# FANUC and the Robotics Basket — InvestMoat Research

_The robotics tape prices FANUC, Yaskawa, Keyence, ABB, and Nabtesco with Tesla Optimus. FANUC's order surge is CNC, Yaskawa's is servo, ABB is selling the OEM, Nabtesco is the industrial joint. Optimus is still Construction._

_Published: September 19, 2026 · Last reviewed: September 19, 2026 · Canonical page: https://investmoat.com/research/fanuc-and-the-robotics-basket_
_Tags: Industrials, Robotics, Business Models, AI_
_Covers: FANUY, 6506.T, 6861.T, ISRG, ROK, ABB, TER, 6268.T, HSYDF, TSLA_

## Summary

The robotics tape treats a FANUC yellow arm, a Motoman cell, a Keyence sensor, a da Vinci console, a Logix PLC, an ABB drive, a Teradyne cobot, an RV reducer, a strain-wave gear, and Tesla Optimus as one AI-hardware sector. They are not. FANUC's Q1 FY2026 order book jumped 36.9% on FA — CNC, including machined components for someone else's humanoid — while Robot orders rose 8.1%. Yaskawa's same-season print jumped on Motion Control; Robotics sales were +2.0% and Robotics operating profit collapsed on ERP and Europe. ABB's Q2 is electrification and a SoftBank sale of the robot OEM. Nabtesco's H1 Component Solutions bounce is the industrial-arm joint, not Harmonic Drive's strain-wave option. The variable that sorts the cohort is whether a unit is already qualified into a cell, an operating room, a plant's control code, or a grid lineup. A humanoid TAM is a different instrument.

## What would prove this wrong

FANUC reporting Robot-division orders down year over year in both H1 and H2 of FY2027 (year ending March 2028) while Tesla discloses external commercial Optimus revenue — customer shipments, not internal training units — in a FY2027 quarter. That would show the humanoid skipped the certified cell. A second trip: Harmonic Drive disclosing AI-robot orders as more than half of consolidated orders for two consecutive halves while industrial-robot demand is disclosed as flat or down.

_Status as of September 19, 2026: **holding**._

---

**Ten names the robotics tape treats as one sector, grouped by whether a unit is already qualified into a workflow.**

| Ticker | Name | Moat | Growth | Valuation | Composite | Note |
| --- | --- | --- | --- | --- | --- | --- |
| **Already in the cell, the OR, or the grid** |  |  |  |  |  |  |
| [FANUY](https://investmoat.com/stocks/fanuc) | FANUC Corporation | 77 | 75 | 80 | 79 | Q1 FY2026 orders ¥281.9B, +36.9%; FA +63.8%, Robot +8.1% |
| [6506.T](https://investmoat.com/stocks/yaskawa) | Yaskawa Electric | 62 | 64 | 76 | 67 | Q1 orders +29%; AC servo +65%; Robotics sales +2.0%; Robotics OP −82.3% |
| [6861.T](https://investmoat.com/stocks/keyence) | Keyence Corporation | 76 | 80 | 78 | 80 | Q1 sales ¥346.6B +32.8%; OP margin 54.0%; sensors already in the cell |
| [ISRG](https://investmoat.com/stocks/isrg) | Intuitive Surgical, Inc. | 80 | 78 | 72 | 78 | Q2 procedures +16%; instruments and accessories $1.73B of $2.89B; 11,710 da Vinci systems |
| [ROK](https://investmoat.com/stocks/rok) | Rockwell Automation | 67 | 72 | 63 | 66 | Q2 organic sales +9%; Intelligent Devices $1.0B, +13%; ARR +6% |
| [ABB](https://investmoat.com/stocks/abb) | ABB Ltd | 72 | 76 | 74 | 75 | Q2 continuing ops orders $12.0B +28% comparable; Robotics discontinued $576M; SoftBank close expected H2 2026 |
| **Robotics is the sidecar** |  |  |  |  |  |  |
| [TER](https://investmoat.com/stocks/ter) | Teradyne | 54 | 78 | 52 | 58 | Q2 Robotics $100M of $1.33B; Semiconductor Test $1.12B |
| **Industrial joint vs humanoid joint** |  |  |  |  |  |  |
| [6268.T](https://investmoat.com/stocks/nabtesco) | Nabtesco Corporation | 64 | 73 | 74 | 71 | H1 Component Solutions ¥45.5B +23.6% (27% of group); Q2 PRG orders +23% YoY |
| [HSYDF](https://investmoat.com/stocks/harmonic) | Harmonic Drive Systems | 64 | 77 | 48 | 59 | FY March 2026 sales ¥59.6B, +7.0%; reducers ¥46.3B; AI-robot market still called nascent |
| **The skip** |  |  |  |  |  |  |
| [TSLA](https://investmoat.com/stocks/tesla) | Tesla Inc. | 70 | 64 | 63 | 64 | Q2 Optimus capacity: Construction at Fremont and Texas; operating margin 1.4% |

