9 min read

Physical AI, Korea and Autoparts companies

Source: http://x.com/i/article/2057649049986371584

tldr: Korea is already being repriced as an AI memory market. But I think the next leg could be Physical AI: robots, actuators, reducers, and the boring precision-machining companies sitting underneath Hyundai’s robotics ambitions.

"The next wave is physical AI" - Jensen Huang, NVIDIA CEO
"the next frontier is going to be the physical AI" - Lip-Bu Tan, Intel CEO

That's a pretty strong signal if you ask me.

Physical AI seems to be where the ball is heading and of all places, I'm looking into the Korean market.

Background for searching in Korea

KOSPI outperformed every other major asset class over the past two years by a huge margin. It is currently up an astounding 2.8x.

Image

It's got to be a bubble right?

Image

https://x.com/jukan05/status/2057634950636241129?s=20 There are definitely signs of froth, but it's still one of the cheapest equities markets in the world with a forward P/E at 8x, even below it's 10 year average of 10.5x.

Image

April 2026, @DeutscheBank So Korea is still cheap???

Let's look into what might be going on.

AI and Memory

The first reason Korea came back on global investors’ radar is simple: Korea makes memory chips.

And memory is now one of the biggest bottlenecks in the AI stack.

Samsung and SK Hynix sit right in the middle of that trade. Memory prices have ripped, HBM demand is exploding, and margins are expanding fast.

Samsung and Sk Hynix are expected to be the global top 5 by operating profit this year. $400b in operating profits.

Image

👉 Foreign capital flows into Korean capital markets

But there's more to the story.

The Korean Policy Angle

One underrated Korea angle is policy.

President Lee stepped into office June 4th 2025 after the tumultuous

martial law crisis instigated by President Yoon.

Stock markets have been straight up ever since. It's marked in the chart above.

Koreans love(d?) Real-estate

For decades Korean households have parked their wealth in real-estate.

Real estate financial exposure relative to GDP stands at 158.6 percent, one of highest ratios in the world.

Image

https://en.sedaily.com/technology/2026/04/30/koreas-assets-65-percent-in-real-estate-economic-structure "the share of non-financial assets (real estate) held by Korean households reaches 65 percent, compared with 32 percent in the United States and 36 percent in Japan"

But President Lee didn't like that.

So he aggressively started hiking taxes on multiple home owners.

Image

https://www.chosun.com/english/market-money-en/2026/05/09/FUWJFFOIX5BQVLDC3CFNK5MU6E/ For a third property, the effective tax rate can get to as high as 82.5%.

Messaging is clear : don't hoard real-estate.

👉 Domestic Real-estate capital flows into Korean equities markets

Crypto to Stocks

Domestic stocks are still tax-advantaged for most retail investors. Foreign stocks are taxed at 22% on realized gains above the annual deduction.

Crypto has repeatedly been delayed, but it is scheduled to move toward a similar 22% tax regime on Jan 1st 2027.

So the relative incentive is obvious: for Korean retail, domestic equities are becoming the cleanest speculation venue.

👉 Domestic Crypto traders transition to domestic stocks

The Government will cover your stock losses

The most direct nudge is the newly unveiled $100b National Growth Fund. Korea is not just telling people to buy stocks. It is creating products with tax deductions, preferential dividend taxation, and even government loss coverage up to 20%. That is a very loud policy signal.

Image

https://fsc.go.kr/eng/pr010101/85267?utm_source=chatgpt.com

Basically, the government is saying: Stop putting everything into apartments and overseas stocks. Put capital into domestic strategic industries.

And the push is not just retail. The National Pension Service(NPS) is also under pressure to avoid forced selling of Korean equities. Officials are reportedly considering raising the NPS domestic equity target from 14.9% to 19.9%, with the effective upper limit potentially rising to 24.9%.

👉 The government is buying Korean stocks

All in all, it's a pretty good time to be in Korean equities.

