Midnight MarketsV4.0
⚠ Not financial adviceNot financial advice. Independent research, not investment recommendations. The author and website make no representation as to accuracy or completeness. Do your own due diligence before acting on anything you read here. Read the full disclaimer →
Sector II · 6 chokepoints · 23 names

Robotics

The Embodied Stack — Robotics

Building a humanoid robot is not like building a phone. It is more like building a small, very angry skeleton that has to balance, see, decide, and grab a coffee cup without crushing it. To do that, you need three things working at once: muscles (the motors), joints (the gear-reducers that turn fast motor spin into slow, powerful arm movement), and nerves and a brain (the sensors and the AI model that decides what to do).

Tesla, Figure, Boston Dynamics, the Chinese humanoid startups, and every warehouse-bot company on Earth are racing to build the skeleton. But the muscles, joints, and nerves are all bought from a tiny handful of companies — and most of those companies are in Japan.

If you own a piece of the joint-makers, you don't care which humanoid wins. You're getting paid by all of them, and the wave is coming whether the headlines that week are bullish or panicked. Same playbook as AI. Same kind of narrow places.

Choke 01

The Joints — Precision Gearing & Reducers

Harmonic Drives · Cycloidal Reducers · Planetary Gearboxes

Three companies in Japan make every robot joint that matters. There are no second sources.

A robot arm needs a gear-reducer at every joint — a small, very precise mechanical device that takes a fast-spinning motor and turns it into the slow, powerful, jitter-free motion needed to put a screw in or pick up an egg without cracking it. Two flavors dominate: "harmonic drives" (compact, zero-backlash — used in robot wrists and elbows) and "cycloidal reducers" (heavier, more torque — used in shoulders and hips). Tesla's Optimus uses both. Figure's robot uses both. Every Boston Dynamics product uses both. Harmonic Drive Systems in Japan invented and still dominates the harmonic flavor; Nabtesco in Japan dominates the cycloidal flavor; Leaderdrive in China is the only credible non-Japanese alternative.

Why this is a chokepoint

If you can't get a harmonic drive, your humanoid does not move. Tesla's Optimus production is rumored to be gated on harmonic drive supply already. Lead times stretched to 12+ months in 2025.

Price

Inventor & ~70% global share of harmonic drives — the joint-of-record for every modern robot wrist

A Japanese mid-cap that invented the harmonic drive in the 1960s and still owns roughly 70% of the world market . Every Tesla Optimus prototype, every Figure 02, every Boston Dynamics Atlas, every Agility Digit uses harmonic drives — and most of them ship from this one factory in Shizuoka. Customer list is essentially a who's-who of humanoid + cobot makers globally.

Price

~60% global share of cycloidal RV reducers — the high-torque joint at the shoulder and hip

Japanese industrial that makes "RV reducers" — cycloidal gearboxes that handle the heaviest joints in industrial robots and humanoid shoulders/hips . ABB, Fanuc, Yaskawa, and Kuka all buy from Nabtesco. Also makes aircraft components and railway brakes — robotics is the growth engine.

Price

China's only credible harmonic-drive challenger — state-backed, fast-improving

The only Chinese harmonic-drive maker with credible scale — closing the precision gap with Harmonic Drive Systems each year. Won meaningful share at UBTech, Xiaomi humanoid, Unitree. Hard to own from US accounts (Shanghai STAR Market). Geopolitical risk in both directions: if China decouples, this stock benefits; if US restricts, sentiment hits.

Choke 02

The Muscles — Servo Motors & Actuators

Servo Motors · Frameless Motors · Linear Actuators

The motors that move the limbs. Less monopolistic than gearing — but the leaders still win.

If the harmonic drive is the joint, the servo motor is the muscle . Every joint needs one or two. Servo motors are precision electric motors that can be controlled to fractions of a degree — modern AI-trained policies depend on having muscles that respond exactly as the model predicted. Three companies dominate: Yaskawa (Japan), Fanuc (Japan), and ABB (Switzerland). They also make their own complete robots, which is why they're both the muscle-supplier and the customer.

Why this is a chokepoint

Less of a chokepoint than gearing — more of an oligopoly. But the leaders have decades of motor-tuning IP that humanoid startups can't replicate quickly. If you're a Figure or a Tesla, you're either buying from these three or building inferior motors yourself.