_Valuation and composite above use each asset's static valuation score; the live site recomputes them against the current market price._

That is the screen. Robotics is not a GICS code in this coverage universe. It is a tape. When a humanoid demo runs, [FANUY](/stocks/fanuc), [6506.T](/stocks/yaskawa), [6861.T](/stocks/keyence), [ISRG](/stocks/isrg), [ROK](/stocks/rok), [ABB](/stocks/abb), [TER](/stocks/ter), [6268.T](/stocks/nabtesco), [HSYDF](/stocks/harmonic) and [TSLA](/stocks/tesla) get pulled as if they were one AI-hardware sector. They share a word. They do not share a cash-flowing unit.

**A factory that has already qualified a FANUC CNC by serial number is not shopping a humanoid next Tuesday.** A Motoman welding cell is the same species of lock without the G-code dialect. A Keyence sensor already written into an inspection recipe is not a biped. A hospital that has already credentialed surgeons on da Vinci is not swapping the console for a general-purpose worker. A North American plant whose lines run on Logix is not ripping out Studio 5000 because a robot walked. An ABB medium-voltage lineup specified into a data center is not a Motoman-class arm — and ABB is selling that arm. Treating those as one trade with Tesla Optimus — still Construction on the Q2 manufacturing table — is the robotics version of sorting a process step with a science project.

## FANUC's order surge is CNC. Robot orders rose 8%.

FANUC reported the first quarter of FY2026, the three months ended June 30, on July 31. Net sales were ¥231.0 billion, up 17.7% year over year. Operating income was ¥53.5 billion, up 26.1%. The company raised the full-year sales guide to ¥948.1 billion and net income to ¥198.0 billion. Orders were ¥281.9 billion, up 36.9%. That is the number the robotics tape can index on. It is not a robot number.

**FANUC quarterly orders by division — the surge is FA (CNC and servo), not Robot.**

| Period | FA orders | Robot orders |
| --- | --- | --- |
| Q1 FY25 | ¥55B | ¥90.6B |
| Q2 FY25 | ¥48.8B | ¥91.8B |
| Q3 FY25 | ¥54.7B | ¥98.8B |
| Q4 FY25 | ¥71.1B | ¥100.6B |
| Q1 FY26 | ¥90.2B | ¥97.9B |

FA is CNC, servo, and related. Robot is the industrial-arm line. Q1 FY2026 FA orders +63.8% YoY; Robot orders +8.1% YoY and −2.6% QoQ. Robot mix of orders fell to 34.7% from 44.0% a year earlier.

_Figures as of July 31, 2026 (Q1 FY2026 results deck). Source: [FANUC Financial Results for the Three Months Ended June 30, 2026](https://www.fanuc.co.jp/en/ir/announce/pdf/2026/reference202606_e.pdf)._

Robot sales in the quarter were still the largest slice of the mix at 41.6%, up 18.7% year over year, with the Americas and China the disclosed growth regions. That is a recovery in the installed-arm business off the destocking trough. It is not the order spike. FA orders rose 63.8% year over year to ¥90.2 billion and took 32.0% of the order book, up from 26.7% a year earlier. On the July 31 call, FANUC said the FA surge was semiconductors for data centers and AI servers, 3C, NEV — and machined components for humanoid robots. **The humanoid shows up in FANUC's print as someone else's parts list, cut on a FANUC controller.**

## Yaskawa's surge is servo. Robotics sales rose 2%.

[6506.T](/stocks/yaskawa) is the name that tests whether FANUC's print is a robot cycle. Motoman is one of the four global industrial-arm standards. It does not own FANUC's CNC serial-number ecosystem. If the tape were right that Q1 was a bid on humanoid arms, Yaskawa's Robotics line should have done the jumping. It did not.

