some strange mixed shifts going on as in, like imagine within the housing sector which is clearly under given what's happening with interest rates it should be.

um, in in severe distress doesn't seem like it is I think partly it's because that you know people who are still working from home basically need more space and so like what that means is. But meanwhile you have commercial real estate which is actually in quite a bit of trouble. You had two big buildings downtown LA that just got turned back to the bank that they they just finished them up and they're like we're not gonna be able to lease these out, we're giving up on the loan and we're turning them back. He was in Bloomberg and said we're giving the keys back to the bank.

Yeah, I'm like it's not that easy. but okay, like the reason you can't lease them is because the business is basically, um, pushed. kind of like that the office real estate cost partly on to their workers. So what does that mean it means? Well, it no longer.

Falls Onto the businesses cost line. So the business looks like it's higher margin. um but the the workers sitting there being like well now I have to pay for an extra room so they need to find an extra room. Then they're going back to the business like well you know you got to pay me more because I'm paying for this.

So so it actually shifts both the cost and then um, the cost into the residential real estate. and then the consumer wins some of that back with with Um with wages? Yeah, but but is that you know that's not inflationary? It's actually just a shift in in where on? Kind of like the corporate consumer balance sheet. uh, the spend is occurring. Uh, and so like I I Still think we're essentially processing some of those changes in behavior.

Um, I Mean you're really? You've really made an argument here that some of the larger increases in wages that we've seen is potentially not actually real inflation at all. It's just moving spend from the company to the individual. Yeah, that's incredible. Yeah, yeah, and but but traditional data surveys can't measure that because this.

this hasn't happened before in technology and the ability to work from home after pandemic, right? right? So it and I think there's a lot of you know, there's a lot of stuff. Certainly, as a response to the pandemic, but also on the technology side as well, there are. As technology accelerates, the macroeconomic statisticians have a harder time actually capturing it into GDP in a way that Um is accurately reflective of reality. So um, for instance, even your smartphone.

It took them a decade before they actually, um, um, realized they needed to to specifically deflate for performance improvements in smartphones. So there was. There's you know, imagine you spend, you know, 100 on a phone every year and then suddenly you go from buying you know, a Nokia phone to an iPhone and so you're spending 4X more for the phone. Uh, that for a decade was counted as, um, an inflationary pressure.

Ah, right. Because the in in the way it manifests is like the cost of telephones was coming down. Sure, the whole time, it's just the cost stopped coming down quite so much. it kind of like flattened out.

Yes, it flattened out. Not because the costs weren't coming down anymore, but because people were buying like a much more Complete product at a higher price. Yeah, now they're buying a video camera and a phone. And why, Exactly how do you adjust? Yeah, and so there was no real attempt to make an adjustment for that until the iPhone 10 came out, which was a much higher price point.

And they're like, oh gosh, we really need to adjust for it. So right, Because the ATT, T-Mobiles or whatever. they were subsidizing a lot of these with the monthly plan. So initially, like you said, you were going from 100 bucks to 400.

But now you're paying a thousand to fifteen hundred dollars for a phone, right? right? And it's a phone, that's you know, massively more performative. It's certainly more performative than the like rotary phone. oh my gosh, beforehand, right? and so. But But they, for A decade of smartphone use, they were using that same cost decline series that was applying to the rotary phone.

so it looks inflationary. When the reality is, with all of the technological Improvement you're getting, it's substantially deflationary Exactly exactly. And so then going back to AI it's like I'm not confident. you know it'll probably take a decade before the macroeconomic statisticians go back and be like oh, we really need to reevaluate, like what exactly is going on here, like how big the economy is.

Um, but I Do think that like one of the you know, one of the things we focused on over the course of the past year is um, how Technologies are converging and and how you have essentially uh, an acceleration one technology is enabling and acceleration in another and I am confident that the acceleration the AI that's happening Now it's not just going to be limited to like a chat bot that is, you know, arguing with you on Bing right? Healthcare it's uh Vehicles sure, whatever. Yeah, it's like like look at you know Tesla's advance and full self-driving during AI day. The most recent AI day they their ability to navigate Lane intersections is based on the Transformers architecture that was originally like introduced just to make language translation better as in it was in advance in kind of like translating you know I don't know uh Dutch to German uh in a general way turns out that that Advance applies directly to how a car can understand an intersection that it's never seen before. That's remarkable.