Price

~20% global servo share + Motoman industrial robots — pure-play robotics

Japanese giant that makes roughly 1 in every 5 industrial servo motors on Earth , plus the Motoman line of welding/assembly robots. Pure-play on factory automation: 90%+ of revenue is robotics-adjacent. Heavy auto-industry exposure means it cycles with EV factory builds.

Price

Highest-margin industrial robot maker — yellow paint, fortress balance sheet

The largest pure industrial-robot company in the world — those signature yellow CNC machines and robot arms on every modern factory floor. Operating margins routinely 30%+ , $7B+ net cash. Sells to automotive, electronics, machine-tool makers.

Price

Swiss industrial — robotics + electrification (also a Choke 03 grid name)

Owns the third-largest industrial robot business in the world plus a top-3 grid electrification business. Spun off the e-mobility/EV charging segment in 2024. The robotics+grid combo gives it tailwinds on two of the biggest secular themes simultaneously. Less explosive upside than pure-plays, but lower drawdown risk.

Choke 03

The Nerves — Sensors & Perception

LiDAR · Machine Vision · Force-Torque Sensors · 3D Imaging

A robot is a blind, deaf brick without sensors. The market is fragmenting in real time.

A robot needs to see (cameras, LiDAR, 3D imagers), feel (force-torque sensors at every joint and on the fingertips), and balance (IMUs — inertial measurement units). Each of these is its own sub-market. Machine vision (Cognex, Keyence) — the cameras that inspect parts on factory lines — is the most mature. LiDAR (Ouster, Luminar, Innoviz) is the most disputed: every humanoid startup is still deciding whether to use LiDAR or camera-only ("Tesla vision"). Force-torque sensors (ATI, Bota) — what lets a robot know how hard it's gripping — are essential and almost entirely private. Event cameras (Sony, Prophesee) are the newest exotic.

Why this is a chokepoint

Sensors are 15-25% of total robot BOM cost and almost all the perception capability. The companies that get qualified into the first 5 humanoid platforms shipping at volume will get qualified into every subsequent program — like getting designed into the iPhone.

Price

Pure-play machine vision — the eyes on every factory line on Earth

A US small-cap that makes industrial machine-vision systems and barcode readers — the cameras and software that inspect parts on assembly lines, route packages in fulfillment centers, and guide robots to pick objects. Customers: every major automaker, every Amazon warehouse, every electronics fab. ~50% gross margins. Heavy exposure to consumer electronics and EV capex.

Price

Digital flash LiDAR — the LiDAR consolidator after the 2023-24 SPAC wreck

Survivor of the LiDAR SPAC carnage. Merged with Velodyne in 2023, became the largest US-listed LiDAR company by revenue. Sells to industrial automation (robotaxis like Motional, smart-city infrastructure), trucking (Embark legacy), warehouse robots, security/perimeter. Less auto-OEM dependent than competitors.

Price

Long-range LiDAR for ADAS — Volvo, Polestar, Mercedes design wins

Once a $30+ darling, now a sub-$5 turnaround. Founder Austin Russell stepped back in late 2025 amid governance issues. Path to profitability is brutal — they need every design win to ramp to volume in 2026-27 to survive the cash burn. Binary outcome stock : either a multi-bagger if Volvo/Polestar volumes hit, or a zero. Position-sizing matters.

Price

The Japanese Cognex — direct sales, 55% operating margins, no debt

Famously eccentric Japanese sensor and machine-vision giant — over 55% operating margins , no debt, no acquisitions. Direct salesforce reaches 350,000+ customers worldwide. Less LiDAR exposure than Cognex; more general industrial. Trades at high multiples reflecting that quality. Compounder, not a moonshot.

Choke 04

The Brain — Humanoid Platforms & AI Models

Foundation Models for Robotics · Vertically Integrated Humanoids

The skeleton-makers themselves. Only one is publicly investable at scale.

The companies actually building the humanoid robots — putting joints, muscles, sensors, and AI brains together — are mostly private (Figure AI, Boston Dynamics, 1X, Agility, Apptronik, Sanctuary, Unitree, UBTech). Tesla is the one mega-cap public exposure. The AI inside the robot is the new race — Physical Intelligence (private), Skild AI (private), NVIDIA's Isaac/GR00T (a free option inside the AI thesis), and a wave of "robotics foundation model" startups. We list the few you can actually trade.

