Ep20. AI Scaling Laws, DOGE, FSD 13, Trump Markets | BG2 w/ Bill Gurley & Brad Gerstner
01:03:11 Share

And this was freeman opening in a very quick, rapid fire interview of the fourteen government agencies that existed at that moment time, which he would abolish, in which he would keep and, you know, will roll IT and put a clip in here, keep them or abolish them. Department of agriculture, abolish. gone.

Department of commerce, abolish. gone. Department of defense, keep, keep.

IT. Department of education, abolish. Gone energy, abolish. How accepted energy ties with a military.

Well, then we shot IT under defense a little bit. That handles the nuclear IT goes under defence, olive. The rest of partment commerce abolish education, abolish.

I built were in my office last night. My boys, my boys put on the cowboy hat, the cowboy hat we made down at your universe.

That looks great on you. It's great if you wear IT for the whole episode.

It's good to be here. It's kind of a surprise this morning and I got on the plane to fly up to seattle, is going to do a pod with socha, as you know, and that we are going to do our reaction pod and a typhoon or something hit seattle last night, literally like halfway up there. I got word that microsoft has no power.

Nobody has in power. Trees are down. So I did A U. Turn and wish them well. I'm going to go back out there on monday and do a pod with such a but given that, I couldn't wait to get together with you and kick around all the things that we've been sharing back and forth .

over the last couple weeks.

stuff no doubt about IT. One of the things certainly popular or threads has been A I scaling laws. You know, like are we beginning to see models top out, particularly on pretrail ing in fact there is this bloomberg headline open a ei google anthropic struggled to build more advanced ai IT then goes on to say that dorian or ChatGPT five gm I and three point five opens are all falling short of internal pretrail ing targets um and then they quote A A darro has saying saying laws are not actually laws of the universe but they are simply imperial regularities he said. I'm going to be in favor of them continuing, but i'm not certain of that know so what what's been on your mind about whether or not we are, in fact, scene continued gains from pre training where .

and I I would add a few other things. He was also quoted as questioning whether scaling laws were continued and for whatever reason, a mark been over a generous and horis also made the same statement. So in a very short window, we got this point of view from a quite a few number of people.

I would add that in dari OS five hour interview with lexus, gently grab pieces of from the try scrip. But he was very positive on scaling laws in that interview, which just recently drop. So um so there may be there may be a disagreement between some people, but the breath of the feedback s suggest something I think sometimes were paying attention to.

I I read that there were people and you that raised this question early on and and I think it's specific to allow limps and I don't think you should say A I scaling laws. I think there was always a question about whether ella limbs would run out of steam and that was tied to three different things, which I, we unite, talked about on july element. And I I even you raise this question a year ago.

One is, will the premier account run out historically, mathematical algorithms that do some type of fit. And this is a very sophisticated m of, but when you take the variables up to a certain level, IT stops adding value. You just get too close to the fed. There's a question about how big the context window will be, and that came up on the in video call day. I think we talk about that.

And then data was brought up and a lot of people I know this was a big point melani Mitchell raised, like, are we just gna run out of data? And there was pushed back that there's synthetic data, but there's been papers published that show synthetic data creates a lot of chaos. And so I do think there were people that were thinking this might happen.

And there was also a belief, and I share this belief, you don't have to. So we can disagree on this, that there was a argument being made by, you know, OpenAIron an d an thropic di d. So you're just going to see the next number drop and the next numbers going to be way Better than the last number that's going to happen routinely.

And they both said, you know, we spent whatever a billion on this model, we're going to spend ten and then we're going to spend one hundred. And and that implies, I would say, at least your scaling or maybe above. And they all said that.

And so if if the the comment you know that you red earlier is true and that this and that are not getting the benefit, there are implications, you know doesn't mean a eyes in trouble or A I is done. I mean, there's tons of positive AI news out there, but IT may mean we're shifting directions. And it's worth, I think, talking about, well, if this is true, what are the applications?

right? No debt about IT you know genson did you say on the video call he he basically said there is no slowdown, but he did say that foundation model pretrail ing is only one vector of scaling. You obviously have post training scaling um and now we have inferences, time reasoning that will scale.

Yeah we have data and data quality that will scale. So I think that you know is a question as we look at in the early phases over the last twenty four months, everybody was mesmerized by these e values on free training. And then I think a lot of people mistakingly, we'll get to this later.

A lot of people mistakenly said if the retraining ever slows down, this is horrible for in video because the only reason in video is going up is because people are buying these big clusters to do pre training, which also is is an entirely true. But I do think it's important to recognize, right, they as as as they can as as you scale up retraining, a lot of the low hanging fruit was lot. And so that makes sense to me that you're seeing a deceleration in the rate of improvement of the free training.

But when you look in the aggregate at all these various vectors on which we're scaling intelligence on a combinatorial basis, that may be just as good or even Better. But I think it's important to your point, like let's steal me on the argument, let's assume that we do you know kind of run out of improvements or see a significant deceleration in the retraining improvements on these models. What do you think that means? Like who are the people who are hurt by that? Who are the people who are who are helped by that?

The first thing that that comes to my mind and and what's to get, can you do agree with this, is just like what in my head, is that this is not the expectation that that salmond IO had twelve months ago, like because they were promoting a thesis that was just going to keep going up and up and they were willing to put ten ex over and over again on the size of the club they were going to train on.

