저자: CathieWood · 외국인
요약 보기
캐시 우드는 테슬라가 자동차 하드웨어(마진 15%)에서 로보택시 기반 SaaS형 모델(마진 70~80%)로 전환 중이라고 주장하며, 옵티머스 휴머노이드는 로보택시보다 약 20만 배 어렵다고 평가함. 다만 2028년 말 유연한 휴머노이드 가능성은 언급. 진행자 Gene Munster 인용으로 AI 추론 시장이 2025년 학습의 5배이며 향후 수백~수천 배 커질 수 있다고 하고, 하이퍼스케일러 capex 성장률 상향 및 소형주 AI 수혜를 강조. 로보택시 확산이 예상보다 빠르다는 견해.
다듬은 글
These are the kind of videos I love. I've got a huge show for you today as Kathy Wood lays out exactly why Tesla's pivoting so hard into robo taxi. And then we have Gan Monster who drops a stat which will blow your mind. And stay tuned as at the end we have a bonus clip from Kathy and Brett Winon from Arch Invest discussing Optimus.
This is a juicy one. Tesla's always been one of your big holdings. You know, in terms of just sales, obviously things have been challenged, but there's a lot of excitement about autonomy. There's a lot of excitement about the Optimus robot. What is your what is your sense of where things stand at Tesla right now?
The EV sales environment has been challenged, but the market is shifting its focus to the robo taxi opportunity. So we're moving from 15% gross margins in autos the hardware to more of a SAS model where robo taxes recurring revenue model is those margins are more in the 70s and 80s and I think as analysts understand that that's where Tesla is going and they start uh building in the model for robo taxis they see they see a tremendous opportunity.
So, it's been interesting to me to see upgrades even by some auto analysts now that they're beginning to understand that this is not an auto company. Maybe they're working with their tech analysts and understanding the stock a bit better. So, when you think though about autonomous uh robo taxis and you think about um robots, what what does that look like?
and and some of the critics would say cars aren't working the way they're supposed to be working right now in terms of what's happening with the EV market and some of these other ambitious efforts are going to take a long time to get to a meaningful escape velocity. Well, we're keeping an eye on how quickly Tesla's footprint is expanding here and it's expanding a lot faster, I think, than many people expected.
And it's interesting to watch Whimo as well. We draw in Texas. We draw the footprint and they're competing with each other. Competition is a good thing. Uh so I think this is actually going to happen faster than most analysts think in terms of the proliferation of robo taxis especially if we get more federal legislation instead of just state byst state.
On the Optimus robot side, there's a lot of excitement in robotics that in the private markets has been a huge space this past year. We look at an Optimus robot. Our chief futurist Brett Winton working with Sam and Daniel have um concluded that an Optimus robot is going to be 200,000 times more difficult to get working correctly relative to relative to a robo taxi.
You'll have waves of Optimus improving with time, but full human is is going to be quite difficult. Nonetheless, the robots that we're seeing, they're starting to make waves. It's January again, and by now you've started making resolutions to make the new year better for you. But the smartest people are already on the road to master that one skill that matters most in 2026.
AI. From just a simple text model in 2019, AI has evolved to detecting diseases, automating work, and even running tasks in your sleep. And that's why I want to introduce to you the sponsor of this video, Outskill. They are hosting a two-day AI mastermind training session live this Saturday and Sunday, 10:00 a.m.
to 700 p.m. Eastern on both days. And right now is the perfect moment to join because you can get in absolutely free as part of the new year upskilling fest. This 16 hours AI training has already built 10 million plus AI first professionals worldwide and is rated 4.9 out of five on Trustpilot. This is where you learn how to build AI agents that plan, write, execute, and report for you.
Automate workflows that even run while you sleep. Use AI to save hours every week and get an unfair advantage at work. Not just that, you'll also learn how to profit from these skills. Seats are limited. Use the link in the description to join. Kathy says Tesla's robo taxi business transforms 15% hardware margins into 70% plus software margins, and Optimus is 200,000 times harder to get working.