Why this is a chokepoint

This is the most narrative-driven of the four robotics chokes. The flag-planting matters: whoever gets to a million units shipped first reshapes the entire supply chain. Tesla is currently leading by a wide margin on stated ambition.

Price

Optimus humanoid + FSD/robotaxi — the only mega-cap humanoid pure-ish play

You know what they do. The thesis here is Optimus, not cars . Musk has repeatedly said Optimus could be a $10T+ business — even if it's 1/10th of that, it dwarfs the auto business. First commercial-ish units targeting late 2026/2027. The bet is that Tesla's vertical integration (motors, batteries, AI inference) gives it a cost advantage at scale that Figure/Boston Dynamics can't match.

Price

$39B private valuation as of Feb 2025 — IPO chatter for 2026/2027

Not yet investable directly — but on every "secondary market" watchlist. If they file an S-1 in late 2026 or 2027, this becomes the most-anticipated robotics IPO ever. For now, the play is owning Figure indirectly through Microsoft (data center anchor), NVIDIA (compute supplier), and the gear-reducer / motor / sensor names listed above.

Choke 05

The Workhorse — Industrial & Warehouse Robotics

Warehouse Automation · Surgical Robotics · Test & Cobots

The robotics revenue that's actually happening today, while humanoids are still in beta.

While the humanoid story dominates X/Twitter, the boring industrial robotics revenue is shipping at billions of dollars per year, right now . Three subdomains have real cashflow and real moats: warehouse automation (Symbotic powering Walmart, AutoStore in Europe), surgical robotics (Intuitive Surgical with the da Vinci platform — 99% gross margins on a $2M machine), and semiconductor test & cobots (Teradyne + its Universal Robots cobot arm subsidiary). These are not "futures." They're "compounders that also have humanoid optionality."

Why this is a chokepoint

If you don't believe humanoids ship at scale in this decade, these are the names you still want to own — they're getting the robotics multiple even if Optimus disappoints. And they're already profitable.

Price

End-to-end warehouse automation — Walmart's exclusive backbone

Builds the entire robotic guts of a modern warehouse — autonomous bots that move pallets and cases at high speed, racking, software. Walmart has committed to deploying Symbotic in all 42 of its US regional distribution centers — a $11B+ contract. Also has GreenBox, a joint venture with SoftBank to operate as a service. $22.4B backlog as of Q2 2026.

Price

da Vinci surgical robotics — the platform monopoly with razor-blade economics

The da Vinci robotic surgery platform — sold to hospitals at ~$2M per system, then bills $1,500+ of recurring "instruments & accessories" revenue every single time a surgeon uses it. ~9,500 systems installed globally; ~3M procedures per year and growing 15-20%. Razor-and-blades economics with no real competitor . The 2024 da Vinci 5 launch refreshes the platform.

Price

North American factory automation incumbent — reshoring pure-play

Sells programmable logic controllers (PLCs), drives, motion control, and automation software to North American manufacturers. CHIPS Act fab construction (Intel Ohio, TSMC Arizona, Samsung Texas), Inflation Reduction Act EV factories, and pharma capacity expansion all flow through Rockwell. Lower beta than humanoid plays but real, durable demand. Trades at fair multiple given the cycle stage.

Choke 06

The Defense Robots — Drones, Loitering Munitions & Defense AI

Loitering Munitions · Counter-UAS · Defense AI · Attritable Drones

Ukraine proved $500 drones beat $10M tanks. The Pentagon is shifting from $50M platforms to $50K drones — and the supply chain to deliver tens of thousands of attritable robots per year doesn't yet exist.

Modern warfare has been turned upside down in 36 months. Cheap, expendable drones overwhelm air defenses. Loitering munitions (think: a flying grenade that orbits a target for hours before deciding when to dive) are now the dominant infantry weapon. The US Department of Defense launched the Replicator initiative to deliver thousands of "attritable autonomous systems" — robots cheap enough to lose, smart enough to win — and the manufacturing base to support that doesn't exist yet. The companies building it are mostly small, fast-growing, and trading like growth stocks rather than defense primes.

Why this is a chokepoint

The DoD wants to move from a few hundred exquisite, expensive systems per year to tens of thousands of cheap autonomous ones. Only a handful of companies — AeroVironment, Anduril (private), Shield AI (private), Palantir, Kratos — are scaled enough to bid on these contracts today. Every Replicator dollar, every NATO reorder, and every counter-drone budget flows through this short list.