And if it's a shift and IT may very well be a shift that leads to insane innovation and continued growth. Lab a lab, blah. I still think it's a surprise ship like I don't think that was what was expected.

And so I think that's worth acknowledged if if that's what happened. Two other things that I think are let's just call me m questions that raise one. The invidia differentiation is we've talked about his greatest that the largest cluster size.

So if retraining, once you going to only equate this tile limbs, I don't think of fsd problem. I think that that is more record eye problem. And that thing may scare away beyond where IT is today. But but for Allan limbs, at least if if we hit a cap, then, you know, does that impact the in video demanded all for the bigger systems.

And then there's a different nation question, which is if the world's shift and inference sooner or faster, and if the path of differentiation for the language model companies becomes more inference base, then retrain based, what does that mean for the hardware systems that people are using? And yeah and then the third thing and I go quicking, you can respond is, you know my ceiling on that pre training allow other people to catch up faster and with open a eye. And and there was an ark out today about a like pu du chinese model that was performing at or near some of the lake deep.

And you those would .

be the three things that I would say that you would want to analyze if you thought I was true.

We've spent a lot of time talking with, for example, to ceos, most of the hyper scales, if not all of them. And then I listened to jensen on the call today, right? They called hopper demand exception on and this is the very tail.

And we may have people thought hopper demand would be over by now, Frankly. And he said it's going to continue well into next year and interest. They call blackwell dean staggering and and said that you know they're increasing the production even against their prior predictions, right? They beat the quarter in in terms of revenue.

So the action on the street seems to support jenson's words on the call, which is there's no slowdown, right? Sam twisted, I think there is no wall. And so there is a debate here and some of that may be semantics, right? Like I definitely think the second derivation is slowing, right? But you're still seeing gains.

And I think the implication we've hit a wall suggests no gains and that there are no different vectors of scale that intelligence. So I think at least as we look at twenty, twenty five foreign video for the supply chain and for the model companies themselves, we think we're going to continue to see improvements in the models. IT won't all come from pre training. We think there's an increasing amount coming from post training, from data quality. And then obviously, from this from this inference time reasoning.

I would add to what you said, I want to listen to the call. Yeah, there was certainly no concern race by chance about either his demand or his sense of what was going on at tarra tecture al level. And microsoft has been in increasingly boysterous as well.

If this thing is true, IT certainly goes against what Kevin Scott was saying on a podcast four, five months ago. And so I go, I doubt that if even if IT worked you, that I would play out the customers as fast as we're talking about. This stuff just came out like weeks ago, but I would not ignore.

I I think it's important to understand I think it's important to know what the drivers are of incremental investment and incremental an incremental performance and and competition and to see where that's going. And I think there are a lot of questions that you know we need to ask and understand about inference. You know, do models that, as I understand IT from the start of so we work with you can, you know, train a model on a big in video cluster, run IT on, you know, a different system.

There are certainly performance test that are in the public that show the rox since reverse and other TPU, including the ones in google. Out performing on inference now on the call of jensen said something that I hadn't heard him say before, which is he talked about context. Windows um you know potentially growing really large and that might then require a larger cluster, which might bring you know a point prior on the previous called this.

And I just said, well, the reason you'll use us for inference is because you ve got all machines land around. This was a step in a different direction, saying, well, if context winter are huge, we might be differentiated yet again here with larger system. So I think those are the things that people need to ask him figure.

There is no doubt. I think one of the implications of what you're saying right is as we looked at free training, we basically watched what appeared to be the increasing commoditization of the retraining and vanity, right? Everybody was catching up pretty quickly, even people who we didn't think we're necessarily in the lead pack.

So if, in fact, we're seeing a declaration right in the rate of improvement, if that's becoming more commodity than IT shifts the baLance of power to post training, which really relies on quality of data, right? And other vectors of scaling intelligence like inference, time reasoning and those things may be less commodity, either because you have a data fly wheel that's Better than your competitors or because you've just had some breakthrough architecturally on other ways to scale intelligence. And so you know, I think twenty twenty five is going to cast more light on those things as well as improvements into the core product like we've talked about before .

around memory and actions, and there was a lot of talk on memory in the past week. So both google geri released a model in their chat, but that has memory. And and most of the microsoft drop the phrase near infinite memory.

Now I putting near next to the word infinite, I think it's a bit of an actually more, and you could say i'm gna live near forever. I'm not sure what that means. Clever, you.

clever. Worse of semantics. But they have, they seem super excited about what theyve on left in memory. And i've talked for a long time about, I think that the more memory one of these systems can have, the way more useful becomes to each individual that uses IT. So I look forward to to seeing what they have and seeing when they release that.

You know, I would I would just make one comment kind over again on on the, you know, if this scaling walls, ceiling is starting to arise. There was talk prior to now. Or if you if you sume that weren't true that these companies were just going to train on bigger and bigger and bigger systems.

So someone train a model on a hundred billion dollar system one day. If it's true that it's top that's not gonna en anymore. And so that that is a IT doesn't mean that the invention may still be sold out to two and thirty. But there was this notion that there just gonna be constantly building larger model for pre training and a lemon that maybe off the table, if this is true.

And let me let me suggest as an investor, at least in one of them, and open the eye IT may not be a bad thing that that off the table.

right?