**Yaskawa Q1 FY2026 (three months ended May 31) — Motion Control did the growing. Robotics did not.**

| Line | Q1 FY2026 | What it tests |
| --- | --- | --- |
| Motion Control | Revenue ¥67.6B (+21.5%); OP ¥7.6B (+50.1%). Q&A: quarterly AC-servo orders at a record, +65% YoY; semiconductor-related AC-servo orders +200%+ | The order spike is servo into semiconductor and data-center equipment — FANUC's CNC analogue, not a humanoid arm |
| Robotics | Revenue ¥56.7B (+2.0%); OP ¥0.9B (−82.3%). Q&A: Robotics orders +14% to a quarterly record; semiconductor robotics +141%; ~¥1.0B Europe restructuring plus ERP production hits | Robot orders recovered. Robot sales and profit did not. ERP and Europe are the disclosed reasons, not Optimus |
| Group | Revenue ¥139.0B (+10.6%); OP ¥8.5B (−19.2%); orders +29% YoY / +8% QoQ. FY guide unchanged at ¥580B sales / ¥60B OP while ERP is assessed | Tape can index the +29% as robots. The mix is servo plus an ERP-constrained arm line |

_Figures as of July 10, 2026 (Q1 FY2026 results, briefing deck, and Q&A). Source: [Yaskawa Consolidated Results for the First Quarter of Fiscal Year Ending February 28, 2027](https://www.yaskawa-global.com/wp-content/uploads/2026/07/20260710_en.pdf); [Yaskawa FY2026 1Q Financial Results Briefing Materials](https://www.yaskawa-global.com/wp-content/uploads/2026/07/20260710_haifu_en.pdf); [Q&A for FY2026 1Q Results Briefing (Summary)](https://www.yaskawa-global.com/wp-content/uploads/2026/07/261Q_QA_EN.pdf)._

Yaskawa's July 10 Q&A is blunt about the split. Quarterly AC-servo and Robotics orders both hit records. Revenue missed the internal plan because production lines shut for the new ERP. Robotics operating profit took the Europe restructuring and an ERP hit on top. Full-year guidance stayed at ¥580 billion of sales and ¥60 billion of operating profit because management is still watching the system stabilize. **A Motoman order book that cannot be built is not a humanoid TAM, and a Σ-servo book into wafer tools is not one either.**

## ABB is selling the robot OEM. The Q2 cash is electrification.

[ABB](/stocks/abb) is the name the robotics ETF can still hold as if the arm were the company. On October 8, 2025, ABB agreed to sell the Robotics division to SoftBank for an enterprise value of $5.375 billion. Q2 2026 already reports that division as discontinued operations. The continuing-operations print is a grid-and-drives compounder funding a Rotork bid with the cash from the sale.

**ABB Q2 2026 — continuing operations against the Robotics line held for sale.**

| Line | Q2 2026 | What it tests |
| --- | --- | --- |
| Continuing operations | Orders $12.042B (+28% comparable); revenue $9.475B (+12% comparable); Operational EBITA $1.925B, 20.2% (+90 bps). Electrification orders +58% comparable; Motion +20%. Order backlog $30.0B. Q2 stranded costs $25M still in continuing ops | The cash is electrification, motion, and process automation. A data-center UPS is not a humanoid |
| Robotics (discontinued) | Q2 revenue $576M (vs $605M); income from operations $47M (vs $102M). SoftBank EV $5.375B; Q2 deck: ~$4.8B net cash proceeds expected, close H2 2026, to fund a ~$5.5B Rotork offer | The robotics tape that bundles ABB with FANUC is bundling a business ABB is selling |

_Figures as of July 16, 2026 (Q2 results, group presentation, and Financial Information); sale announced October 8, 2025. Source: [ABB Q2 2026 results](https://new.abb.com/news/detail/137496/q2-2026-results); [ABB Q2 2026 Group results presentation](https://library.e.abb.com/public/47f0bfb2ecaf494b99846882bffa5409/ABB-Q2-2026-Group-results.pdf); [ABB Q2 2026 Financial Information — discontinued operations (Robotics)](https://www.schweiz.biz/wp-content/uploads/ABB-Q2-2026-Financial_Information.pdf); [ABB to divest Robotics division to SoftBank Group](https://new.abb.com/news/detail/129685/abb-to-divest-robotics-division-to-softbank-group)._

> **Do not index a CNC order, a servo order, or a robot-OEM sale as a humanoid TAM** — A yellow arm in a cage and a humanoid that is supposed to skip the cage are different products. A CNC that machines a flexspline for a humanoid joint is a third thing: FANUC selling the process step that makes the option, not the option. A Motoman order that cannot ship because of ERP is a fourth. A SoftBank cheque for ABB Robotics is a fifth: someone is paying for the OEM, and it is not the listed ABB equity after the close. If the tape cannot tell those apart, it will pay a humanoid multiple for a machine-tool recovery, a servo cycle, or an electrification compounder.