What's your take on Uh Tesla Obviously with Uh with Elon Musk's acquisition of Twitter, there's been a lot of rumoring that ah, maybe he can take AI technology from Tesla and sort of bring that over to to Twitter and maybe even create a competitor. some people have rumor to chat Jpt. Do you see that as an ambition of Musks to get into more of that neural AI which then folks tie into the future? Tesla Bot I mean I I Think that it's clear that Twitter has a unique data set and should be able to um develop its own language models that should be quite powerful. Um, one of the I think the the character limit on tweets actually may have unintentionally made their data set more powerful than it would otherwise be.

Really so shorter being better. You're saying because because it forces people to basically compart to quantif quantize their ideas. like if I send you a block of text right and you are a machine and you need to like break it into unique ideas that might be actually hard. Yeah, and um, because people can then amplify the individual ideas.

Um, it means that they they have just a unique data set. Yeah, right right Because if you have a thread, uh, a computer can now see which part of that thread is being most Amplified as opposed to here's the complete block, you figure out what's the most Salient computer Yeah, Exactly exactly. And so and so kind of the fact that there's metadata at the idea level metadata meaning retweets and likes. Sure, Um, and and that essentially the human work of breaking up larger ideas into not just sentences, but to whatever is the 280 character like maximum like, uh, call it, call it minimum viable idea means that they have a really interesting data set right? and so.

but I don't think you know it's you know the it's not the like. it's in advance that's in Tesla that's going to go inside Twitter Instead, it's like I think Twitter will, um, try to utilize all of the Uh language advances that are being published on and attended the likes of Open AI are using I Do think that um, on the robotic side, You know it's really fascinating the degree to which actually using these language models can enable robots to be more performative. And now Tesla purchased their own robotics facility from Germany uh, their technology. Uh, the German company in 2016 I Believe it was.

uh, so obviously we know they're they've got the autonomy. They've got robotics. What? What do you see for their advantages? Since we've last spoken over, let's say a Ford where for example, you just had Mr Farley the CEO talking about. Oh, we realized we had an extra mile of cabling in the car and and here's everyone saying oh Ford's the best at manufacturing.

They've done it for a hundred years. Meanwhile, what the lightning manufacturing? Line's been closed now for 3 3-4 days. Expected to be down for another week on battery issues. What's your take on really? Legacy Auto's ability to to be able to compete at all in CV space I I Think it's unclear that they can I mean I think Tesla has just on a drivetrain efficiency point of view.

I think a four-year lead, three or four year lead uh into the that's just on manufacturing. Yeah, it's basically just like um how performative uh drive train can they make per dollar? uh drive train inclusive to the battery system? Uh and and what that me? I think that the challenge for Legacy auto is they still it seems to me are on like a four or five year kind of like design to um kind of production. Innovation Cycles So and they they're stuck in a very like root nice kind of half decade Cadence of like this is how we game plan and then launch in advance. Uh and and it seems as if they um you know it's like once they get a year down the track, they're kind of like tracked into where they're going to hit.

Uh and so then the problem is in in uh you kind of saw this with the model 3. it's like Tesla launched the model 3 in 2017. it took all of like a Seattle five years to to release models being released this year that are roughly as performative as that 2017 model 3.. right? And so you know in March 1st I think Tesla's going to announce a vehicle.

That'll you know they'll claim they're going to be able to produce for half the cost. You think they're gonna announce a car? Oh yeah, yeah yeah yeah. I mean that they've They've definitely indicated that they're likely to. Uh, so I think they'll It'll be a new model that um, you know will be um, very economical and tuned to be a robo taxi probably? Uh, and that, um, you know it, it'll It'll just like set the entire industry off sites on a cost basis Again, as in you know now kind of like industry is trying to like design.

how are we going to scale and and you know they're just I I Just think that they're designing for a cost of manufacture that is going to not have them able to profitably compete with where Tesla's at. but Charlie Munger says Byd is better I mean I Think that for one thing the Chinese market is is like a very different Beast than the U.S market and I do average vehicles with 13 14 000 there. our average use a new car price 47 000 somewhere around yeah I I Well I don't know about the Chinese Average but but I There is like a whole class of vehicles that's more like a neighborhood electric vehicle, so it depends on how many of those you bundle into how you do the averaging. Um, but you know in net I think within China It's more likely that Tesla sits at a premium tier and is not you know, competitive with with real mass volume manufacturers hitting the Chinese market that we'll see.

