Amram Capital · Research
Robotics & Physical AI · Selection List
The Robotics Playbook — Named Instruments
A specific, global companion to the thematic note. Universe is unconstrained by wrapper or domicile; pick what fits your access and tax position.
The structure is the quality-weighted barbell set out in the note: lead with the compounders and the shovels, hold humanoids as a small option, and buy only the exposure a diversified technology book does not already contain. For single stocks, the third column shows the share of company revenue that comes from robotics specifically, so the distinction between a robot maker and an automation enabler is explicit. This is a selection menu, not a portfolio, and not a recommendation to buy any specific security. Live prices are intentionally omitted; verify them, the exact listing line and spread at execution.
LEAD = the names we would build first
Robotics rev. = share of company sales from robots
Cost = fund TER; single stocks have none
01Quality core
~55% of sleeve
Businesses that already earn and survive a disappointment: the automation oligopoly and the surgical compounders. This is where the real money sits, bought with patience.
Diversified automation funds — pick one
| Instrument | Ticker · listing | Cost | Role & rationale |
|---|---|---|---|
| iShares Automation & RoboticsDiversified ETF · Acc · IELEAD | RBOT / RBTX · LSE | 0.40% | Cheapest and largest robotics ETF, ~150 holdings spanning automation, vision and AI. Broadest low-cost way to own the core. |
| Global X Robotics & AIDiversified ETF · US & UCITS linesLEAD | BOTZ · Nasdaq / LSE | 0.50–0.68% | Concentrated on the hardware moats: Fanuc, Keyence, Intuitive, Nvidia. Buys the Japanese automation cluster that is awkward to hold directly. |
| ROBO Global Robotics & AutomationDiversified ETF · equal-weight | ROBO · NYSE / UCITS | 0.80–0.95% | The original, purest robotics ETF. ~80 names, equal-weighted. Most thematically clean, but the priciest of the three. |
Robotics-revenue single names — the moats
| Instrument | Ticker · listing | Robotics rev. | Role & rationale |
|---|---|---|---|
| FanucIndustrial robots + CNCLEAD | 6954 · Tokyo | ≈41% | Robot segment ~41% of sales in the year to Mar-25, running 43–47% since, and the primary profit driver. The real robotics-revenue single name. |
| Yaskawa ElectricRobots + motion controlLEAD | 6506 · Tokyo | ≈48% | Robotics is the largest segment, ~46–50% of revenue and ~45% of operating profit, though lower-margin than motion control. Genuine robotics exposure. |
| Rockwell AutomationControl & software | ROK · NYSE | <5% | Broad factory automation and control, not a robot maker; AMRs (OTTO) are a small piece. Automation-ecosystem exposure rather than robotics-pure. |
Look-through. Fanuc and Yaskawa are genuine robotics businesses, with roughly 40 to 50% of sales from robots. The diversified funds and the enablers below give you the broader automation cycle that robotics rides, not robot revenue itself. ABB has been removed: robotics is only ~7% of its sales and is being sold to SoftBank, after which the listed company carries essentially none. Keyence has moved to the enabling layer: it is a sensing and vision leader with ~0% direct robotics revenue.
Surgical & specialised robotics
| Instrument | Ticker · listing | Robotics rev. | Role & rationale |
|---|---|---|---|
| Intuitive SurgicalSurgical roboticsLEAD | ISRG · Nasdaq | ~100% | Pure robotic-assisted surgery (da Vinci + Ion). The cleanest robotics revenue on the list. ~60% share, de-rated to ~37x fwd. See dedicated tear sheet. |
| StrykerOrtho robotics + med-tech | SYK · NYSE | single-digit | Mako robotics is a growth driver inside a broad, durable med-tech; robotics is a single-digit share of group revenue. |
| Globus MedicalSpine / ortho robotics | GMED · NYSE | minority | Excelsius navigation and robotics are a minority of musculoskeletal revenue; strategic but not the bulk of the business. |
02Enabling layer
~25% of sleeve
The shovels. Demand here is agnostic to which robot brand wins, which is precisely the point. The actuator and magnet chokepoints have the clearest scarcity cases.