I mean, I do think there was this question around return, around returns and where the money was gonna from. And i'm not worried about IT from the perspective of the video because I think um again, well, we're all hyper focus on these events, right? The real world people are focused on use cases.

So if you break down consumer and enterprise here for a second, right, you heard Johnson referring to this a lot on the call. Remember microsoft said of its ten billion dollars in a revenue. It's almost all inference.

We know in video has said half of their revenue is influence. And he was asked on the call if that mix would become higher in the coming years. And he said, I expect IT and it'll be and .

I would be great for the world when .

IT does yeah because I would mean a lot more people are getting benefits from IT. But let's talk about this vector around around consumer A I, right? You mention my staff are referring to, you know, infinite memories, very clear. Now he said, next twenty twenty five gonna the year of memory.

You and I started off this year hoping that we know twenty twenty four might be the year of memory because we think these agents are gonna come infinitely more useful to us when um we don't have to prom them about all the things we've done in the past. And we've also seen everybody began to roll out actions every company. Now german, I has talked about that we saw computer use out of topic.

You know, I took note that sam said recently that tools and agents that can do things like book airlink tickets are at least as important as Better models. We will have Better and Better models. But I think the next giant breakthrough will be with agents and their actions.

I think that these, again, they're not either or right. The core models will get more capable multi I modality inference, time, reason, post training. But I also think we finally see out of these companies a real cases around product improvements, memory actions, voice, voice, right, that just dramatically increase the utility of these these things, which then leads to massive increase in token production or inference.

And that influence has got to run somewhere, right? So that doesn't. And when I look at ChatGPT, you can see the numbers that out there publicly IT certainly does not seem to be slowing down, in fact, if anything, accelerating.

Um so I think those are indicators, right? Because i've heard the claude and other models are doing perplexity are doing well as well. I think those are indicators on the consumer side, right, that consumers are finding more and more utility. And I think twenty five with these product improvements, I think that could be another step function unlock in terms of consumer .

use in an enterprise use. I mean, I all those things, memory, voice, um they can be utilized in a in a Price in a fashion for sure and and will be and so yeah, I think I think you're right about that. I think those innovations will matter for exact same reason you said which is utility. I think snowflake .

set on their call tonight thirty two hundred of their customers. So they have about ten thousand enterprise customers. Thirty two hundred of them are using some of their A I use cases AI products now in daily use.

So again, starting to see some of that indication. I noted that microsoft, which is on a sixty six billion dollar run rated, as your has said, they expect q one and q two as your revenue to accelerate. Think about that, bill, as is growing at thirty four percent on a sixty six billion dollar base, and they say they expected to accelerate.

So it's it's really significant growth on big software businesses. And I think what we've seen in the season is that software certainly has not decelerated. If anything, it's accelerated. Um after a couple years of digestion, I think around this spite you know if I had one take away, having watched all of this, I would agree with you and you've been you know rattle my cage about this for a year and I think you'll ve been proven at least partially right that we're seeing a deceleration ation in the rate of improvement on pretrail ing. Um but my conclusion is I don't think IT necessarily matters to the consumption.

Uh, on the video side, certainly IT would be higher, could be even more, I suppose, if they were able to produce that many more, which I don't think they can. But I think that there's you know still in a very strong position. And I think these models, I think, you know, for example, open the eye, could double, triple the number of users of ChatGPT with Frankly, not making any improvements in the core model because consumers today barely scratched the services on core model capabilities, product improvements like voice, memory and actions. I think any of themselves could drive, uh, tremendous traction and revenue increases. And that but I but I do think he'll still continue to see some of those model improvement.

Maybe one area of that, I know that we've actually seen a tremendous return, I think, on the investment made in AI, and we saw some news this week is on national, a full self driving regulation, right? So the trumpet administration announced that they want the dot to develop a national framework to regulate self driving rather than the state by state framework, right? So most states, I think, require certain to be able to do in texas and florida have no restrictions, but at least the indications are out of the administration where that they want a national policy, which makes sense to me.

I think that's very beneficial to tesla. And a way more and to anybody else is in the the full self driving game, you get to deal with one regulatory authority rather than trying to, you know, deal with the complex nature of fifty of these folks. That happened about the same time our teams doing a tremendous amount of work around fsd thirteen, which we expect to roll out in the next next few weeks.

okay. So you and I did the pot earlier in the year, bill, on fsd twelve point three as a big breakthrough. Is the first ima learning model really moving away from these determination models where we saw step function in terms of capability? Elon IT said they were going for making ten percent improvements a year to ten x improvements in, in a much shorter period of time.

So we know about fsd. Thirteen is the first model we've trained on the Austin supercluster right? Parameter size is about five times bigger.

And what I hear is the island's priorities around that are safety, safety, safety, safety, right? He wants different driver to be able to understand different driver modalities in different situations, right? So that you can you know, the metric which they use is n PCI miles per critical disenchantment.

okay. And my partner, freeda, did a, you know, tweet like a couple weeks ago at the start of this year, we saw twelve point one come out by the middle. This year when they released twelve point five IT showed a hundred x improvement in mile's per critical disenchantment.

And then with the launch of fsd thirteen in the next couple of weeks, we expect another ten ex improvement in terms of that n PCI. But here was the the crazy stuff for me. So you've already at a thousand x Better than we were at the start of this year in terms of miles per critical engagement.

