日本には技術があるのになぜ勝てないのか。自動車からロボットへ移る世界競争の正体とは【マイキー佐野 投資学】 — Transcript
Full transcript
- 0:02It's no longer just about how many cars
- 0:04a manufacturer sells; the current
- 0:06reality is that they're now factoring
- 0:09in how many robots they can deploy into
- 0:11society.
- 0:13Hi, everyone, Mikey here. I'm filming
- 0:15another YouTube video today.
- 0:16Let's do this.
- 0:17Let's get started.
- 0:18We’ve touched on robotics and robots
- 0:20quite often on this YouTube channel.
- 0:22For example, with the situation between
- 0:25the US and China, I have a question:
- 0:27why are automakers like Tesla in the US
- 0:29or Toyota in Japan working on things
- 0:32like smart cities and robotics? Yeah.
- 0:35It’s a simple question—why are car
- 0:36manufacturers making robots in the
- 0:38first place?
- 0:39Do you know the name of the Tesla robot
- 0:40? It’s got a pretty edgy name.
- 0:42Wait, hold on. It's not the Terminator
- 0:44one, is it?
- 0:45More like Transformers, maybe.
- 0:47My apologies. What was it again?
- 0:49It's Optimus. Besides them, companies
- 0:53like Xpeng, Xiaomi, and Hyundai in
- 0:55South Korea are also developing
- 0:57humanoid robots. We're moving into the
- 0:59physical AI era, and I was just
- 1:01watching those Chinese robot sports
- 1:02competitions—I almost died laughing.
- 1:04Yeah. It's pretty great, isn't it?
- 1:06Yeah, it's absolutely hilarious.
- 1:08So, they’re developing humanoid
- 1:10robots while simultaneously moving into
- 1:13mass production. What’s happening now
- 1:15is that the criteria for evaluating car
- 1:18manufacturers is being redefined.
- 1:22It’s getting harder to evaluate them,
- 1:24right? That’s exactly what it is. The
- 1:27reason for this is that an increasing
- 1:29number of automakers are entering the
- 1:31robotics field, especially the humanoid
- 1:33sector.
- 1:35Right. There’s actually a technical
- 1:40and operational necessity for this,
- 1:42such as leveraging the synergies with
- 1:44AI architectures developed for
- 1:45self-driving cars, or utilizing their
- 1:47manufacturing strengths from their own
- 1:49factories. To put it simply, the next
- 1:54generation of EV competition is all
- 1:56about Level 3 and Level 4 autonomous
- 1:58driving. For instance, even though many
- 2:03Japanese people say they don't need EVs
- 2:05or self-driving tech, the key point is
- 2:07that the same underlying technology is
- 2:09used. It uses things like onboard
- 2:14cameras and LiDAR to analyze visual
- 2:16information. It recognizes the
- 2:18environment in real-time, finds the
- 2:20best path, and moves accordingly. It's
- 2:22just like a car, isn't it? It’s very
- 2:24similar, right? By training things like
- 2:28end-to-end deep learning models or
- 2:30Vision Language Models, we can actually
- 2:32apply self-driving tech to robotics.
- 2:38These autonomous AI algorithms for
- 2:40perception, reasoning, and
- 2:41decision-making are mathematically and
- 2:44structurally quite similar to the human
- 2:46brain. For automakers, AI is used for
- 2:51things like self-localization. It’s
- 2:55called SLAM, or sometimes an "occupancy
- 2:57network," which refers to AI models
- 2:59that continuously predict what exists
- 3:01in a space on a grid-by-grid basis.
- 3:07Technologies that reconstruct 3D space
- 3:09directly from camera feeds or LiDAR are
- 3:12becoming widespread, seen in things
- 3:14like Tesla's FSD or the autonomous
- 3:16systems from Chinese manufacturers.
- 3:21This is robotics technology. Yeah. So,
- 3:26the biggest hurdle in developing this
- 3:28AI and robotics is collecting data from
- 3:31the real world. The AI most people use
- 3:33is basically based on internet data,
- 3:35right?