Um, but I I Think that there's like. It's very clear to us that that, um, almost all of the auto market is going to become electric drivetrain based. Uh, you know we think within five years it'll be 90 plus percent of the auto market. We also think the auto market is going to be smaller because Robo taxis will take up more miles than traditional Vehicles So there's going to be a compression and Top Line Sales but you end up with like 60 million EVS sold in 2027..

Uh, and and we think you know Tesla's worldwide. That is worldwide. Yeah, uh. Tesla is going to command a plurality of that.

You know, some 20 million vehicles or so. Uh, and um, it's hard for me to see how at least a legacy automaker is able to successfully scale and compete. And then the startups, which I think designing an EV from the ground up is is like a better way to approach the problem. There's a reason that Ford ended up with way too much harness in its vehicle like you said, and and it's because you know they didn't probably didn't take a first principles approach to design and um, yeah, I I Just think that there's like huge problems in In Legacy automakers.

Um, even leaving aside the pressure they'll face from Robo taxi and um, you know the fact that they have a lot of auto loans on their books that probably aren't going to look very good over the next few years. Yeah, did you I imagine you did. But I thought it was so ironic maybe unironic actually that a month ago Mr Toyota the CEO of Toyota was at a conference. uh, talking about how he thinks he's part of the silent majority that people want hybrids and so they're going to double down on hybrids and then within 30 days we hear he's getting kicked out new CEO is coming in who's going to dramatically transform? Toyota So here we are in 2023 and only by April or they're going to have a new CEO who's just now going to start their pure Electric Division What's your response to That is just is that just an expected transition? I mean I Our our perspective has always been that um, electric vehicles just make more sense.

They're more performative, there's less complexity. I mean with a hybrid, you're really having to almost build two simultaneous Drive trains and so um, it. and I'm sympathetic to the idea. I Think the reason people think hybrids are attractive is because they say, listen, the average person only drives 30 miles in a day anyway, right? So why do they need a battery that's 250 miles If they can just drive 30 miles, you know, using electric and then anything incremental to that, they can just use a gas engine, but then you introduce like much more engineering complexity.

There's also like for plug-in hybrids, people don't plug them in because because it's like, well, I'll just drive it anyway and so then it. You never actually stay in the electric mode for as long as people underwrite too. Um, and you introduce a lot more weight to the vehicle and you sacrifice performance. And ultimately, um, and in addition, you, you end up with a vehicle that's not as safe because you still have a big internal combustion engine in the front of the car to get shoved into you in a in a head-on collision.

and so you end up with a left safe vehicle that's less performant and uh, and you know, more expensive. and I I just don't think a consumer will ever optimize to it. It's hard for me to see how that would be the dominant mode. Um, no.

I've been under this impression that we could probably get down to and I've mentioned this a few times on. YouTube There are always people who say no way, it's never going to happen, but I want to hear your opinion on it. My idea is that we probably only really need a a 20 kilowatt battery for. for most people, that now brings down the weight of the vehicle because you're carrying around a battery that's one-third the weight.

Uh, certainly one-third the cost. maybe even less. And so it has an 80 to 100 mile range that's more than enough for your daily commute going to the grocery store as you mentioned, probably not. I mean it's like I think that was range anxiety is not rational as in as in um I think even the early EVS like the GM bolt they were kind of.

that was their idea was hey, we'll deliver a vehicle that doesn't have. You're not going to go on a road trip with it, but it'll be your electric around town car right? I think for the U.S buyer in particular: when somebody buys a car, they in their head say what is the most exciting thing I'm going to do with the car this year and it needs to meet that need. So and they never do it anyway. Yeah, and then they they might.

It might even be over four years. It's like, why do people buy pickup trucks? It's not because they're hauling things around every day, it's because it's like they will feel really good when they need to dump a mattress in the back and they're able to do so and so. I think when you're making a big purchase like that, it's you like purchase for that that tail event. Um, and and that's just the reality they consumers in America that that might be.