Now if you do some back at the envelope, which he did in her tweet, elon retweet IT at launch in q four of this year. IT looks like version thirteen will be somewhere around twenty five thousand miles per critical disenchantment. And we estimate that way mos right around that same level, maybe twenty thousand miles.

So they already caught up, we think with with way. And at the current rate of improvement, right in two, one of next year, they could be at one hundred thousand miles per critical disenchantment. By the middle next year, five hundred thousand miles per critical disenchantment. Now just to put this in perspective, there's a human accident every five hundred thousand miles, right, that self reported with human driver.

So if twenty twenty four was the ChatGPT moment for full soft driving bill, where we started seeing emos on every street corner, and sanford cisco, where you and I had our, you know, and Michael dell tweet about at our tesla fsd twelve three moment, then if this was the ChatGPT year, I think next year really is the year of achievement of a safety standard that allows robot taxi to go into action, probably in queue to the pick couple cities. And that doesn't mean that you can pull the steering wheel out of every tesla because, you know, elon is said, we're not there yet from a safety perspective. But the way, though, do IT for a robot taxi is you can put the, you can put the car on robot taxi mode.

So think about IT this way. If i'm driving down sand hill, rode out here, I have an expectation i'm sitting behind that will have an expectation as to how fast i'm gonna. And if the test is not going that fast, i'm going to disengage and drive the car, right? So because i'm sitting behind the wheel now, imagine i'm in a robot taxi, i'm sitting in the back, i'm leaning back, i'm on my phone, i'm watching something I will tolerate, just like I tolerate in a way.

Mo a different speed of driving, write a different safety standard, if you will. So I think that's how they get safe enough in a couple markets to be able to drive robot taxi is to just put IT in a safer driving mode. Um you know when i've watched these cars do IT they modulate modes depending upon the type of streets that they're on already um so that for me was paying attention to um that regulation combined with the improvements in fsd thirteen. I think that's going to be big news as we look forward at twenty twenty five.

There's a couple things I would I would respond to that I think you're gonna have to play out. First of all, as of now, tesla hasn't had cars driving around without drivers picking up people, and we most got years under their belt doing that. And there's gonna some learnings to come from that.

They're just dar, and so they're gonna. I think i've read that they're going to maybe to start doing that the next few months or something like that. But in the state city, allowing way more, you have to apply, you have to get approval.

So there's a step that they need to take. And I think there's some natural learning that comes from that. I mean, you can hear how they talk about space x and and there's a similar thing here, like until you get the cars out in the rope.

Um most people believe we most still has oversight from a data center somewhere where people are watching when engagement happen or this engagement happen so they can take over and pay attention. And so is that same infrastructure you're going to be built a test? I don't know.

Um that might be an important thing as well um to the national standard I unquestioned think I can be helpful. There are states in this uh fine country of bars i'd like to override regardless of national standard, most notably california. And so you know that would done but but of course you would make more sense for this to be done nationally.

And so I be grave. They stepped in you. There's A I always made a big purpose of open source in this area.

There's a lot of stuff that could be beneficial from a safety standpoint if everyone were IT. We're doing at the exact same way. And I could even imagine a communication network, the status of of a traffic light, for example.

Should there is no need for that to be in third, I mean, we're going to use how much compute to look at the light and determine whether it's red or Green. That's a state machine. It's either red or Green and that that can be known.

And so you could, as we build these intelligent systems, start to have the state of different things and which could include, you know road openings and road closures and and you know construction work is going on in traffic and all kind of other things that should be built into the system as well. So I do think that I do think a national kind of standard could be great. IT could also be really bad, like if you get over regulation in this massive you know stuff that ever in the in comments might just use IT to protect themselves.

But my um my own belief is that IT you've got you've got a who's got the most cars running autonomously on the street right now. Tesla, who's got all the advantages you just talked about, you've got uber who's got this huge network of users unless certain who the next player is after those three. That's got a strong argument right now. Oh, actually that is in the us. Obviously, a china has tons of autonation driving you know products on the street as we speak.

Well, let's talk about that because I think that's an interesting question um you know kind of a market structure question that i've been thinking a lot about lately, bill, certainly on the public fun side of our business, right, when I have a disruption of this magnitude in an industry that's a multitrillion dollar industry that as you know, twenty thirty different public players in the supply chain OEM eeta like I just start thinking who are the winners and who are the losers here, right? Exactly how it's feeling in two thousand and eight.

You see the iphone, know you've got nokia, you've got motorola, you've got a blackberry, you've got, you know, all this stuff, you're like somebody y's winning and somebody y's losing. I remember my mother didn't understand the concept of in right and two thousand and nine um I tell her to buy apple and to short nokia and that was my kind of onic example for her, you know and I worked out okay. I actually went against her first but but then IT worked out okay.

Let's just talk about EMS for a second. I think the consensus view in the world is that all these OEM, so i'm talking the U. S, O, EMS, the santis, the gms, the ford, the europeans, the workers wages, the bmw, the Mercedes, the japanese, the toyota saturated the chinese, the fsd becomes a commodity, and that they all end up with some version of IT.

Maybe there's one provider like invidia or you know mobile ee or somebody else who's providing this capability into these cars. And that you know think of that is maybe like this android version, that they're all this box, they are gonna plug in their cars and they're all going to go compete, you know, as though they were before um with tesla, that's one version of the world. The other version of the world is that this is a tech tonic disruption.