## Ten prints, ten jobs

**What the latest print actually is, against the robotics label.**

| Name | Latest print | What the robotics label is pretending |
| --- | --- | --- |
| FANUY | Q1 FY2026: sales ¥231.0B (+17.7%); orders ¥281.9B (+36.9%); FA orders +63.8% to ¥90.2B; Robot orders +8.1%; FY sales guide raised to ¥948.1B. Q&A: FA demand includes machined components for humanoid robots | That the order book is a bid on humanoid arms. It is a bid on CNC, including CNC that cuts humanoid parts |
| 6506.T | Q1 FY2026: sales ¥139.0B (+10.6%); orders +29%; Motion Control ¥67.6B (+21.5%); Robotics ¥56.7B (+2.0%); Robotics OP −82.3% to ¥0.9B. Q&A: AC-servo orders +65%; Robotics orders +14% | That Motoman is the robot cycle FANUC is hiding. The cycle in the print is servo. The arm line is ERP-constrained |
| 6861.T | Q1 FY2026 (Mar 21–Jun 20): sales ¥346.6B (+32.8%); OP ¥187.1B (+44.7%); OP margin 54.0%; gross 85.0%; overseas +24.0% local currency; Asia +33.9% LC; Japan +19.9%. FX ~+¥26.1B sales / +¥18.9B OP | That a factory-automation sensor compounder is a humanoid inspection TAM. The cash is a sensor already qualified into the cell |
| ISRG | Q2: revenue $2.89B (+19%); worldwide procedures +16% (da Vinci +15%, Ion +36%); instruments and accessories $1.73B; 468 da Vinci systems placed (246 da Vinci 5); installed base 11,710 (+12%); FY da Vinci procedure guide 13.5–15.5% | That surgical robotics is the same TAM as factory humanoids. The cash is a procedure on a credentialed console |
| ROK | Q2 FY2026: reported sales +12%, organic +9%; Intelligent Devices $1.0B (+13% reported, +9% organic); total ARR +6%; FY reported sales midpoint ~$8.9B; adjusted EPS guide $12.50–$13.10 | That Rockwell is a robot OEM. Logix is the plant OS a robot has to talk to |
| ABB | Q2 continuing ops: orders $12.042B (+28% comparable); revenue $9.475B (+12% comparable); Operational EBITA 20.2%. Robotics discontinued: revenue $576M, OP $47M. SoftBank EV $5.375B, close expected H2 2026 | That ABB is a robot OEM in the basket. The listed equity is exiting the OEM and buying actuators for process plants |
| TER | Q2: revenue $1.329B; Semiconductor Test $1.122B; Product Test $107M; Robotics $100M (Universal Robots + MiR) | That Teradyne is a cobot compounder. Robotics is 8% of the quarter; the cash is ATE |
| 6268.T | H1 FY2026: sales ¥167.4B (+16.8%); OP ¥15.8B (+72.4%); OPM 9.5%. Component Solutions sales ¥45.5B (+23.6%), orders ¥48.0B (+26.0%); Q2 PRG orders +23% YoY / +11% QoQ. FY guide raised to ¥344B sales / ¥32.6B OP. Script: RV mini deliveries started H2 2026; customers include FANUC, Yaskawa, KHI, KUKA, ABB Robotics | That the industrial RV is a humanoid actuator. It is the joint already designed into the caged arm, 27% of group sales |
| HSYDF | FY March 2026: sales ¥59.6B (+7.0%); operating profit ¥2.57B; orders ¥61.6B (+16.2%); reducer sales ¥46.3B (77.8% of sales). Filing: won AI-robot orders as customers move toward mass production; still calls the AI-robot market nascent. FY March 2027 guide: sales ¥68.0B, operating profit ¥6.2B | That the strain-wave gear is already a humanoid volume story. The cash is still industrial robots and semiconductor equipment |
| TSLA | Q2: revenue $28.2B (+26%); operating income $0.4B, operating margin 1.4%; automotive $20.5B. Manufacturing table: Optimus at Fremont and Texas listed as Construction, after S/X lines were decommissioned, with production anticipated later this year | That the multiple is a robot company. The robot line is a construction status. The cash is cars, energy, and services |