- 3:35That's right.
- 3:36But autonomous vehicles are constantly
- 3:37moving through the physical world,
- 3:39aren't they? So, the background is
- 3:40shifting, and automakers clearly have
- 3:42the advantage when it comes to building
- 3:44robots.
- 3:45Oh, I see what you mean.
- 3:46Yes. Because they can gather real-world
- 3:48data. For instance, in their own
- 3:51factories or automated EV assembly
- 3:53lines, they are already running tests
- 3:55with humanoid robots during the
- 3:57development phases. That's why
- 4:00automakers are incredibly
- 4:01well-positioned, aren't they? They have
- 4:03the world's largest scale for
- 4:05verification already in place, which is
- 4:07a massive strength.
- 4:09By deploying robots into various work
- 4:11environments—like transporting parts,
- 4:14quality inspections, screw fastening,
- 4:16or component sorting—they can collect
- 4:18vast amounts of data within the factory
- 4:21, keeping everything in-house and
- 4:23running at high speeds. At least from
- 4:26an automaker's perspective. Tesla is
- 4:29making Optimus, but EV makers in China
- 4:31and Korea are also leveraging their
- 4:33strengths to take various approaches to
- 4:36building robots. But even though
- 4:37Japanese companies are building them,
- 4:39there isn't much news about it.
- 4:40Regarding Japan specifically, for
- 4:42example, China is incredibly strong
- 4:44when it comes to robotics, right?
- 4:47There’s a Chinese automaker called
- 4:49Xpeng, and they’ve positioned
- 4:50themselves as a "Global Physical AI
- 4:52Company." They are carrying out a
- 4:55complete vertical integration of
- 4:57robotics alongside their autonomous
- 4:58driving technology. In their Series A
- 5:01alone, Xpeng brought in Alibaba and
- 5:03Tencent, raising 900 million dollars,
- 5:05which is one of the largest funding
- 5:07rounds in China's physical AI sector.
- 5:11They plan to start mass production of
- 5:14their physical AI by the end of 2026,
- 5:16and aim to have them in commercial
- 5:18stores, showrooms, and offices for
- 5:21full-scale use by 2027. And this robot,
- 5:24well, it’s got that "chuunibyou"
- 5:26flair. What?
- 5:26What is it?
- 5:27It's Iron.
- 5:28With "Man" attached.
- 5:30So it's Iron Man. Xiaomi is an
- 5:32automaker too, though they were
- 5:33originally a phone manufacturer, right?
- 5:35And this is what became a car
- 5:36manufacturer. They're combining their
- 5:38know-how from smartphone mass
- 5:39production with the manufacturing speed
- 5:41learned from EVs to create a humanoid.
- 5:43That’s also a pretty great name,
- 5:44isn’t it?
- 5:45This is CyberOne.
- 5:46It's cool.
- 5:47And after that, there’s this
- 5:49quadruped robot, which they call
- 5:51CyberDog.
- 5:52They aren't overthinking it at all.
- 5:54Xiaomi is incredibly strong in IoT, so
- 5:56they’re integrating people, cars, and
- 5:59homes into a single OS. They’re
- 6:02incorporating robots into that, aiming
- 6:04for both the automation of their smart
- 6:06factories and the consumer market. Now,
- 6:08take Hyundai from Korea.
- 6:10Right.
- 6:11They originally started out building
- 6:12heavy machinery. They switched
- 6:14completely to electric, acquired Boston
- 6:17Dynamics, and now possess top-tier
- 6:19dynamic control technology for walking.
- 6:21They’re now deploying these in
- 6:23next-gen smart factories in places like
- 6:25Singapore, showing how Hyundai is
- 6:27pursuing vertical integration through M
- 6:29&A. Startups like BYD and Geely are
- 6:34also partnering with humanoid makers
- 6:36like UBTECH or Agibot, bringing them
- 6:39into their EV assembly lines to
- 6:40accelerate the practical use of
- 6:42industrial robots. So, why are car
- 6:49manufacturers pinning their hopes on
- 6:51the robotics sector? Well, they’re
- 6:53under pressure from price competition.