The tesla currently sells two million of the world's eighty million cars. But in the future, this isn't going to be about buying a car. This is going to be about a mode of transportation. And the tesla could end up restructuring in the entire industry, end up being tens of millions of cars. You may only have two or three or four global players, which you know we have a lot of other industries.

And so the question I have for you is do you you know what is your pattern recognition? Do you think that we're going to have all the same OEM? They're all going to have some system. It's going to be basically commoditized? Or is this a an extinction moment for some .

of these oim? Such a hard question. I mean, if so many different ways this could play on. One of the things that that I questioned in my mind is who's gna own these vehicles. And so you know, on one hand, you have the way most situation where they own all the cars, but they also know what that means from a cap. Same point. And if you if you look at the study kind of the statements they've made in conjunction with uber, they are at least going to experiment with some models where maybe they don't have to pay for all the cars or licensing their technology.

The other OEM that we talk about and and people ever dark talk about that you were talk about maybe large debt companies get behind and known these things, like maybe the containers work on a in a railyard right? Like who owns these things? They move around that someone owns and right people listen um so that's a big question you know and and the tesla argument is that they'll just beyond by the people that on the tesla, which is is great from a tesla, have the capex burrow out because i've talked him in to buy these cards.

But what percent of people are going to lease out their car every day? That's what would you guess? I have no idea what to put on that number.

I mean, airbnb does exist. So IT maybe not zero, but most of us still have our own house, right? So where is IT fall? I think.

you know my own hunch on that is that it's going to be a higher percentage, particularly is the Price points on the car is fall right now, the leasing cars for, you know two hundred bucks a month. And as the subscription Prices fall, I think you're going to have a whole category of people who don't mind IT, all right, earning incremental revenues on their car. I imagine they've done an internal survey work on this.

And I think if you look at the places were built, they're likely to launch the robot taxi. Number one is going to be in marked places where they have sufficient safety, right? So where they have a density of data that tells them they can put IT robot taxi mode, they won't have any problem having that could be somewhere like maybe the bay area, somewhere like Austin.

And then secondly, it's going to be a market where they have a lot of testers sitting on the sidelines so they come seat at with twenty five or fifty tesla. But in both of these places, they happen have tesla factories. There are a lot of people with teslas and there are a lot of people who will drop him into the pool.

So those will be good test markets. But if I think of a market bill like new york city, right, new york city seems to me to be a problem because there's massive demand for cars in new york city. But there aren't a lot of text list sitting around latten, right, that can be tapped into for supply.

So I don't think that will work equally well in all markets, but I think IT is. A massive advantage. Remember out of there, I think they may have seven million cars on the road today, something like that, about two and a half million of those.

I think our hardware for and it's going to require hardware for in order to run all these cars. So you we're really talking about multi year undertaking. But if you said to me over under on, you know, middle of next year where we have evidence of a single market or two where robot taxi is working, I think we will. I'm going to take the over on and that is a huge difference from where the world was eighteen or twenty four months ago.

I just really don't know what person of people will let their car go out. You know terroll exists. It's tiny, like tiny compared to what we're talking about. It's a nice business, but but that's where people lets .

everybody who owns that tesla has tesla APP on their phone and all they have to do is talking, says yes this times a day. I just think if they make IT easier enough, IT could surprise us. But in the industry.

and IT smells little funding. You know, you, your wife says, honey, I really need you to go out to the drug store right away. I've got this problem.

And you're like, 呃， i like the car. Go SHE like, what? And I do, you know exactly what's .

gonna happen here. The arbitration, the same arbitrage. And I saw an uber, right? You're going to have enterprising Young people. They're gonna least four of these among .

two hundred box I can box.

and they're going to be in this whole system. I just think the marketplace will fill up.

It's more sense to me than than than yeah some majority of humans letting their car go ever time.

And and then just to this industry structure question because I think this is one that's gonna play out over the course of the next three, five years. You what are we going to be long and what are we going to be sure? What are we think gonna really happen?

I'm skeptical. I'm skeptical. You know, we spend a lot of time looking at the chinese manufacturer free to spend a lot of time in china.

I'm skeptical on china. They don't have the sensors on the car. They don't have the data.

Um we think models and and access to the best chips are gonna matter. Certainly, they have advantages over non chinese manufacturer, but i'm skeptical on the european and manufacturer. I'm skeptical on the japanese manufacturers.

I could be wrong and a mentally flexible, the industry may end up looking exactly like the industry does today. But I think there is a path where this is, that event, that extinction of the tectonic event that leads takes us from twenty ims down to three or four. And if we SAT here in five or six or seven years, and we were on the past to three or four, a global to makers, IT wouldn't surprise me.

Maybe we can have an expert on on the tiny thing. I've certainly seen videos online to suggest, know the way more equivalent are already, you know, working in china. So that would that be a data point? We could free to went.

rode in every single one of, I think there were twelve. He took videos in every car and set IT back to the team. And some some of were totally terrifying, obviously anodos yeah everybody from Willy to buy due to uh to D D to B Y D eta has their version of .

IT is another question, I think, which is both and test hand potentially licensing. And so if there be an interesting um maybe that happens down the road between the his the historical incomes and these two new companies on whether that type model plays out seems quite likely to me.

