Inside the machines that build every robot

Hi {{first_name|Investor}}

Apologies as today’s issue is a bit later than usual. I just got back from London, UK, and I had a great time. Coincidentally, my trip happened around the same time Leopold’s $45B fund was having a meltdown with Citadel buying its public equities at a massive discount. This is an ardent reminder that one must practice risk management, especially when it comes to leverage. I’ll write more about that in this Thursday’s educational post.

But if you’re a long-term investor like I am, volatility is part of the game when investing in high-growth stocks. Thus, I’m looking towards the future - the humanoid (aka robots) buildout.

The framework issue mapped the humanoid supply chain three levels down and made the case that the asymmetry lives below the component layer. This issue is the descent. The first deep dive, the grinding economy, runs free in full so you can see what these look like. The other three chains live in the full guide, and the download link sits behind the paywall.

This chain runs on precision grinding, the manufacturing step that every humanoid joint on earth passes through, whichever robot maker wins and whichever actuation design they choose.

Robot capex arrives at the grinding machine before robot revenue arrives anywhere else, and the order data says it's arriving now. Robot demand is still a rounding error for every firm named below; the capex pattern is what tells you where the leverage could build.

A planetary roller screw converts a motor's spin into straight-line force through threads that roll against each other under load, millions of cycles, with tolerances measured in microns. A harmonic reducer transmits torque through a thin flexible gear ring whose teeth stay in constant contact with the outer ring. A joint bearing carries all of it on raceways polished to a mirror.

All three share the same manufacturing reality. Once a screw or gear blank has been through heat treatment, the only path to final precision is grinding, on a machine rigid and accurate enough to hold tolerances dozens of times finer than a human hair, thousands of parts in a row. The grinding step is where a screw earns its precision grade, and the precision grade is what separates a $3,000 roller screw from a $30 lead screw.

That makes the grinder the gatekeeper of the entire actuation chain. Chinese industry commentary is unusually direct about this: high-precision thread grinders are consistently flagged as an import dependency, and the Jamestown Foundation estimates China still imports around 90% of its high-end machine tools, mostly from Japan. China has localized the screws and reducers themselves at rates above 90% by domestic accounts. The machines that make them are the layer that resisted. Every high-end joint passes through a grinder, few companies build those grinders, and fewer still make the wheels they consume.

The machine tool cycle is measurable in a way the robot story isn't, and three data sets point the same direction.

The listed field is small, which is the attraction and the constraint.

Qinchuan Machine Tool (SZSE: 000837) is the most complete China-listed expression of this chain. Its machine tool division sells the thread grinders that make roller screws and the gear grinders that make reducers, while its Hanjiang Machine Tool subsidiary uses that same equipment to produce screws itself, with company statements citing P0-grade accuracy (the top precision class) and small-batch roller screw production in customer validation. Its high-end "industrial mother machine" production base (Chinese policy language for the core machine tools that all other manufacturing depends on) was slated for completion at the end of 2025, with capacity releasing through 2026. One company, two levels of the chain.

Hiecise Precision (SZSE: 300809) is the localization spearhead. China's leading roll-grinder maker pivoted into precision thread grinders and booked a 100-machine order in March 2025 for roller screw machining, reported as the first bulk domestic order of its kind. It's the name to watch for evidence that Chinese grinders are reaching import parity, and repeat orders matter more than the first one.

Chongqing Machinery & Electric (SEHK: 2722) owns UK-based PTG Holroyd, a world leader in thread, rotor, and gear grinding machines. A Chinese state-linked conglomerate holding a British crown-jewel grinding franchise is a structural oddity, and export-control attention is a live risk in both directions. The listing is also thin, which limits practical position building.

Tsugami (TSE: 6101) builds precision thread and form rolling machines in Japan, with its China operations separately listed in Hong Kong as Precision Tsugami (SEHK: 1651). Rolling forms threads by pressing rather than grinding them, which matters for the cost-down phase of the screw story: as volumes scale, makers will rough-form by rolling and finish-grind only where precision demands it, so Tsugami sits on the volume side of that split.

Nidec (TSE: 6594 / OTC: $NJDCY) carries a machine tool arm built from the former Mitsubishi Heavy machine tool business and OKK, which sells gear grinders and gear cutting machines, alongside the motor business already supplying humanoid joints and its own reducer ambitions. The humanoid theme touches Nidec at three points, though each is a small slice of a very large company.

Klingelnberg (SWX: KLIN) is the Swiss-listed pure play on gear precision: grinding machines on one side and gear metrology on the other, with the company reporting that the biggest pull for both continues to come from Asia, led by China, India, Japan, and Korea. The gear measurement franchise is the quieter moat, since every reducer maker chasing humanoid tolerances needs certified measurement whether they buy Swiss or Chinese grinders.