Compute & sensing — the brains and eyes
| Instrument | Ticker · listing | Cost | Role & rationale |
|---|---|---|---|
| NvidiaCompute & robot foundation models | NVDA · Nasdaq | — | The physical-AI compute standard: GR00T models, Isaac and Omniverse. Flag: likely already held via broad tech, so size only the incremental exposure. |
| KeyenceSensors, vision & measurementLEAD | 6861 · Tokyo | — | Premier machine-vision and sensing franchise at ~50% margins. A pure enabler (its vision guides robots) with ~0% direct robotics revenue; one of the great industrial compounders. |
| CognexMachine vision | CGNX · Nasdaq | — | Machine-vision pure-play; cyclical but a direct sensing read. |
Actuators & motion — the chokepoint
| Instrument | Ticker · listing | Cost | Role & rationale |
|---|---|---|---|
| Harmonic Drive SystemsPrecision actuatorsLEAD | 6324 · Tokyo | — | Harmonic-drive gearing: the precision-actuator chokepoint in essentially every humanoid joint. The purest shovel in the theme. |
| NidecMotors & actuators | 6594 · Tokyo | — | Motors and actuators at global scale, with EV and robotics read-through. |
Materials & magnets
| Instrument | Ticker · listing | Cost | Role & rationale |
|---|---|---|---|
| Lynas Rare EarthsRare earthsLEAD | LYC · ASX | — | Largest rare-earth producer outside China and the cleanest ex-China scarcity holding. Preferred over the basket on thesis purity. |
| VanEck Rare Earth & Strategic MetalsMaterials ETF | REMX · NYSE | 0.54% | Basket access to the complex. Caveat: heavy China weight dilutes the Western-supply thesis, which is why it sits behind the single names. |
| MP MaterialsRare earths + magnets | MP · NYSE | — | US rare-earth mining and magnet manufacturing, vertically integrating. The Western magnet-supply chokepoint and the most robotics-relevant of the three. |
03Humanoid optionality
~20% of sleeve
A small, deliberately-sized call on an early inflection. Entered on weakness, not at highs, and sized so it can fail without damaging the book.
Humanoid-pure vehicles
| Instrument | Ticker · listing | Cost | Role & rationale |
|---|---|---|---|
| KraneShares Humanoid & Embodied IntelligenceHumanoid-pure ETF · equal-weightLEAD | KOID / KOIB · Nasdaq & LSE | 0.69% | The cleanest single-line frontier exposure. Diversified, equal-weighted across the humanoid value chain. See dedicated tear sheet. |
| Roundhill Humanoid RoboticsHumanoid-pure ETF · active | HUMN · US | 0.75% | Active and concentrated, with Tesla and Nvidia above 20%. The high-conviction alternative to KOID. |
| ARK Autonomous Tech & RoboticsActive thematic ETF | ARKQ · US | ~0.75% | Broader autonomous, robotics and space; Tesla-heavy. A wider, more diluted frontier wrapper. |
The shortlist we would actually build
Five lines that express the whole barbell without redundancy.
Core breadthRBOT · or BOTZ
One diversified automation fund. RBOT for cheapest breadth; BOTZ for the hardware-moat tilt. Fanuc or Yaskawa if you want a robotics-revenue single name.
Surgical anchorISRG
Start here. The one name already on weakness, with the pullback condition met, and ~100% robotics.
Actuator shovelHarmonic Drive · 6324
The purest enabling-layer holding; demand-agnostic to the winner.
Rare-earth shovelLynas · LYC
The cleanest ex-China rare-earth scarcity holding. MP Materials is the alternative if you want magnet integration.
Humanoid optionKOID / KOIB
The diversified frontier line, sized small and entered on weakness.
Execution discipline
Valuation gates size. Build through tranched limit orders anchored to drawdowns from recent highs, not calendar dates. Do not chase strength; the complex has run hard.
Weakness first. The one place this discipline is already satisfied is the surgical anchor, de-rated on competitive fear rather than deteriorating numbers. That is where to begin.
Count the correlation and the currency. Almost every line here is dollar, yen or yuan denominated, and several re-buy Nvidia and Tesla you may already own. Size the incremental exposure deliberately rather than acquiring it by accident.
Watchlist & on hold
ABB Robotics (SoftBank) — the pure-play robotics arm is leaving ABB for SoftBank; if it is ever re-listed independently it becomes a genuine robotics single name worth revisiting.
Unitree — private; IPO expected 2026. The most-watched pure humanoid name. Wait for the listing.
Figure AI, Apptronik, Agility, Skild AI — private; accessible only via secondaries, with the usual premium and liquidity caveats.
Premium-to-NAV vehicles — any closed-end or trust trading at a large premium to underlying value is disqualified regardless of thesis quality.