Speaking speaking of speaking of iran real quick .

when I think that I forgot to mention on the national footprint. Ah um IT may be important if they're gonna put that together. Assuming that this is A A issue that the the country wants to support, wants to see leadership in, they may need to to consider some type of uh insurance reform um or in litigation uh limitations built into the whole thing.

Because I do think that we have this huge problem in amErica where litigation is so rampant. And I can just imagine some lawyer in front of a jury talking about how the computer killed somebody and and trying to extract tens, hundreds of millions of dollars and that that could be a real problem for the industry. So if they're going to consider all the other safety stuff, I think they should also think about liability limitations um especially on like just you know arbitrary Jerry awards and things like that.

Well, I know listen, I think that that's a perfect segway to talking about dogs and what's happening with this tectonic election we've just had. And I think those those really big ambitious objectives, which Frankly, we haven't seen a lot of you. We haven't seen a lot of ambitious efforts to reform.

I think over the course, the last humid administrations, maybe both republican, democrat, the one thing about the current administration IT strikes me as a level of ambition that I don't know that i've seen you know in my lifetime, bill but one is the department of government efficiency um so this is you know the unofficial government department which is run by vaaka and elon and is overseen trying to reform government spending, government regulations. Frankly reducing the size of the federal government with the overall objective of getting control of our national deficit or our national debt, which we all agree is pretty egregious. And and one of the the videos S I saw going around, and I saw several people to eating IT, including iron, was of somebody you and I.

I think i've a lot of affection for a milton freeman. And this was freemen opening in a very quick wrap ID fire interview of the fourteen government agencies that existed at that moment time, which he would abolish, in which he would keep and, you know, will roll IT and put a clip in here, keep or abolish them. Department of agriculture, my abolish gone.

Department of commerce. Abolish gone department of defense. Keep, keep IT department of education. Abolish gone energy. Abolish, accept the energy ties.

And with the military, well, then we shot IT under, defend the little bit that handles the nuclear. So IT goes under defense. But we abolish the rest of the asked department, agriculture, abolish commerce, abolish education, abolish um etta. And I think in tweet that a kind of sets some boundary conditions right of a level of ambition that I don't think many people they heard the words, but I don't think many people said now they're not really going to do there. They're not really serious about going after that.

What do you expect out of the department of government efficiency? Um and you know kind is the agenda going to be that ambitious? Or do you think that will hit the wall of some congressional reality and maybe they're just setting the boundary way out there that they can negotiate back from?

Well, first of all, I I mean, I think if if you extrapolate the growth of government, you know two or three decades into the future, you know we would get to a point where such a large percentage of the population and work for the government that you IT IT was self and people like and and many people, uh, very smart people have also highlighted that our debt is is too large. The U. S.

Country debt is too large, and the interest payments are now a huge part of federal budget. And how can you possibly solve that if you don't drink the size of government? Now prior to the election, um the the paper said that both candidates are going to spend the same amount.

So this kind of new department is is, I think, or talk to do what most of the the media thought trump was gonna, I would add. I highly recommend everyone watched the two hour interview between lex remen and and malaya, argentina. I think a lot of this is being provoked by him.

He was recently in town and hung out with with, with the dash team, dosh team. If you take a matter word for what has been accomplished in argentina in such a short window of time, six to nine months, it's pretty spectacular. And maybe it's important for someone to fact check that. But but really remarkable what can happen if you improve efficiency and you things like you get rid of a turn regulation around um ring control and housing improve for everyone you and for me like that second of course of course that's what would happen.

And so how do how do you tear this apart, guy? We we could simultaneously, um if they're successful, see a reduction in government and an acceleration or proliferation of growth um because you find out governments kitten, in the way, as I said on the past two or three podcast, you know governor superior pennsylvania keeps getting accolades for moving regulation out of the way. It's so ridiculous like if that if that's what you get celebrated for, we should just rewrite regulation and and get IT off the books.

And we we know there are states that are more productive and are growing faster right now and actually increasing their population because people can build things, they can build factories, they can build you know, distribution center and everything can happen. They can build solar farms, you know, a lot faster. And so I do think it's super important. Um I really encourage people to watch the malaysian. The part is that most interesting because I think the easy push back is, although never be able to get this to happen, washington you know in count in scots and you just can't move IT are these people will protect themselves the methodology that i'm going to assume was elan's idea um is basically to shine a flashlight on the ideot racy that exists. And you know, transparency can be a hell of a disinfectant.

And while you and I have started this, I I took a flash look at twitter and viva just announced that is going to pause his own podcast to launch a new protest with he called those cars where they're going to give regular updates on the cost cutting and what they're going to do and they're by we asked people to send in their favorite, you know, example of government waste, they're gona they're gonna talk about IT. And I think when you highlight to the american population really idiotic and they're going to get behind this thing. And right.

and that's that's a point here, you know, comes to this question of can they get IT done, right? Can they get IT done? And listen, there been plenty of president clinton ragan eeta who've talked about getting government control spending under control that talked about empower ing the individual versus the state.

And then they meet the leviathan in of washington. Maybe just a little background here on the executive powers that relates to this. So there's a constitutional doctoring that we've been talking about, an r thread, the doctor of impounded, and it's the ability of the executive branched with holder, delay the spending, a money appropriated by congress, right? And this was.