The names you can't buy complete the picture: Reishauer and Kapp Niles dominate continuous generating gear grinding, Gleason spans the gear machinery spectrum, Studer leads cylindrical grinding, and Drake and Holroyd's peers own much of the thread-grinding niche. Private ownership across the club is a large part of why the listed few matter so much.

A grinding machine is a one-time sale. What it consumes isn't, and this layer is where I found the single most overlooked economics in the whole humanoid map.

Grinding hardened steel at micron tolerances requires superabrasive wheels, usually CBN, cubic boron nitride, a synthetic crystal nearly as hard as diamond. The wheels wear with every meter of thread and every gear tooth ground, they get replaced on a schedule tied directly to production volume, and switching suppliers requires requalifying the entire grinding process. The machine is the razor, the wheels are the blades, and the blade buyer is locked in tighter than most software customers.

Asahi Diamond Industrial (TSE: 6140) is Japan's listed diamond-tool specialist, and one industry report estimates it supplies over 60% of the grinding wheels used for industrial robot reducer gears, a figure I would treat as directional rather than precise, but the direction is the point. It also markets vitrified CBN wheels for super-finished bearing raceways and the diamond dressers that recondition grinding wheels between cycles, which is a consumable for the consumable. If reducer and screw output scales the way the capacity announcements imply, Asahi's addressable wheel volume scales with it mechanically.

Noritake (TSE: 5331) is the diversified counterpart: Japan's grinding wheel franchise inside a group spanning ceramics and electronic pastes. The group posted record sales of ¥142.9 billion for the year ended March 2026 with net income up close to 10%, but the abrasives segment itself saw sales decline, with strength coming from the electronics side. The robot consumable story is ahead of the abrasives numbers today, and Noritake is a patient way to hold the option while the electronics business pays the bills.

OSG (TSE: 6136) owns the global benchmark franchise in taps and thread-cutting tools, and threads get cut or rolled before they ever get finish-ground, so OSG participates in every thread regardless of finishing method. The company just raised its first-half fiscal 2026 guidance to ¥92 billion in sales, citing stronger-than-expected global demand, and lists robotics among its served markets. Sandvik (STO: SAND) is the same logic at carbide scale, a spread bet on machining activity everywhere with humanoids as one tailwind among many.

Two ways to hold this chain's names: Klingelnberg on machines and Asahi Diamond on consumables sit closest to the theme, while Nidec, Sandvik, and Noritake are the diversified route, broad industrials with the humanoid upside attached.

The earliest public read on whether screw and reducer factories keep equipping is the JMTBA monthly order release, specifically China orders and the precision machinery category. The March and April records are why this chain leads the descent.

Screw maker capacity announcements and, more importantly, utilization commentary. Announced capacity in China already far exceeds near-term robot demand, so the question becomes whose lines actually run.

Hiecise repeat orders and Shuanglin's integration of its acquired grinding-machine maker, as the two cleanest markers of grinder localization. If domestic machines start winning follow-on orders at P0-class precision, the Japanese and European franchise erodes faster than I assume.

Klingelnberg's order intake and any first disclosed humanoid-linked machine orders from the Japanese builders, as confirmation the demand is reaching the high end rather than staying domestic Chinese.

An inflection in the abrasives segments at Noritake and Asahi Diamond, which would be the first sign the consumable annuity is switching on. This signal fires last and matters most, because it's the one that recurs.

The cycle comes first. Machine tools and abrasives are industrial cyclicals, tariff caution is already in builders' commentary, and a manufacturing downturn would swamp the robot signal in every one of these names.

Localization cuts the other way at the machine layer. The whole reason the Japanese and European builders hold pricing power is China's import dependence, and China is attacking exactly that dependence with subsidy support. The consumables and metrology layers are stickier, but the machine franchise is the most exposed part of this chain's thesis.

Process substitution is the quiet risk. If thread rolling, whirling, or forming technologies improve enough to shrink the finish-grinding content per screw, the wheel volume story shrinks with it. Cost-down pressure guarantees screw makers will try.

And the scale point from the introduction cuts hardest here: the humanoid re-rating case needs the theme to keep building for years, and none of these names will move on robot news alone in the meantime.

That's one chain of four. Behind the paywall:

The sensor finishing chain. The two Taiwanese specialists every robot eye is a candidate to pass through, the filter niche worth a patent war, and the coating machines behind every optics line.

Steel, heat, and measurement. The mills, the outsourced hardening network, and the referee layer that certifies every joint before it ships.

The security layer. The first humanoid hack, the passport every robot needs, and the security chip just certified into NVIDIA's robot brain.

A watchlist like the one above comes with each chain, and all of them roll into the 35-name list inside the 26-page guide.