That's me indexing to the American buyer because I'm more familiar with that. Yeah, and I think the neighborhood electric vehicle segment is really interesting and hasn't been penetrated with a compelling product. And it could be interesting. as in 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 do a car, why not do like a 15 000 Apple car that was an around town car. Yeah, or twenty thousand dollars either even. And you have like a set of buyers who would probably be like oh, that's it if it looked cool.

Yeah, yeah. like you would massively open up your design engineering space because you're no longer designing it to hurdle along the highway, right? 70 miles per hour, right? And so it could be like a really interesting product. Now on the robo taxi side. you know the um, there, because it's driving a bunch of people around.

You still need a relatively big battery because they'll probably have to go 200 or 300 miles. maybe even a day, right? And so then it then it's like you actually don't get down to that small battery size. That's interesting. So your thought is, uh, this next vehicle that Tesla might announce wouldn't necessarily be something smaller battery.

It's Robo taxi. It's the direction I think so and I think I Mean if I had to guess I I think that there's a optimization they could do in providing a robo taxi that at least has two or maybe four separate passenger compartments that are independent of each other. As in, oh wow. So part of the problem with Robo taxi in our view is that you will, um, uh, you'll massively increase the number of miles that people drive on roads.

Okay, because it's going to be much less expensive and much less convenient to ride around in a car. Sure, Uh, and so it's like I don't go to downtown L.A because it's like an hour and 15 minute schlog and it's you know, expensive and an Uber and driving myself. It's like, you know, exhausting, exhausting. Uh, but if instead of costing three dollars a mile, it costs 30 cents a mile.

Yeah, uh, and I'm just riding in the back seat. Not only will I do it, everybody will do it. And so traffic will get worse because because the pain of sitting in traffic isn't as high and it's not as costly to go from place to place. Sure.

Okay, and and so then, like you know, Uber had this great idea. They're like we'll do Uber pool people along the route to ride with other people along the Route and um, but that's weird people. Yes, people didn't use that because then you're like sitting next to a stranger, right? you know? and so, uh, you definitely won't want to do that in a robo taxi because in a robo taxi, you're either by yourself, you don't even have the Uber driver, right? You're just by yourself or with a stranger. No way.

You choose the stranger to save yourself. Like five cents a mile? Yeah, right. But if you can create a robo taxi that has like separate compartments that open up and it's like your own little pod. and since it's like you don't actually care how long it takes anyway because you're just, you know, whatever.

watch Netflix Anyway, then you could have something that enabled that kind of pooled concept to work. That's interesting. They might do that because they also have some um FCC patents that, but in my opinion are starting to suggest maybe they're trying to get Starling into cars that could potentially align with this idea. Who knows.

Sure. I mean I think that Starlink on all I Mean you know they already have Starlink for RVs I think I Think that um, you know and this is Sam's work. Who covers reusable rockets for us? But uh, we think there's you know, tens of billions of dollar opportunity and attaching star links to moving our low earth orbit satellite constellation comps to moving Vehicles cars, planes, RVs boats, whatever. Yeah, maybe you? Yeah, it's I Think there's a like, then you're weighing like, um, yes, the cost of the antenna versus like the utility you get uh, out of having it on the vehicle but it seems like um, if they're doing something like um, uh, enabling people like connectivity in their ride.

Absolutely. So that's actually interesting. So initially because you all have a very strong bull thesis on uh, self-driving Auto taxi Um, the um I think your your Tesla projections almost exclusively in your bowl case, give most of your weight to the taxis. It's like two two-thirds of the value of Tesla from five years five years from now is in our view, driven by Robo taxi versus at the 132 electric vehicles.

Um, and but but I I suspect the product they'll they'll release will be A. This is an inexpensive vehicle platform to manufacture that's optimized for adaptation to Robo taxi, but also we can sell it into people who you know are the um, kind of like median car buyer, not just the average car buyer in the US which is a lower price point so that would be the 25-ish thousand dollar car plus maybe some inflation since then, right right? And yeah, don't know what the price point would be, but think about, you know, half the half the cost to produce of the model 3. And so if the average model 3 is selling for 50, yeah, maybe 25 or 30.. And why do you think they would announce that now? Why? Because I think that they are gearing up to manufacture it because they are even before cyber and semi? Well, no right.