_Figures as of May–July 2026 (latest reported period per name); ABB Robotics sale October 8, 2025. Source: [FANUC Financial Results for the Three Months Ended June 30, 2026](https://www.fanuc.co.jp/en/ir/announce/pdf/2026/reference202606_e.pdf); [Q&A Summary of the Telephone Conference for FY2026 1Q](https://www.fanuc.co.jp/en/ir/announce/pdf/2026/qasummary202606_e.pdf); [Yaskawa Consolidated Results for the First Quarter of Fiscal Year Ending February 28, 2027](https://www.yaskawa-global.com/wp-content/uploads/2026/07/20260710_en.pdf); [Q&A for FY2026 1Q Results Briefing (Summary)](https://www.yaskawa-global.com/wp-content/uploads/2026/07/261Q_QA_EN.pdf); [KEYENCE First Quarter of FY2026 Financial Results Materials](https://www.keyence.co.jp/pdf/FinancialResults_202607_en.pdf); [ABB Q2 2026 results](https://new.abb.com/news/detail/137496/q2-2026-results); [ABB Q2 2026 Financial Information — discontinued operations (Robotics)](https://www.schweiz.biz/wp-content/uploads/ABB-Q2-2026-Financial_Information.pdf); [ABB to divest Robotics division to SoftBank Group](https://new.abb.com/news/detail/129685/abb-to-divest-robotics-division-to-softbank-group); [Intuitive Announces Second Quarter Earnings (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/1035267/000103526726000047/q226ex-991earningsrelease.htm); [Rockwell Automation Reports Second Quarter 2026 Results (Exhibit 99)](https://www.sec.gov/Archives/edgar/data/1024478/000102447826000020/q2fy26ex99.htm); [Teradyne Reports Second Quarter 2026 Results (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/97210/000119312526321933/ter-ex99_1.htm); [Nabtesco Summary of Consolidated Financial Statements for FY2026 Q2](https://www.nabtesco.com/cms/wp-content/uploads/Summary_of_Consolidated_Financial_Statements_for_FY2026_Q2_e.pdf); [Nabtesco Results Briefing Material for FY2026 Q2 (script)](https://www.nabtesco.com/cms/wp-content/uploads/Results_Briefing_Material_for_FY2026_Q2e_script.pdf); [Harmonic Drive Systems Consolidated Financial Results for FY March 2026 (Japanese GAAP)](https://www.nikkei.com/markets/ir/irftp/data/tdnr/tdnetg3/20260513/fx6gjx/140120260511521357.pdf); [Tesla Q2 2026 Update (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/1318605/000162828026049213/exhibit991.htm)._

Intuitive is the cleanest control. Q2 instruments and accessories were $1.73 billion against $685 million of systems. The installed base is 11,710 da Vinci systems. Full-year da Vinci procedure growth is guided to 13.5–15.5%, closer to the midpoint, after a year that ran hotter. That is a razor-and-blades print on a trained surgeon. It is robotics. It is not a humanoid TAM, and it does not re-rate when Optimus changes construction status.

Rockwell's Q2 is a controls print. Intelligent Devices — drives, motion, sensing — did $1.0 billion, up 13%. ARR was up 6%. Reported sales growth of 12% is not a robot OEM inflecting. It is the North American plant OS cycling with discrete automation, data centers, and reshoring. A humanoid that walks a factory floor still has to speak to that layer, or replace it. Replacing it is a multi-year revalidation. The tape that bundles ROK with TSLA is bundling the layer with the option.

[6861.T](/stocks/keyence) is already in the cell the humanoid is supposed to skip. Q1 FY2026 sales were ¥346.6 billion, up 32.8%. Operating margin was 54.0%. Asia was +33.9% in local currency. That is a factory-automation and semiconductor-inspection recovery on a direct-sales sensor franchise, with yen weakness adding about ¥26.1 billion to sales. A Keyence vision system specified into a pharma or battery line is a qualification event. It is not a biped, and it is not a cobot sidecar. The tape that pulls Keyence on a humanoid headline is pulling the inspection layer because the word factory is nearby.

Teradyne is the name the market can buy when it wants "robots" in a liquid US listing. Q2 Robotics was $100 million. Semiconductor Test was $1.122 billion. Robotics is a real cobot and AMR franchise — Universal Robots defined the category — and it is 8% of the quarter. Buying TER for physical AI is buying an ATE duopoly with a cobot sidecar. The sidecar can grow. It cannot be the print that justified the re-rating until it is no longer a sidecar.