  • AI chips supercycle
  • Real-estate capital to Stocks
  • Crypto-capital to Stocks
  • Government subsidizing losses/supporting stock markets directly

That is why I’m not just looking at Korea as a memory trade anymore. I’m looking for the next theme that foreign capital and Korean retail can both understand.

Physical AI fits that perfectly.

Physical AI in Korea

Korea is very much in a memory super-cycle atm.

But with all the policy shifts and the change in perception from foreign investors, it is also has potential to become a physical AI trade as well.

I see this in my Korean semi and robotics watchlist. Down days on semis lead to up days on robotics and vice versa.

When Samsung went on strike, robotics stocks flew.

Foreign capital seem to be catching wind of the trend as well. The three top foreign capital flows in May were into Hyundai Motors, Doosan Robotics and Rainbow Robotics.

Image

https://x.com/Jaymin_Alpha/status/2055892362623414327?s=20

Hyundai owns Boston Dynamics

The most obvious reason to care about Korean Robotics is Hyundai.

In Korea, Hyundai Motor Group is not just a car company. It is one of the country’s dominant chaebols, with Hyundai Motor, Kia, Mobis, Transys, Glovis, and a massive supplier ecosystem underneath it.

Hyundai owns Boston Dynamics, and the strategy is no longer just lab demos. In 2025–2026, Hyundai and Boston Dynamics started talking openly about scaling robot manufacturing, deploying Atlas at Hyundai, and applying robots to real industrial work like automotive manufacturing, logistics, and inspections.

https://www.youtube.com/watch?v=3aQWvdCac9o Building a prototype is one thing, but scaling production is much much harder.

Image

https://x.com/elonmusk/status/1308284091142266881?lang=en Boston Dynamic's Atlas is going into Hyundai factories.

Hyundai plans to deploy Boston Dynamics’ Atlas humanoid robots at its U.S. manufacturing plant in Georgia starting in 2028. Reuters reported that Hyundai aims to manufacture 30,000 robots per year, beginning with tasks like parts sequencing before moving into more complex assembly work by 2030.

Hyundai Mobis is providing actuators for Boston Dynamics.

Boston Dynamics said Hyundai Mobis(012330) will supply actuators for Atlas and help accelerate mass-production plans. Hyundai Motor Group also said Mobis is working with Boston Dynamics to develop high-performance actuators, standardize key components, and optimize designs for manufacturability.

Image

https://bostondynamics.com/news/hyundai-mobis-forms-strategic-collaboration-framework-with-boston-dynamics/ Actuators are pretty important to the humanoid story.

Humanoids and Actuators

Actuators are the motors + sensors + mechanical drive systems inside every joint of a humanoid robot.

A full humanoid can require 40+ actuators, depending on the design. McKinsey estimates that actuators could account for up to 60% of humanoid robot cost, and has flagged them as a potential bottleneck as production scales.

Image
Image

This is a great clip from a Chinese actuator manufacturing company called Zeroerr.

https://x.com/KraneShares/status/2019112595618164959?s=20 ## Breaking down the Actuator

A humanoid actuator is not just a motor.

It is a compact joint system that usually combines a motor, precision reducer / gearbox, bearings, encoder, sensors, brake, housing, and controller into one module. The job is simple to describe but hard to execute: convert electrical energy into smooth, precise, high-torque movement inside a very small space.

The key bottleneck is often the precision reducer / gearbox.

The motor creates rotation, but the reducer converts that high-speed rotation into usable torque for the robot joint. For humanoids, this has to be done with low backlash, high durability, high precision, low noise, and consistent quality across thousands or millions of units.

This is why I care about companies with experience in precision gears, reducers, micro-reducers, actuator modules, and mass-production machining.

Image

The market can easily understand “robots.” The harder question is: who can actually manufacture the joints at scale?

That is where the Korean supply chain starts to get interesting.

Current Korean Humanoid Market

There has already been some repricing in the market as can be seen in the chart below. These are some of the companies I've been tracking.

Image

And as can be seen in the price performance over the past year, there has already been some repricing occurring in the market.