I Think they as in, they'll announce that they're doing it. They'll reveal just like they did with the Cyber truck. they'll reveal what it looks like and then they'll go through kind of engineering for manufacturing so that'll take a year or two. Yeah, yeah.

Wow. Wow. Okay, so uh, all in on Tesla except when Jerome Powell gets angry. huh? Well, I mean listen, across the board, Innovation stocks have underperformed measured on a year-over-year basis.

There's no question about that. I Think that the Um. What's really fascinating for me is that the the rate of technological Improvement across the technology platforms that we focus on has never been faster like in at least in my career of of looking at technology and markets. It's uh, you know, in the AI space, it's gone from um I was having a hard time keeping up with the academic papers to then having a hard time keeping up with the product launches.

Yeah, yeah, right. Uh and and so like the rate and it's not like the academic papers have slowed down. You know, if anything, it it's kind of having these tools to play with. Um then enables people to do more experimentation and Discovery and then that's feeding directly into all of the other technologies that we focus on.

What do you think about? Uh, it is. I Don't believe you all invest too heavily in chip manufacturing? More software driven? What's your take on that? Are you staying away from labor intensive and and commodities intensive intentionally? Or is there a different reason? Maybe for example, even an Asml equipment manufacturer on Intel You know, Yeah, I mean there's like I think structural issues with Intel and in that the the Um the CPU is no longer on the critical path to Performance in in a any kind of like compute application as in it's really your AI accelerator. And so if if the CPU is just like a glorified kind of like router as opposed to an actual thing that's driving performance, then then why bother spending Intel margins when you can do like an an arm CPU that's just as performant. maybe even more performant for Less cost.

So it seems like they're moving purposefully almost now into manufacturing because I feel like they're starting to see the writing on the wall. right? right? And so then then there's the like on The Foundry side is manufacturing chips. An interesting business it is. and you know, we, um, kind of like Tsmc is a really interesting company, right? Uh and uh.

It is a heavily Capital intensive business and and requires a lot of investment to get to that next node to deliver that next performance. Uh, and and you know tends to be lower margin than stuff that we're interested in. But I Think that like the argument for the likes of Tsmc is that you you know because AI is going to be in a whole variety of in-use devices. Like you're going to have some kind of chip in that camera that that is more compute intensive than exists in it today.

Um then that means that Tsmc will have a a larger variety of in buyers of its chips so it should have more pricing power. uh in that situation? Um, but it's it's you know we tend to focus on kind of uh there I Think there is like we do invest in companies that are the enabling tools for the technologies that we focus on. but we really try to like to the extent possible focus on the platform itself, the end product. essentially.

So what? in other words, sort of like what we first started talking about. We we don't want Google.com We want the answer so we don't really care about the chip. We want what the chip does for us well and ultimately what that does. Yeah, ultimately I want I Want you know to deliver a good return for my end clients? But yeah I think I Think that like when we do the analysis of companies and uh we look at in even entire sectors, we look at where in the value chain we think cash flow is likely to accrue and then how those companies are valued but go back I think with AI there the analogy to um, kind of the early.com errors is pretty apt.

and there if like in 92 or maybe 91 like the company to invest in was Cisco You know it wasn't Amazon didn't exist yet, right? Like you know it wasn't Uh, AOL hadn't even spun up. Sure, right. And it's because like providing the tools that connected all of these apps, all of these computers together. Uh, was the right spot to invest.

and I think that you know we're still in the early days of kind of connecting together the world with AI So that could mean you know in Nvidia. On the other hand, we have massive aggressive expectations for AI Hardware investment. We think I think 2030. We think it's on the order of two trillion dollars a year will be being spent on AI compute infrastructure and so some share of that will flow to Nvidia Yeah, Um, but you know it's still valued pretty heavily even with that expectation.

Sure. So so to some degree you know the challenge of capital markets is that the thing that everybody expects to happen is the thing that's already priced in. Yeah, Yeah, exactly. Yeah, No kidding.

yeah. So what now? Uh I I Always want to bring this up as an opportunity for for all of your defense, a lot of folks in the mainstream media like to say oh, Kathy and the team at ARC it's Prof Fitness Tech But then you look at some of the companies, you actually have free cash flow. There's just a lot of stock based comp that might show a negative net income and supporters. What's what's say you in response to sort of that mainstream argument? Well ultimately the the you know you get paid on companies based on the cash flow not that they're generating in this moment, but the free cash flow they generate over the life of the investment Sure, right.