- 6:56Compared to the auto industry, the
- 6:57humanoid robot business is expected to
- 6:59have overwhelmingly higher gross
- 7:01margins and business growth potential.
- 7:03Car manufacturing involves thousands or
- 7:05even tens of thousands of parts, which
- 7:07costs an incredible amount of money.
- 7:09Plus, with strict safety standards in
- 7:11place, it’s a capital-intensive
- 7:13industry with thin margins. You’ve
- 7:15probably heard that even Chinese
- 7:17manufacturers need to sell hundreds of
- 7:19thousands of units to turn a profit, as
- 7:21EV makers only have a vehicle gross
- 7:23margin of about 10%to 20%. So, for
- 7:25companies like Xpeng, their gross
- 7:27margin on automobiles—well, these are
- 7:29EVs—is around 12%. Compared to that,
- 7:31humanoid robots actually offer higher
- 7:32added value and carry a premium, which
- 7:34is why they’re said to have higher
- 7:36profit margins. Xpeng says they can
- 7:38achieve a gross margin of over 50%.
- 7:40That’s impressive, isn’t it?
- 7:41Exactly. I see. Well, a company that
- 7:43recently went public, which you might
- 7:44have seen on the news, is Unitree.
- 7:46I've been saying for a long time that
- 7:47this company is great, and they just
- 7:49listed in China. As for their gross
- 7:51margin in 2025, it’s going to be 69%.
- 7:54They're the company famous for making
- 7:55those quadruped robots that can walk
- 7:57down stairs.
- 7:58Yes. Also, a huge portion of robotics
- 8:00costs comes down to the actuators. The
- 8:03motors, the speed reducers, the screws,
- 8:05and the encoders. These integrated
- 8:07components are what I'm referring to.
- 8:10They account for 50%of the total, but
- 8:12with manufacturing scaling up and the
- 8:14supply chain maturing, what’s
- 8:16happening is that Chinese manufacturers
- 8:18are cutting costs significantly. Since
- 8:22things like actuators, speed reducers,
- 8:25and encoders make up 50%, their falling
- 8:27costs are boosting overall profit
- 8:30margins. Actually, Goldman Sachs and
- 8:32Bank of America have been analyzing
- 8:34this. The cost itself was around $
- 8:3635,000 as of last year. This is
- 8:38expected to drop by half to under $
- 8:4117,000 by 2030. Especially in China.
- 8:44The reduction in cost drives up margins
- 8:46, and it’s predicted that this can be
- 8:48maintained for some time. So, why pivot
- 8:50to robotics?
- 8:51Well, the robotics business is
- 8:53overwhelmingly different from the
- 8:54automotive business. And that is
- 8:56because it’s a one-off sale. You sell
- 8:58it and that’s it, right? To keep
- 9:01increasing profit margins, you need a
- 9:03business model that is multi-layered
- 9:05and allows for recurring revenue. Take
- 9:08EVs, for example; there's a reason they
- 9:10are so focused on software—it’s
- 9:12because they can charge you every time
- 9:14there’s an update. The automotive
- 9:17industry is constantly pressured by
- 9:19price competition based on specs and
- 9:21the commoditization of vehicles. For
- 9:25industrial or commercial robots, the
- 9:28selling price isn't based on specs, but
- 9:30rather on how much they reduce human
- 9:32labor costs. The robot itself, that is.
- 9:36And that could be the future trend. If
- 9:38it offers a cost-effective alternative
- 9:40to annual labor expenses, people will
- 9:42buy it. Since you can recoup the
- 9:44investment quickly, it creates a real
- 9:46incentive. You hear terms like SaaS a
- 9:49lot these days, but there's also RaaS.
- 9:51Mm-hmm. Robot-as-a-Service.
- 9:53I see.
- 9:54Software and skill-based billing.
- 9:56Yes.