Image

Hyundai Mobis 012330

Mcap $37.4b / Revenue $42.17b / PE 17.05x 1 year performance : 125%

Large-cap robotics supply chain proxy. Not a pure humanoid play, but if Hyundai/Boston Dynamics scales, Mobis is the obvious tier-1 beneficiary. Safer, more liquid, but less torque than the small caps.

ROBOTIS 108490

Mcap $2.9b / Revenue $26.87m / PE - 1 year performance : 429%

Cleanest Korean actuator narrative. Revenue is tiny versus market cap, but 98% actuator exposure and customers like Tesla, Google, and Unitree make it the premium pure-play. Great story, but valuation already prices in a lot.

SPG 058610

Mcap $1.88b / Revenue $235.7m / PE 294x 1 year performance : 362%

Precision control gear motor supplier with a more established revenue base than the tiny reducer names. At ~8x sales, it is not cheap, but it is also not in the same “pure concept stock” zone as SBB/ROBOTIS.

The key debate is whether SPG deserves to be valued as a scarce humanoid actuator supply-chain bottleneck, or just as a high-quality precision motor company with robotics optionality.

SBB 389500

Mcap $373m / Revenue $4.97m / PE - 1 year performance : 241%

High-purity Korean harmonic reducer story. SBB is expensive because the market sees it as one of the few domestic players sitting directly on the humanoid actuator bottleneck: precision reducers, robot drive modules, and now integrated actuators.

The bull case is not current revenue. Current revenue is tiny and the company is still loss-making. The bull case is that SBB becomes Korea’s scarce reducer/module supplier if humanoid supply chains localize around Hyundai, defense, industrial robots, and semiconductor automation.

The risk is that this is still mostly future optionality. Prior capacity expansion ran far ahead of real robot demand, utilization has been low, and at ~75x sales the stock needs a very large order ramp to grow into the valuation.

Keyang Electric Machinery 012200

Mcap 164m / Revenue $268.96m / PE - 1 year performance : 476%

Previously cheap auto-motor company that got repriced into a humanoid joint-module narrative. The Hyundai Transys contract through 2030 gives the story real validation, but the key question is whether the actuator business shows meaningful revenue.

This is more of a “cheap legacy business plus robotics call option” than a clean humanoid pure-play. The market liked the Hyundai Transys link because it gives Keyang a credible path into robot joint drive systems, but the stock still needs evidence that this is not just a small contract attached to a low-margin motor business.

NEOOTO 212560

Mcap $109m / Revenue $160m / PE 10.94x 1 year performance : 129%

Cheap precision gear supplier to Hyundai with real automotive revenue, positive earnings, and a much lower valuation than most Korean humanoid-adjacent names.

More importantly, NEOOTO explicitly highlights in-house development and prototype testing of precision micro-reducers for space robots and mobile robot actuators, plus ultra-precision gear machining down to 5㎛ tolerance. That is much closer to the actual humanoid reducer bottleneck than generic auto-parts exposure.

Less proven as a humanoid supplier than ROBOTIS/SBB, but the risk-reward is much cleaner if the market starts valuing precision reducer capability instead of only confirmed humanoid customers.

Summarizing the rabbit-hole

Physical AI is next, and Korea is suddenly one of the more interesting places to look.

Boston Dynamics is a real contender in humanoids. It is owned by Hyundai, and Hyundai Mobis is the obvious large-cap supply-chain proxy. But the more interesting opportunity may sit further down the stack.

Humanoids are basically walking actuator machines. Actuators are a huge part of the cost, and precision reducers / gearboxes are one of the hardest parts to manufacture at scale.

Korea already has decades of experience mass-producing precision auto components for Hyundai/Kia. Some of those suppliers may now have a path into robot joints, actuator modules, and humanoid components.

So the bet is simple: a boring Hyundai auto-parts vendor could rerate into a physical AI supply-chain stock if the market starts valuing precision gears and reducers as the next bottleneck.

Disclaimer

I own shares of NEOOTO at the time of writing. I may buy or sell without notice. This post is my personal research and opinion, not investment advice, solicitation, or a recommendation to buy or sell any security.