## Nabtesco is the industrial arm. Harmonic Drive is the option.

**Two reducers, two jobs — cycloidal RV in the caged arm against strain-wave in the humanoid option.**

| Name | Latest print | What the robotics label is pretending |
| --- | --- | --- |
| 6268.T | H1 Component Solutions ¥45.5B of ¥167.4B group (+23.6% sales, +26.0% orders). Q2 PRG orders +23% YoY / +11% QoQ on China and Korea auto-robot capex, machine tools, and semiconductor equipment. Script: RV mini deliveries started H2 2026 to chase smaller robots and Physical AI. Customers named: FANUC, Yaskawa, KHI, KUKA, ABB Robotics | That an RV recovery is a humanoid actuator ramp. It is restocking of the industrial joint, 27% of a doors-and-marine company |
| HSYDF | FY March 2026 reducer sales ¥46.3B of ¥59.6B. Filing records AI-robot orders as customers move toward mass production and still calls the market 創成期 — nascent. FY March 2027 guide: sales ¥68.0B, operating profit ¥6.2B | That strain-wave is already the mix. The cash is still industrial robots and semiconductor equipment; the option is in the multiple |

_Figures as of Nabtesco H1 FY2026 reported July 31, 2026; Harmonic Drive FY March 2026 reported May 13, 2026. Source: [Nabtesco Summary of Consolidated Financial Statements for FY2026 Q2](https://www.nabtesco.com/cms/wp-content/uploads/Summary_of_Consolidated_Financial_Statements_for_FY2026_Q2_e.pdf); [Nabtesco Results Briefing Material for FY2026 Q2 (script)](https://www.nabtesco.com/cms/wp-content/uploads/Results_Briefing_Material_for_FY2026_Q2e_script.pdf); [Harmonic Drive Systems Consolidated Financial Results for FY March 2026 (Japanese GAAP)](https://www.nikkei.com/markets/ir/irftp/data/tdnr/tdnetg3/20260513/fx6gjx/140120260511521357.pdf)._

A six-axis industrial arm wants a handful of high-stiffness cycloidal RVs. A humanoid wants dozens of compact strain-wave gears. [6268.T](/stocks/nabtesco) is designed into FANUC, Yaskawa, KUKA, and ABB Robotics joints. [HSYDF](/stocks/harmonic) is the listed high-end strain-wave the tape pays for as if Optimus were already the mix. H1 demand at Nabtesco recovered from China and Korea auto-robot capex. That is the installed industrial cycle. RV mini started deliveries in H2 2026; it is a small-series SKU chasing smaller robots, not evidence the skip has arrived. **If the humanoid scales, the joint that gets paid first is Harmonic Drive. If the caged arm restocks, it is Nabtesco. Bundling them is bundling the installed gear with the option.**

**Learned interfaces sort the cohort. A gear has none. A humanoid is trying to skip the ones the installed names have spent decades embedding. ABB's lock is the grid and the process stack, not the arm it is selling.**

| Ticker | Learned Interfaces | Regulatory Lock-In | Bundling | Proprietary Data | Transaction Embedding |
| --- | --- | --- | --- | --- | --- |
| **Already in the cell, the OR, or the grid** |  |  |  |  |  |
| [FANUY](https://investmoat.com/stocks/fanuc) | strong | strong | intact | strong | weakened |
| [6506.T](https://investmoat.com/stocks/yaskawa) | intact | intact | intact | intact | weakened |
| [6861.T](https://investmoat.com/stocks/keyence) | intact | intact | intact | intact | strong |
| [ISRG](https://investmoat.com/stocks/isrg) | strong | strong | strong | strong | strong |
| [ROK](https://investmoat.com/stocks/rok) | strong | intact | strong | intact | weakened |
| [ABB](https://investmoat.com/stocks/abb) | intact | strong | strong | intact | intact |
| **Robotics is the sidecar** |  |  |  |  |  |
| [TER](https://investmoat.com/stocks/ter) | intact | weakened | intact | intact | weakened |
| **Industrial joint vs humanoid joint** |  |  |  |  |  |
| [6268.T](https://investmoat.com/stocks/nabtesco) | na | intact | intact | intact | weakened |
| [HSYDF](https://investmoat.com/stocks/harmonic) | na | intact | intact | intact | weakened |
| **The skip** |  |  |  |  |  |
| [TSLA](https://investmoat.com/stocks/tesla) | weakened | intact | intact | strong | intact |

The matrix is the mechanism, not a decoration. FANUC's panel, G-code dialect, and alarm system are a 40-year muscle-memory lock. Intuitive's console is a career. Rockwell's Studio 5000 ladder is the plant's documented logic. Those three show **strong** learned interfaces. Yaskawa's teach pendant is real and narrower — **intact**, not the global machinist dialect. Keyence's application software is **intact**; the deeper lock is **strong** transaction embedding into the inspection recipe. ABB's 800xA and drive tools are **intact** learned interfaces with **strong** bundling and **strong** grid-code lock-in — on the stack that remains, not on the arm. Teradyne's cobot interface is real and narrower. Harmonic Drive and Nabtesco are *n/a* — a reducer has no operator to retrain. Tesla's learned-interface pillar is **weakened**: FSD and Optimus are supposed to erase the interface, not become one. That is the bull case stated as a moat rating. It is also why bundling the ten names is a category error.