- 9:58Subscriptions become the core,
- 10:00generating continuous cash flow, where
- 10:02you pay extra to add specific skills or
- 10:04features to expand functionality.
- 10:08Oh, so it becomes an add-on system.
- 10:10That’s why people expect
- 10:11manufacturing companies to hit profit
- 10:13margins similar to software companies.
- 10:16Plus, with manufacturing costs dropping
- 10:18and the ability to run 24/7, you can
- 10:21cut down on labor costs as well. That
- 10:23leads to a steady increase in revenue.
- 10:25It’s not just robots fighting robots;
- 10:27it’s really becoming one human versus
- 10:29one robot.
- 10:30Well, it is a replacement for part of
- 10:31the work.
- 10:32Ah, right.
- 10:33Yes.
- 10:35And what you need to understand is the
- 10:37importance of the components. It is a
- 10:39fact that automakers are creating
- 10:41significant barriers to entry for robot
- 10:43companies. The reason is, as I
- 10:45mentioned, they’re collecting tons of
- 10:47data, including real-world data, and I
- 10:49talked about how there’s a massive
- 10:51field for testing. But when it comes to
- 10:54parts, the hardware and software
- 10:56components of EVs and humanoids
- 10:58actually overlap quite a bit.
- 11:00Oh, I see.
- 11:01In terms of the hardware supply chain,
- 11:03for example, things like automotive
- 11:05lithium-ion batteries and battery
- 11:07management systems—the main power
- 11:09modules can be repurposed directly. For
- 11:12instance, the torque stators, rotors,
- 11:15and power inverters developed for EV
- 11:17electric drivetrains form the
- 11:19foundation for what we call frameless
- 11:21motors, which are used in robot joints.
- 11:24Then there’s also neodymium, which is
- 11:26necessary for torque actuators. These
- 11:28are magnets, and they’re also used in
- 11:30materials like rare earths or in
- 11:32braking systems. Another thing is that
- 11:34reducers can be integrated into the
- 11:35same supply chain as the EV business.
- 11:37So, procuring in large volumes brings
- 11:39down manufacturing costs, right?
- 11:41That’s another advantage as well. You
- 11:43know how there’s this whole debate
- 11:44about whether EVs are necessary or not?
- 11:46Those people are only looking at EVs,
- 11:48but in reality, it’s worth doing EVs
- 11:49precisely because the technology can be
- 11:51repurposed for the robotics industry.
- 11:53That’s definitely true.
- 11:55Right? That’s why people complaining
- 11:57about EVs are only looking at cars. And
- 11:59I’ve been saying all along that EVs
- 12:00are necessary. The reason is, as I said
- 12:02, it can also be used for the robotics
- 12:04industry. And the robotics industry is
- 12:06an incredibly strong sector in Japan.
- 12:08So, as EVs become more popular and
- 12:10continue to evolve, it creates a flow
- 12:12where the supply chain for robotics can
- 12:14actually be secured as well.
- 12:16Yeah. It’s a package deal.
- 12:17Yes.
- 12:18I understand now.
- 12:19And the interesting part is, as I
- 12:21explained earlier regarding the
- 12:23software AI, EV autonomous driving
- 12:25systems and humanoid control are often
- 12:27built on the same technical foundation.
- 12:30They use cameras for real-time spatial
- 12:32awareness and sensors to recognize
- 12:34obstacles, so the control algorithms
- 12:36used are often the same. However, they
- 12:38aren't exactly the same. To put it
- 12:40simply, robots are more difficult.
- 12:42Yes. Yeah. Yeah. Yeah.
- 12:44In the case of an automobile, it's
- 12:45actually very simple. As long as you
- 12:46don't hit anything, you're doing it
- 12:48right.
- 12:48That's true.
- 12:49But for robots, you need something more
- 12:51like contact mechanics. You have to be
- 12:53able to change your force according to
- 12:54the environment. For example, what
- 12:56happens if you try to hold an egg with
- 12:58the same force you'd use to hold a
- 12:59cardboard box?