Or the Market's anticipation of that free cash flow once the market begins to anticipate it, right? Uh, and so the um, what we pay a lot of attention to is the balance sheet sensitivity of companies. So you know if how much money do they need to, uh, grow at the rate the we think they're capable of growing and um, can they, uh, internally generate those needs And if not, how are they going to raise the cash to do so. But whether or not a company is investing heavily in kind of like a strategic platform that we think is very large or not. Uh, actually, we would prefer them to be investing heavily, not trying to distribute cash out to shareholders or pay out a dividend or buy back shares.

I Think that like generating a lot of excess cash means that you don't and haven't identified provide a good spot to deploy that excess cache. And the reason we invest in the Technologies we invest in is we because we think that there's this huge value accrual Tailwind behind them and so companies would to the extent that they are able to spend money to increase their probability of and strategic footprint in these Cornerstone Technologies That's in the best interest of of shareholders and their investors. That's an interesting idea that if companies are conducting BuyBacks they've lost sort of the idea of what should we do with this money. Let's just buy back the shares because they fall in a bit now.

I Think you all have large uh Ambitions or or I shouldn't say Ambitions uh large projections for connected TV You also have a large investment in Roku and uh, not in Trade Desk. Trade Desk just announced large BuyBacks and Roku just increased its net loss. What's your take about Trade Desk is that sort of an example where Ca Trade Desk is doing the BuyBacks They don't know where to place the money. Let's stick with Roku I mean I Think that I'll just say that we can make a very clear case for Roku and the value accrual that we think is going to happen with Roku I mean if you look, they just reported a quarter where um, their all of their engagement Saps beat even the base case in our in our publicized valuation model.

Like a way to think about Roku is is that they currently generate around three cents per hour streamed in net revenue to the company. Okay, and so look at, think of um, you know, traditional cable when you're paying for a cable package, you're paying like on the order of a dollar per hour, right? And even if you're paying for Netflix like given kind of their streaming Trends might be 20 25 cents per hour. Sure, so Roku shouldn't command that because they they serve as like a lead generation for for Netflix right? But but um, we think that as uh, kind of traditional TV People cut the cord and move on to Roku Uh, they will generate an ecosystem that should generate five cents per hour that's embedded our expectation for the company. and um, you know the the engagement and viewership trends that they're seeing are right in line with our base case forecast.

So if you combine kind of like five cents per hour uh and in net revenue and uh, that kind of Engagement Trends Continuing, then you get to our price Target which is 600 or so a share, right? And so um, I think I Think it's a really interesting exposure because largely the advertising dollars haven't followed people into the streaming ecosystem yet right right when and if they do and trade desk could be an interesting exposure on this basis as well. But we we personally think that like or we at ARC think that like having the operating system for for the TV platform is the better spot to kind of accrue that lead generation to add dollars. but the that there should be an entity that sits between all of those ad dollars and distributes it out to viewers and Roku's positions. you know I Think something really that resonated with me that you mentioned was this balance balance sheet Sensitivity I Analyzed very briefly just the earnings of uh, both Roku and Shopify this morning and even though they both have losses substantial losses on a net income basis a lot of which is stock based comp, their balance sheets are incredibly robust.

I think Shopify sitting on over five billion dollars of free cash I think Roku's somewhere on 1.6 roughly from my memory. So I think you're right, they have plenty of time to to essentially realize those. Visions is that one of the big priorities? When you all at ARC do a sort of an analysis is can do they have the balance sheet to support that Vision Well I mean for us the first like the way in which we um kind of call it our decision making flows for is first at the top down. like is this company positioned in front of a technology that we believe in that I believe is going to um kind of um basically has a structural Tailwind sitting in a In in a sector where we think cash flow is going to extra normally accrue and then is our point of view differentiated from the rest of the market? Just in terms of that top down, that is a consideration already at the beginning.