So, you're going from a car business with 15% gross margins, which is actually really good for the automotive space to something that looks more like a SAS company. 70% 80% margins on robo taxi revenue. And I think that's why you're seeing automotive analysts suddenly get interested in Tesla again. Right now, Tesla sells you a car and that's mostly it.
Maybe some software upgrades and some supercharging revenue. But a robo taxi fleet, that's revenue every single ride forever. The car becomes an asset that generates cash instead of something that depreciates in your driveway. And Tesla doesn't need to convince you to buy a $50,000 vehicle anymore. They just need the ride to be cheaper and more convenient than the alternatives.
Now, the Optimus stuff is interesting, too, because ARK actually tried to quantify how hard it is. Brett Winton and the team have concluded that getting a humanoid robot to full capability is 200,000 times harder and more difficult than robo taxi. Think about what a robo taxi has to do. Navigate roads, follow traffic laws, handle weather.
Hard but bounded. A humanoid robot is in your house. It needs to handle literally anything. Open any drawer, fold any shirt, navigate any room. The edge cases are essentially infinite. That math explains Tesla's prioritization. You've got one opportunity which they're on the cusp of solving robo taxi and achieving meaningful revenue with incredible margins and another with even bigger long-term potential.
That's Optimus, but it's still in the earlier development stages. Roboaxi is ready to scale now. Optimus is the next wave. Doesn't mean it's not valuable. If you're trying to figure out what moves the company forward in the next few years, it's Roboaxi. The robots you're seeing now, the pick and pack stuff Kathy mentions, that's wave one.
full human capability comes later and Tesla's building towards it. Gene monsters are making the exact same argument about AI infrastructure. The market's measuring the wrong thing. Everyone's focused on training spend slowing down, but in France actually using these models is already way bigger and it could get a lot bigger than that.
Check out this clip from Jean. Anyway, so if that's all the case and you mentioned that the AI trade has passed two hurdles and that it seems to still this tide be lifting all boats here. I mean there are some firms like your Denny research which have underweighted um certainly the mag 7. Um what do you think is the third hurdle this year and are you sticking with tech broadly or or more specifically?
So I think that the third hurdle I guess more coming up more recently of course we had the good um news from TSMC. We had the good commentary from uh Nvidia at CES but the third is going to be about capex guidance. Just a quick view on that. It was 30% growth is what the hyperscalers analysts were expecting hyperscalers to grow in 22 and 2026.
30% growth. That was the expectation a few weeks ago. That's now inched up to 34% and so that's moving higher. That's the next hurdle of course at the end of this month. But just a couple quick thoughts. First in terms of the uh kind of what's going on the bigger picture why I think that we'll get good news later this month.
I think those capex numbers are going to be probably plus 40%. is investors are making a fundamental mistake when they think about this AI infrastructure buildout. The conversation is wrapped around the idea of infrastructure for training and the reality is is that the inference market Jensen talks about this but we have an affiliate company intelligent alpha which uses a lot of these models.
We see how long these models are thinking. that inference market uh last year in 2025 was five times as big as the training market over time this could be hundreds thousands hundreds of thousands of times bigger so this party is going to continue longer as far as kind of the small versus large cap piece uh I made a prediction earlier stand by it this back in December that small cap's going to outperform we have an ETF LOUP which focuses on those small cap uh you saw it this week AMD up 14% on that TSMC Why is that?
Why is that? Just because the the gains have kind of run their course for the big guys. That's one piece to it. But if you just look at if you're a believer, ultimately it comes down to do you believe that this infrastructure piece that comes back to that inference. If you believe that that's the case, just mathematically these smaller companies should have higher greater earnings revisions.
And so uh still a big believer. I think Apple's going to be a wonderful performer in the first half of the year, Google for the year. But I think this small cap trade, it it happened last year. I think it's going to continue throughout this year. The inference market was already five times bigger than training last year and could scale hundreds of times larger.
Genie is saying that when Wall Street talks about AI infrastructure spending, the conversation is almost always about training, building bigger models, more parameters, more compute to create the next GPT. But inference, the actual using of these models was five times bigger than training in 2025. Five times.