- 13:00Because of that, contact mechanics are
- 13:02incredibly difficult. In comparison, a
- 13:03car just needs to avoid collisions. And
- 13:05follow the rules.
- 13:06It really depends on the purpose,
- 13:07doesn't it? Various things. Yes.
- 13:09Yes. Also, remember that the number of
- 13:11moving parts is different. For a car,
- 13:13you just have the brake, the
- 13:14accelerator, and the steering wheel.
- 13:16But robots are designed to have the
- 13:17same shape as humans, right? They have
- 13:19over 50 of them. Like the joints. For
- 13:21example, fingertips, wrists, elbows,
- 13:23shoulders, hips, knees, and ankles.
- 13:25Because these need to be coordinated
- 13:27simultaneously, the calculations
- 13:28increase astronomically. There's also
- 13:30the aspect of balance and speed; for a
- 13:32car, you're basically just turning four
- 13:35tires steadily on the ground. So, even
- 13:37if the AI's judgment is delayed by 0.1
- 13:39seconds, the car won't just flip over
- 13:41on the spot. But for a robot, you have
- 13:43to make it walk bipedally and
- 13:45coordinate over 50 parts at once. So,
- 13:47if it’s about to fall, you need
- 13:49techniques to catch it at the last
- 13:51second, making the continuous control
- 13:53much more severe than for a car. And
- 13:55camera movement is also extremely
- 13:57important, whereas a car's body is
- 13:59naturally fixed in place.
- 14:00Correcting camera shake in video is
- 14:02relatively easy. But with robots, they
- 14:04move up and down violently when they
- 14:06walk, don't they? So having to make
- 14:08decisions while your vision is shaking
- 14:10is extremely difficult. Yeah. Robots
- 14:13are...
- 14:15As automakers get serious about
- 14:17robotics and physical AI, the way the
- 14:19stock market evaluates these companies
- 14:22is actually changing. It’s already
- 14:24happening. So no matter how much Toyota
- 14:26increases its sales, they aren't really
- 14:28doing EVs, right? They aren't into
- 14:29robots, right? They end up being seen
- 14:31as having a low valuation. Investors
- 14:34apply specific multiples to the
- 14:35businesses a company develops. That's
- 14:38why traditional automakers are seen as
- 14:40having limited growth and being
- 14:42susceptible to cyclical trends. So, for
- 14:44example, look at the P/E ratio. The
- 14:46price-to-earnings ratio is usually
- 14:48around 5 to 10 times, and the P/S ratio
- 14:50is only about 0.3 to 1 times. Meanwhile
- 14:53, AI platformers have P/E ratios of 30
- 14:55to 100 times. Their P/S ratios are also
- 14:575 to 15 times higher, as they are
- 14:59easily awarded a premium valuation. So,
- 15:01Chinese manufacturers like XPeng are
- 15:02really starting to ramp up sales, but
- 15:04compared to Toyota, their revenue is
- 15:06still overwhelmingly lower. I can
- 15:08understand why Chinese manufacturers
- 15:09suddenly get such massive valuations,
- 15:11making them worth more than companies
- 15:13like Nissan or Honda. This is exactly
- 15:16where the premium and multiples come in
- 15:18, because they aren't just seen as
- 15:20making cars; they're viewed through the
- 15:22lens of EVs, AI, and robotics. So, it's
- 15:27not about how many cars an automaker
- 15:29sells anymore; once they are viewed as
- 15:31physical AI or robotics companies, the
- 15:34market begins to factor in how many
- 15:36physical AI platforms they can deploy.
- 15:39So, there are people who say, "You
- 15:41don't need to do EVs at all," but do
- 15:43you think they haven't thought about
- 15:45anything? I really thought they hadn't,
- 15:47but now it makes perfect sense. I see.
- 15:49So, automobile manufacturers are
- 15:51basically stuck with the automobile
- 15:53business model. Even when it comes to
- 15:55EVs.