Yeah, do we have a different opinion than others because if it's expected, then it's priced in right? Okay, okay, right. But it's also an easy way. I mean I think that you know we So we look at Technologies like VR as an example where we've kind of been like you know we do our work, somebody else does their or we don't look at the other work before we do it. we finish and then we compare and we're like oh, we're the same as we're actually below where other people are, then it's not going to do us any good to then try to like, continue to to pull on that string yes and so then then it's kind of like given our expectations for the technology in the company's position, like what um, could the top line look like for it and what margin structure should the company be able to uh to accrue getting to that Top Line number and then if that's interesting, then it then we back into okay, this is where they're at on a balance sheet perspective.

This is what their burn looks like if they're still burning cash and then what do they need to get to that kind of crossover point where they you know where likely the point at which the market begins to realize and recognize the cash flow accrual potential of the business. Ah That's fascinating yeah. I Love the idea of if it's already uh, uh, something that's a popular belief, you're going to look elsewhere I Think that's fantastic now. VR It didn't sound like you're too bullish on it.

Yeah I mean I mean the it cuts so massively against the trend in technology and user Behavior With technology, it's hard uh, at least in the current form factor for us to see it. uh, really working I Think that, um, we've done cost decline work on on VR headsets. and so we think by I think it's by 2032. We think for somewhere between a thousand and three thousand dollars, you'll have a headset that you'll put on and Visually it will be indistinguishable from reality.

So that's like I Don't know if that's actually even a meaningful like. conceptually, there should be something interesting before that happens, but in terms of like thinking about it, it's still. it's still really expensive for what it is for the end consumer. And and so the um, you know there's rumors that Apple's headset is going to cost three thousand dollars.

Yeah, that's there is no that is not a product that will work at that price point. Interesting. Uh, and so um, then you know the the reality of these headsets seems to be mostly they're given as Christmas gifts and then they sit on somebody's shelf. People don't use them.

So Nick who covers the consumer Internet space for us? He he has one and it's like Nick You have to let me know when you actually use one of these sets, you know because because uh, the and there's a real chicken and egg problem between. like the um, if you only have a small audience that buys them, then no game developer is going to devote a big budget to creating a game specific for that experience. Without a game experience Like a game specific for that experience, it's going to be hard to get people to buy it. And so um I I Think that for me like a subjective indicator I'm looking for is um when do I read a news story about some poor teenager who has like died of starvation because he's been so absorbed by his VR experience that he didn't resurface to get food or something right? Like there have to be the like dramatic kind of this is a a a terrible stain on Humanity that these things exist out of news stories before we're close to a product that is going to be Mass adopted.

There have to be early adopters that are losing themselves in it before it's something that somebody will spend you know, a thousand dollars of and then stick a headset on. Yeah, and that's not even remotely close to happening. in other words, right? right? So making the extreme argument. obviously we don't want anybody to die, but because we know nobody's even taking it close to the fringes of extreme, it's probably not right.

No. I'm No. I feel very sorry for the teenager I'm just saying like if that were happening, yeah, it would be on the front page of the New York Times or something because it's such a salacious story. Be compelling also for them people.

So I want to try that. It's not. It's not like there's you know there there is a market for that kind of media. and so because that media doesn't exist means these things aren't being used that much.

Yeah, right. and and um, the the and then, like the tendency is not to just have like one screen that I'm only looking at is to have like six screens, right? It's like instead I'm watching TV and on Twitter at the same time and so I just think it's it's it. If VR takes off in the way that it's being presented, it's like a huge swerve away from kind of the direction that people have been going VR and your Robo taxing see. But I mean you know, for one thing, like you'll just have a screen in the robot actually.

but realistically, you'll you'll be on your own phone. You'll be. you want to put somebody else's headset on? Probably no, the taxi headsets. It can be a little weird.

Yeah, uh. so y'all were in Matterport briefly. Yeah, not anymore. What happened? I mean you know in the process of kind of the downturn over the past year, there is a winnowing that happens in portfolios.

As we say. you know it's not that we don't like this company or that company, but um, you know we try to concentrate portfolios during downturns to really essentially to wind up the spring even tighter for for the turn. uh and so um, it's not that there was a, you know, some kind of, um, salacious event or anything that that uh, it's more like a a prospect of kind of like relative value. and where do we have more confidence? Yeah, yeah, I mean I I agree with you anyway I Love Matterport.