The math here gets kind of crazy when you extend it out. Training a model is a one-time thing. Expensive, sure, but you do it once and you're done. Inference happens every single time someone uses the model. Every chat GPT query, every co-pilot suggestion, every AI generated image. Multiply that by billions of users and you start to understand why Gene saying inference could be hundreds or thousands of times bigger than training eventually.
This is why Gan thinks hyperscala capex guidance is going to surprise to the upside. Consensus moved from 30% growth to 34% growth over the last few weeks and Gene is saying it could hit 40% because the analysts are building models based on training demand flattening out but inference demand is just getting started.
Microsoft, Google, Amazon, they're building because people are actually using this stuff. And yeah, this has implications for where you want to be invested. The Max 7 has already priced in a lot of the AI story, but smaller companies with AI exposure, they potentially got more room for earnings revisions because the street hasn't fully modeled what inference demand does to their numbers.
Jean still likes Apple and Google, but he's saying the small cap AI trade that worked last year probably continues. The tides lifting, smaller boats have room to rise. And now, as promised, here's the bonus clip from Kathy and Brett. If you're enjoying this video, please consider subscribing. We're really trying to hit 50,000 subs.
Enjoy this clip. I'm gonna give a plug to our research. It's been really interesting for me to watch uh our humanoid robot um research evolve because we've been saying this is really hard. It's really hard. It's 200,000 times harder or maybe I'm exaggerating than a robo taxi. Uh and this is going to take a long time.
people are overestimate and yet I think given our latest research haven't we determined that it's possible especially if Elon's driving uh the program to to have a humanoid robot um fairly flexible humanoid robot uh in 2028 that's not that long I would say late late 2020s is possible but there's a there you I mean and Sam you can jump in here too but that is preconditioned on you're throwing a hundred billionish dollars at AI compute to to as in
and and that's what's magical about markets it's kind of if people believe it then people will fund it if enough people fund enough of them then one of them will kind of like figure out how to make it work and then be able to attract the capital to make it work uh and if people don't believe it at all then you will never actually push money into the game and so it will never work.
Now Elon is like a unicorn here because he has he has a cash flow generation engine that can allow it to self-fund regardless of the the kind of like consensus belief. Right now we're in a lined up period where people are very excited by it and it requires a lot of capital and so there's a lot of companies that are like making bids but there will be a winnowing I can guarantee that.
Oh there absolutely will be. But just I wanted uh to say that um I think this is telling us something about risk appetite generally starting to increase uh because chat GPT moment AI more and more people casting around saying okay this is going to have significant ramifications I better I better I better be positioned right.
Mhm. And I think that the the ROIC potential here is absurd, right? If you can produce a humanoid robot that does for $25,000 that what what's the productivity delivery if it's human like Sam in operating whatever we model 20 24 hours? Yeah. Yeah. It's Well, it's like the human full labor cost over 10 years is like $500,000.
So, right, take a robot and say, you know, even call it $100,000, you know, there's 5x, but then that's like 5x raw, not including the Yeah. Yeah. Cash for cash. Yeah. You know, there's there's there's like moreover more complexity than that. But the net economics are so obvious that then people are willing to take a risk on if I can deploy capital to get that thing to market.
It's clear that it's an amazing it's worth the upfront capital thrown in. And that's I think where we are both on the AI side clearly the the the foundation model companies are making that bet because they're not just paying for inference. they're uh committing to you know tens and hundreds of billions of dollars of spend on underlying compute to power the next models because there are we think and I many agree trillions of dollars of opportunity in kind of like supercharging productivity of knowledge workers humanoid robot it's the same thing on building physical stuff in the world there are trillions you know maybe even tens of trillions of opportunity in delivering like a truly working human form factor so Then in that context, what's hundred billion dollars?
You know, it's it's a a worthy risk for somebody who's the richest guy in the world. And if he's going to try to do it, then, you know, funders are going to try to find others that can compete with him. So, I think that that's it's both risk appetite, but it's also the size of the potential returns here are unprecedented.
And so, of course, that's motivating capital in it's taking capital from really kind of poor uses of capital into more interesting uses. YouTube isn't just entertainment. It's one of the best client acquisition tools because it builds trust at scale. We've helped businesses grow from scratch to a $100,000 month just by launch and let's see how YouTube could help your business scale.
Comments 0