A question is to whether or not the executives branch actually had this authority. Um congress passed and act in ninety seven for the impounded control act, which basically said is restricted the the president's ability to withhold funding. Sane congress is the one that allocates funding.

If we allocate funding, then the executive brain test, implement the law, has to spend the money. Now, not withstanding that fact, there was a big in trump, one, the trump administration, without three hundred and ninety one million boxing military aid to ukraine that congress had appropriated. And some argued that this violated the ica.

Now why do I bring this up? I bring this up on the one hand, because there is going to be a constitutional authority question here. But over the weekend, I talk to several members of congress, both in the senate and in the house on the budget committee. And I will tell you there is strong agreement.

And IT wasn't just from republicans, right? There were people who were trying to figure out how they can support and interact with the department of government efficiency on both sides of the ae, right? And so I think that not withstanding the ica, if they can get congress to CoOperate, then there is no constitutional issue.

And then the final one bill, which is the one that I think is the most powerful here, here at the end of the day, what we've seen time and time again, the power of twitter, the power of x going directly to the people with your message saying, here's what government standing in the way of here's what we want to do, I think that overwhelmingly powerful shaming the government right really into making um what I think a lot of people will agree or logical cuts and so i'm boo lish. And you know one of the things I keep peering in the investment world is lots of people are trying to read between the lines, right? They're well, what do you think the president really says or what do you think is really going to happen? And I said, stop trying to guess.

Just listen to the words after the election. Trump is literally recording on ex reiterating exactly what he's going to do. People are tweet about what the department of government efficiency is going to do.

You are going to have these doge casts. And I think that's exactly what they're going to attempt to do. And I think this is going to be once in a generation reset and baLance of power between the individual in the state.

And Frankly, for one, I think it's well needed. I hope it's successful. Certainly theyll be some things they get wrong, just like anybody takes on an ambitious project will get some things wrong. But I think you can just undo those.

By the way, one of the things that already come up is I think two of the largest departments in the country have failed on its like three to five over the past years. If a single company on our public markets fails and on IT or or has to change dors as we recently saw a super micro you get not taken to the gun, which is how .

about that down ninety percent?

Yes, dear, it's considered A A focal of of the worst kind like IT IT. I've already seen interviews where someone's asking these department heads about these failed audits and they're like saying it's OK and so that should not be OK like the dollars are massive. These people work for the for the citizens of the country. And if they have an audit requirement, they should be meeting IT and you know that that's unshocked to find out that's true. Well, one thing I want .

to do I an a quick exercise because I wanted to just see i've heard a lot of people say it's unachievable. You can get the budget to baLance in four years. This is um and which I totally disagree with.

So you if you look at the federal revenues, they're about five point two trillion dollars, just over five trillion dollars expected this year. Government spending seven trillion. So we basically have a two trillion dollar deficit that spends up dramatically, by the way, from the forehand half trillion that we spent, two thousand and nineteen.

So encode IT, we just lost our mind and we've never regained our mind. So I just said, what's possible here, what would not even require draconian changes if you just grow revenues from here three to four percent over the course of the next four years, bill, and you cut costs in the first year by five percent, second year by five percent, thirty year by two percent. So all that does this get you back to trend line from two thousand and nineteen as though we increase spending three percent a year from two thousand and nineteen? The budget baLances, it's tiny.

These are tiny changes in cost, and I shared that with with the vacuum and they both had separate independent reactions, which is not nearly ambitious enough, right? And and here here, like there's no reason in my mind that we can do more. But what's encouraging to me and exciting to me, and I think will be well received by the bond market, is showing a four year plan just to get us to baLance right, just like a company does. Show me the long term, a boundary conditions, tell me where you're trying to go and then let's use the power of the bully popped the power of improvement to get people in line.

Agree, that would be interesting to watch. And once again, I I would encourage people to watch. I D say watch just because I I think malays is such a fascinating human.

And i'd watch her on youtube to seeing see its face of expression, although it's dub because they actually did interview vanished. But he's a very unique individual like he was he was in iraq ban. He he obviously has studied the history of economics like no other.

The cyber note and freemen.

which comes out in this thing, and you know if if you like I said, if what he said is true about what they accomplish IT maybe the fastest impact a president of a country has ever had on on, on the trajectory of that country, a country. And so anyway, i'm going i'm going to agree with you. I'm in the optimistic camp. It's about time. Um there is zero out my mind that making the government more efficient in our country um would be very positive for everyone that lives here.

So maybe just a quick market uh, rap and then and then we will bust out here. I get a lot of questions. In fact, i'm doing C N B C tomorrow, bill.

And and the question I get right now, like I always do in kind of november is what what about next year, right? You know and so people are beginning to you know, everybody's who's been in the market. We've had a really good year.

Lots of our our friends, i've had a really good year. So now we're starting to play a little bit more locked down poker, taxloss harvest, all the things that one does as you had to head into december of this year. And people start thinking about what is the portfolio that I want to start twenty, twenty five with.

And so maybe just at a high level, what I would say is I always start with kind of just market backdrop. So let me give you a sense. At the end of twenty, twenty two, you remember how bad twenty two was.

You know, mike Wilson was saying we're going to have a hard landing. Larry summers will say in interest rates are going to seven, eight percent. You know, we had seen, you know A A tremendous cell off almost thirty percent in the naznai.