## Qualification is the switching cost. The humanoid thesis is a skip.

The industrial robot that already ships is not a general-purpose worker. It is a serial-numbered device inside a qualified cell. Aerospace machining standards name FANUC CNC models. A Motoman welding cell is PPAP'd to a model and a controller. A hospital credentials a surgeon to a console, then consumes instruments on every case. A pharma or food line is validated to a Logix revision. A Keyence sensor sits in that same validation. An ABB medium-voltage lineup is certified into a site. None of that is a software seat. **A replacement has to re-clear the cell, the credential, or the validation — or skip all three by being a worker that needs none of them.**

- **The cell (FANUY).** Robot-made-by-robots at Oshino, CNC share, and OEM qualification by serial number. Q1 Robot sales grew. Q1 orders that the tape can misread as Robot were FA, including CNC demand FANUC itself tied to humanoid components. Physical AI, on the same call, is still mostly trial implementations.
- **The peer cell (6506.T).** Motoman in welding, handling, and wafer-transfer. Q1 Robotics sales +2.0%; Robotics OP −82.3% on ERP and Europe. The order spike was AC servo +65%. Chinese OEMs compete harder here than they do on FANUC CNC. This is the control that says the robot cycle is not the print.
- **The sensor in the cell (6861.T).** Direct-sales vision and measurement already written into inspection recipes. Q1 operating margin 54.0% on a factory-automation recovery. Swap-out is a process re-qualification. It is not a biped.
- **The procedure (ISRG).** 11,710 systems and a guided 13.5–15.5% procedure year. Instruments are the cash. A competing console has to retrain the surgeon and reopen outcomes data. That is robotics with a medical-device clock, not a humanoid clock.
- **The plant OS (ROK).** Logix and FactoryTalk. A robot, cobot, or humanoid that does useful work in a North American discrete plant still routes through this layer unless the plant rips it out. ARR at +6% is the software attach on that lock, not a robot OEM print.
- **The grid and the sale (ABB).** 800xA, drives, and electrification specified into plants and data centers. Q2 Robotics revenue $576 million, presented as discontinued, against $9.475 billion of continuing-operations revenue. SoftBank is buying the OEM. Rotork is what the proceeds buy. The listed equity after the close is not a robot company.
- **The sidecar (TER).** UR and MiR are the cobot and AMR standards at the human-scale end of the floor. They already displace some caged arms in smaller cells. They are 8% of Teradyne's Q2. The other 92% is test equipment for the chips a robot might someday run.
- **The industrial joint (6268.T).** Cycloidal RV designed into FANUC, Yaskawa, KUKA, and ABB Robotics arms. H1 Component Solutions was 27% of group sales. The rest is doors, rail, aircraft, and marine. RV mini is a small-series chase of smaller robots, not the skip.
- **The option joint (HSYDF).** Strain-wave reducers designed into a platform and tending to stay for that platform's life. FY March 2026 reducer sales were ¥46.3 billion of ¥59.6 billion. The AI-robot orders the filing reports are a beginning. The filing still calls the market nascent.
- **The skip (TSLA).** Optimus is the claim that you do not need G-code, a Motoman teach pendant, a da Vinci credential, or a Logix rewrite — you need a worker. Q2 lists that worker as Construction. Automotive was $20.5 billion of $28.2 billion. Operating margin was 1.4%.

## If the skip works, the cell is the thing being abstracted

The other side is not a cartoon. It is the reason the tape exists. A general-purpose humanoid that can be walked onto a line without rewriting NC programs makes FANUC's learned interface the stranded asset. The operator panel becomes a legacy skill. Yaskawa's teach pendant goes with it. Keyence's inspection recipe is the thing a vision-enabled worker was supposed to make optional. Chinese robot OEMs do not even need the humanoid: they only need good-enough arms at a lower ASP, which is the risk FANUC already names around China capex and which the Q1 Q&A already touched — some China customers took competitor product because FANUC lead times stretched. That displacement can happen while Tesla's manufacturing table still says Construction. The fortress can fail on price without failing on a skip.