Don't get me wrong, but I also know they've grown users like crazy selling the uh, the iPhone version of the Matterport and I'm not a big fan of it. The uh, the larger cameras they use for you know, purposes like scanning this or real estate is fantastic. but we know what's happening in real estate commercial real estate right? And there is some like sensitivity to to in markets and kind of like the the macro cycle of course. Yeah yeah.

so uh is something you mentioned uh made me think of yeah you need to create sort of people people. You need people using the headsets to create the games and but you have to have something compelling to to get people to buy the headsets. It made me think of uh, and nobody likes referencing this company anymore. But Nicola They had a very interesting idea about the hydrogen battery electric vehicle hybrid and that by having people who had a hydrogen vehicle would induce hydrogen gas station demand.

but in the meantime they could use the electric vehicle battery charging stations. Is there even a remote future for that ever coming back? I Mean the work that we've done on hydrogen fuel cells suggests no, but you know we with every technology we look at we will revisit it periodically and say is you know, does the cost decline that we've done still hold like what it have the unit economics changed for in buyers. The problem with like particularly Hydrogen Passenger cars but even in the trucking space is that the a hydrogen station. it's very expensive called a million dollars uh or a couple million dollars whereas a supercharger is like a hundred thousand dollars is that kind of order of difference right? So uh, from the consumer side, it's really hard to infill all the infrastructure you need to make it useful.

Uh, and then the cost of the fuel cell itself is is still very expensive and so kind of. the work that we did suggested that you would need it a passenger vehicle selling at Toyota Prius like levels for a decade to get fuel cells. cost competitive with electric Drive trains. Oh wow.

okay I understand I See uh now obviously then that brings us to Lithium. Everybody We've seen Lithium 10x in the last few years. Everybody is thinking how can I invest in profit off of lithium but you all don't invest in commodities we have in the past I mean we've invested in Albemarle before we're not right now and I think with commodities the reason not to is certainly the the amount of people talking about how how can I invest in lithium makes me think it's probably not a good time to invest. Interesting.

Uh, and you know with commodity markets the price signal is very powerful and it's not like the if you look at the entire raw material stack going into electric vehicles. Cobalt for us was more of a concern because Cobalt is co-produced with I think it's copper and Tin maybe. Anyway, it's co-produced so as in if I give you a big price signal for Cobalt that won't necessarily Drive the economics of bringing on more Cobalt Supply because you could have a Cobalt price signal without the underlying elevating and so you wouldn't increase production lithium. On the other hand, um, you know a miner sees the lithium price and it's like a big blinking dollar sign like produce more lithium, right? right? Yeah, you want to get rich.

Go mine Lithium. And if you look at Reserve to production ratios and lithium, they are, um, extremely high. Meaning there's tons of lithium out there. You just need to bring the mines online if you look across.

If you look at our forecast for electric vehicles which is more aggressive I think than at least anyone. To my knowledge, the um, the amount of capital expenditure in aggregate required to support that to bring on additional lithium supply is is 30 billion dollars. Okay, so relative to it's something like the Oems have announced 600 billion in capital expenditures for uh, electric vehicle programs. Sure, so it's actually not that big an amount of money to bring on the lithium.

and I in in the history of commodity markets when people are pointing at oh, there's going to be a huge supply problem five years from now that usually doesn't turn out to be a supply. you're supposed to run out of oil 10 years ago and then so if you look at I think you can go back over any commodity super cycle and they people are always like, oh, we're going to run out of copper, We're going to run out, of, you know, Reserve to production and copper has basically been 40 years perpetually even though we've kept using it, right? So it's like there's always you know you always are growing Reserve or growing production. but you're also growing reserves. like proportionately.

So um, it's just not a Um in Commodities it's it's not a, You know where you have a clean price signal. you know Supply should come on, are we going to war with China I Hope not. Yeah I mean I think I I Don't know. This is probably like way outside.

Actually, it's probably so far outside my wheelhouse. I'll just be like very fair. Okay, okay, I don't know. So if you stay out of Russia and Ukraine as well mostly then I mean yeah.

I I Think there's a reason that in our technology forecasting, we don't make it contingent upon political development. and it's because political developments are so idiosyncratic and bizarre and a function of like the people that are in the room. and I think things that are like, not forecastable, Um, that you're better off, um, just accepting that that's going to be noise in the forecast rather than trying to. So that's something that, uh, obvio