So IT felt really terrible. But the set up for twenty, twenty three was actually pretty good. Just reverting to the mean would be a pretty good year.

Now we've had two pretty bullish years for technology, but we just had this historic election. So when I look at twenty five, here are the things I put on the process of the ledger, right? We're going to get lower taxes, right?

Not only the extension of the tax cuts that we had ten years ago, but IT looks like we're going to get some additional tax cuts on top of that, that the corporate rate may be go even lower from twenty one to fifteen. You may get adjustments down in the capital gains rate. You may see the reintroduction of assault.

Um so a lot of a lot of tax stimulus, we're gonna get less regulation like you and I just just discussed, rates are headed lower. They may be going lower, slower than we thought because the economy is doing well, but the fed has tons of flexibility. Our growth is still high, right GDP still ripping.

We see these corporate reports coming out. They're doing great. We are baLanced employment, right employments not so tight that you can't hire anybody, but our unemployment rate is still at historical lows and then were early in this tech and infrastructure super cycle.

The you and I started the pod talking about what's on the what's on the the negative side. Um well IT looks like we make IT you know a lot of terrifying. As we've seen to our letter now proposed for commerce, he's advocated as as trump.

We need to level the plane field and then negotiate from there. So level in the plane field is putting terrorists on a lot of products and a lot of countries that would that would presumably raise Prices. We also see we're going to a get reduced government spending.

And remember, government spending is a key driver of economic growth. So you'll put stimulus in by way of tax cuts, but reduce spending. And then you've got all that against the fact that market multiples are near all time highs.

So the s trading at twenty four times, I think the nzd x closer to twenty nine times. Um so in my mind, it's not a recipe for an all in moment, right? Because there's not blood in the streets.

We're closer to trumpets in the air then blood in the streets, as buffet would say. And he says cell when there are trumpets in the air. So IT all comes down to top line growth and bottom line growth for next year.

And when I look at that, there were two reports today that kind of stick out to me, right? On the one hand, we were short red a little bit of target because we're worried about terrace and we're worried about the the retail consumer holding up, especially around, you know, these highly levered consumers. When IT comes to housewares and things that target has exposure to, they miss their numbers. They took down guidance. Stock was taken out to the woodshed down twenty .

five percent. And I saw some data that suggested other deliveries are on the rice so that you have any concern about the consumer for sure.

I think that listen, we know that they've spent through all the stimulus money, right, that they're not getting the pay raises in all the overtime that they were a couple years ago. Um and and we know that when you look at their credit, they're really spending a lot on credit now. And we see that same data on auto delinquent es.

And so if you just go through the things that trump says he's gonna do, right, imposing tariffs bad for for retailers like target. So that's that's one side of the equation. Then on the other side, you have somebody like snowflake or palo alto report tonight.

They beater means, but expectations are low, snowflakes up twenty percent and after hours. So just think about those that spread trade targets down twenty snowflakes up twenty one as super high expectations, super high valuation, one as super low expectations. And in the case of a company like snowflake, there's massive room for efficiency, right? They're getting the efficiencies from my ee.

The they can they don't need to hire at the rate that they were hiring before, and they still just grew twenty nine percent in the order. Um you know so that that far from the massive declaration that many people were projecting. And so I do think that the beneficiaries and technology, they're both beneficiaries of this top line growth but also bottomline margin expansive.

You know I was looking back in preparation for this ah for the pod was saatchi on monday. Since he became CEO of microsoft, the net income margin has expanded by a thousand basis points. You right like if you look at the story of microsoft last ten years, revenue is doubled, but earnings have almost quite dropped because of the expansion of of the earnings margin in that business.

And that's what I think we're still going to seen in technology. Um you know and it's gonna accelerated with a eyes. So I had in the next year, not all in, but i'm definitely constructive on the backdrop.

I just think it's gonna a stock pickers market. It's going to be a bunch of stuff that you want to avoid. But you can find you know your five, ten, twenty tech companies that are enjoying the tail winds of AI and also enjoying the possibility of this margin expansion. I think they can continue to do reasonably well. So that's our positioned.

I'm here. You says you can kind of have your taking needed to if you pick the right courses here, like even if the consumer has some struggles, you get to your portfolio line in the right way, you're going to you're going to be able to maneuvre through that.

Yeah, another way i've saying that I don't expect a twenty five percent up here for abroad based index, you know and if you look at IT, I don't know where IT is as of today. But there is one point in the year, bill, that if you took the top ten tech companies out of the S M P, right, the S M P was not up twenty, but I was actually down five yeah right.

And I think that's the type of diversion that I think we're going to see again next year that they're going to to be a lot of winners um either as the result of changing government regulations, changing government policies, lower taxes, always the result of this tex super cycle. But there are going to be a whole lot of losers, right? We just talked about the ones in fsd.

IT may very well be that this is terrible for a lot of you know, OEM, but really great for tesla. We have to yet to be see, i'll tell you as an investor, and we could leave IT on this. The rate of change has never been higher in technology.

And the rate of change of this election is massive, right? The magnitude of the policy change is massive. You combine those two things and let's just say we've been working overtime.

Make sense to me. One of the things we may want to talk about in the future is, is whether truck administers gonna lean in the AI regulation as well, which could could have implications for everyone around anyway.

thanks on do you two.

As a reminder, everybody just our opinions, not investment advice.