- 15:55But for humanoid robots, you can get
- 15:57high profit margins from the early
- 15:59stages, especially with what I
- 16:01mentioned earlier: subscription-based
- 16:03billing. Because of this, it's said
- 16:05that a lifecycle with 60%to 70%profit
- 16:08margins could be quite sustainable.
- 16:10Even among experts. So, by reaching
- 16:13into those areas, the future potential
- 16:15and profit margins are clearly higher.
- 16:17They are seen as having growth
- 16:18potential, which is why they are valued
- 16:21so highly. That’s what I really love,
- 16:23honestly. Wait, are you talking about
- 16:24Optimus and things like that?
- 16:26Iro—
- 16:26Iron.
- 16:27Iron Man or CyberOne, those look cool,
- 16:29right?
- 16:29Man, Toyota’s design is lame. W-what
- 16:32is this? Pu-Punyo? T-HR 3.
- 16:34I really wish they would lean more into
- 16:35something like Marvel.
- 16:36Yeah. That’s what Japan is missing.
- 16:37Yes. Only guys probably get this.
- 16:39That's true.
- 16:40Yes. With that, thank you all for
- 16:42watching until the end today. I’ll
- 16:43summarize today's discussion in a
- 16:45report, so if you'd like to receive it,
- 16:47please check out my official LINE
- 16:48account. Also, if you found today's
- 16:50video interesting, educational, or
- 16:52helpful, please subscribe to the
- 16:54channel, leave a comment, and share it.
- 16:56Today, just like the past two weeks
- 16:58before filming, we were watching the
- 17:00Chinese robot Olympics videos.
- 17:01Those are seriously fun.
- 17:03They really are. I definitely want our
- 17:05viewers to watch those, too.
- 17:06Maki-san, what impressed you the most
- 17:08when you watched them?
- 17:09The ones that get back up when they
- 17:11fall.
- 17:11Ah.
- 17:11That 100-meter dash where they hit the
- 17:13wall, get destroyed, and then
- 17:14extinguish themselves with foam—that
- 17:16was super fun to watch. It actually
- 17:17reminded me of something. AI, the
- 17:19Spielberg one...
- 17:20Ah, yes. The one with Haley Joel Osment
- 17:23. Yeah, yeah, exactly, it reminds you
- 17:25of those robots all piled up in a heap.
- 17:27And there are the ones where they fight
- 17:28, right? The ones wearing pants and
- 17:30boxing.
- 17:31That’s definitely going to be a real
- 17:32sport in ten years.
- 17:33I bet it will. Yes.
- 17:35Back in the day, when I was in high
- 17:37school in America, robots...
- 17:38So, well, they're like
- 17:39remote-controlled toys, but these
- 17:41various robots are outfitted with
- 17:42weapons and gears, battling it out
- 17:44until they beat the opponent's robot
- 17:46into submission.
- 17:47Oh, like the beginning of Big Hero 6.
- 17:49It was super fun to watch. Oh, there
- 17:51was like this
- 17:52show that had that.
- 17:53Right. Huh. That's interesting.
- 17:54Back in high school, I did stuff like
- 17:56programming robot dogs, and we'd bring
- 17:58robots to class and work with them like
- 18:00this. It was really interesting.
- 18:02Yeah, yeah, yeah. Anyway, everyone
- 18:04should watch China's Robot Olympics;
- 18:05it'll definitely make Mikey's robot
- 18:07modification videos even more exciting,
- 18:09so please go check it out.
- 18:10If you had to recommend a robot movie,
- 18:12what would it be?
- 18:12Well, I like A.I. like I mentioned
- 18:14before, but for this YouTube channel,
- 18:16let's go with Transformers.
- 18:17I'm more of a...yeah, yeah, yeah.
About this transcript
This page contains the full transcript of 日本には技術があるのになぜ勝てないのか。自動車からロボットへ移る世界競争の正体とは【マイキー佐野 投資学】 by マイキーの非道徳な社会学, generated from the public captions YouTube serves with the video. The transcript has 3,064 words across 549 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.