Price

Maker of Switchblade — the de facto US loitering munition. Just merged with BlueHalo to add directed-energy and space.

The US specialist for small unmanned aircraft and loitering munitions . The Switchblade 300, 400, and 600 are small "kamikaze" drones a soldier can carry in a backpack, launch from a tube, and steer to a target dozens of miles away. They've been used extensively in Ukraine, are deployed across 100+ allied militaries, and have a 42,000+ installed base. In May 2025, AeroVironment closed a $4.1B all-stock merger with BlueHalo, adding directed-energy weapons, satellite operations, and space-domain awareness — making AVAV one of the few US defense names with a fully integrated drone-plus-counter-drone product line.

Price

The defense-AI software chokepoint — Maven Smart System is the de facto DoD AI platform. Also fastest-growing commercial AI software company.

Palantir builds the software platform militaries and big companies use to turn messy real-world data into decisions. Gotham is the defense product (used by US Army, Navy, Space Force, Israel, Ukraine, UK MoD). Foundry is the commercial equivalent. AIP is the new AI agent layer that sits on top of both. The Maven Smart System — Palantir's contract from the DoD's Project Maven — has effectively become the operating system for joint-force AI: sensor fusion, target identification, command-and-control for the entire DoD. It is the rare US AI software product that is genuinely fielded at scale, not a pilot.

Price

XQ-58 Valkyrie loyal-wingman drone + jet engines + hypersonics — the affordable autonomous-aircraft pioneer.

Kratos makes large autonomous combat aircraft, jet engines for tactical drones, and hypersonic targets . The flagship product is the XQ-58 Valkyrie — a runway-independent, low-cost (~$3M) unmanned fighter that flies alongside crewed aircraft as a "loyal wingman." Selected for the Air Force's Collaborative Combat Aircraft (CCA) program. Kratos also makes most of the hypersonic test targets the US uses to qualify missile defenses, and is the dominant US supplier of small turbojet engines for cruise missiles and target drones.

Price

Drone-in-a-box autonomous inspection + counter-UAS systems — small-cap with explosive growth.

Two-business holding company: Ondas Autonomous Systems (Airobotics + American Robotics — automated drone-in-a-box for critical infrastructure and counter-drone) and Ondas Networks (private wireless for rail and energy). The hot business is counter-UAS and autonomous drone deployment — the same Replicator / counter-drone tailwind that lifts AeroVironment.

Price

Defense electronics subsystems — RF, EW, processing modules inside missiles, radars, and drones.

Makes the rugged electronics modules that go inside missiles, radars, electronic-warfare pods, and high-end UAVs. Mercury sits one layer below the primes (Raytheon, Lockheed, L3Harris) but above the chip vendors. Q3 FY26 EPS $0.27 vs $0.07 estimate; record bookings. The turnaround thesis is intact. Smaller, less liquid than the primes but exposed to every modernization program — including the very loitering munitions and counter-UAS systems in this section.

Price

Defense AI consulting + integration — the "AIP for the rest of DoD" partner.

The largest defense AI consulting and systems-integration firm . $38B backlog. Booz is the rare consulting firm that gets paid to deploy Palantir AIP, Anduril Lattice, and other defense-AI platforms into the field. Quietly positioned as the "neutral party" that connects all the new defense-AI vendors to DoD customers. Anduril partnership announced 2025 makes it a key integrator. Lower beta than the pure-play drone names.

Price

Small-drone program of record — Army Short-Range Reconnaissance winner.

Very small cap. The thesis is binary: can Red Cat manufacture at the volumes the Army actually wants? Skydio (private) is the major competitor and arguably has better commercial drone tech. Watch SRR delivery rates and any follow-on international interest in the Teal 2 platform. Size very small.

Price

US drone motors, parts, and FPV components — Berry-compliant supply chain for the Replicator era.

Q1 2026 revenue +296% YoY; raised drone-revenue guide to $10-14M. Needham PT $22. Sub-$200M market cap means single contracts move the stock 30%+. Pure microcap risk profile — but a unique exposure to the "no-Chinese-parts" Pentagon procurement mandate. Note: also covered as private/quasi-private peer to Anduril (private) and Shield AI (private) , which are not investable directly.