Yaskawa is the honest way the skip, or a cheaper skip, shows up first in the installed arm rather than in a biped. Motoman competes with Estun and Inovance in cells that never had a FANUC CNC serial-number lock. Q1 already showed the cost: Robotics revenue almost flat, Robotics operating profit collapsed, while Motion Control did the growing. If Robotics sales inflect and take mix while FANUC Robot orders stay muted, FANUC's CNC fortress is the thing being walked around — not by Optimus, by a peer arm and a Chinese ASP. That would not make Yaskawa a humanoid stock. It would make the FANUC print's Robot line the lagging tell, and the tape would still be wrong about Tesla. Being right about Motoman share is not the same as being right about a worker that needs no cell.

Harmonic Drive is the honest way to hold the other side inside this coverage set. If humanoids scale, each robot needs dozens of precision reducers rather than a handful per industrial arm. Design-in at the joint is a qualification lock of its own, and it does not require FANUC to win the arm. FY March 2026 already records AI-robot orders as customers shift toward mass production. The stock has been paid as if that shift is the company. The filing's own language is still 創成期 — nascent. The counter-case that is actually in the numbers is narrower than the multiple: Harmonic Drive can be right about being in the joint of whoever wins, and still be early. Being early is not the same as being wrong. It is the same as being priced as if the nascent market had already become the mix.

Nabtesco's own script is the industrial side of that argument talking. RV mini deliveries started in H2 2026. Management names Physical AI as a reason to expand the lineup. If compact industrial robots and AMRs take joint count the way a humanoid would, the RV versus strain-wave split collapses from the industrial side — Nabtesco gets paid for smaller robots without Tesla shipping a worker. H1 does not show that. H1 shows China and Korea auto-robot capex plus machine tools and semiconductor equipment, and Component Solutions still 27% of a doors-and-marine group. The counter-case is that RV mini is the beginning of a mix shift the tape is allowed to pay for early. The print is still the caged arm restocking.

Universal Robots is the displacement that does not wait for Optimus. Collaborative arms already take work that used to sit in cages, in plants that never certified a FANUC cell at aerospace serial-number depth. If UR and MiR inflect from $100 million a quarter toward a material slice of Teradyne, the sidecar framing is what breaks — not because humanoids shipped, but because the human-scale end of the floor was the real robotics market and TER owned it. That would not make TER a humanoid stock. It would make the 8% line the growth the multiple had been borrowing from Semiconductor Test. Watch the mix, not the label.

ABB's sale is the counter-case that the tape can be wrong on mix and still own a compounder. SoftBank is paying $5.375 billion of enterprise value for the robot OEM. Masayoshi Son's language on the October 8, 2025 announcement was Physical AI. That is a bid on the skip, from a buyer who is not pretending the listed ABB equity is the OEM after the close. If the sale closes in H2 2026 and Rotork is absorbed on the guided timetable, ABB shareholders own electrification funded by someone else's humanoid cheque. The tape that kept ABB in the robotics basket would have been wrong about what the equity was, and right that robotics was worth a strategic premium to someone. Being wrong about the mix is not the same as robotics being worthless. It is the same as owning the wrong residual.

Keyence's Q1 can be read the same way from the inspection layer. A 54.0% operating margin and Asia sales +33.9% in local currency is a factory-automation recovery the quality-industrial multiple already knows how to pay. It is also what a physical-AI build looks like if every new cell still needs vision. The skip thesis says the worker replaces the recipe. The Keyence print says the recipe is compounding. Both can be true for a while: more cells, more sensors, and a humanoid still in Construction. The counter-case that matters is not that Keyence is a robot company. It is that paying a robotics multiple for a sensor already in the cell is cheap if the cell count is what actually grows, and expensive if the multiple was for Optimus.

Intuitive is the other place the skip thesis can be half-right. Competing surgical platforms do not need to be humanoids. They need to be good enough that a hospital trains a second cohort of surgeons. If da Vinci procedure growth settles below the 13.5% floor of the FY2026 guide, the lock was category leadership, not a monopoly, and the robotics-as-procedure story still holds — just with a thinner franchise. That is a medtech share fight. It is not evidence that Optimus belongs in the same basket. Mixing those two failures is how the label keeps getting paid.

Tesla can still be the skip. Construction is not a refusal. Fremont S/X lines were decommissioned for it. Production is anticipated later this year, and the Q2 table lists capacity at both Fremont and Texas as Construction, not Production. The cars-and-energy company underneath printed $28.2 billion of revenue at a 1.4% operating margin. A construction status on a second product line does not make the first product line a robot company. It also does not make construction a rounding error if units ship to customers. The test is customer revenue, not a factory tour.

## Customer Optimus against falling FANUC Robot orders — or a gear book that is already majority-humanoid

- **The skip.** Tesla disclosing external commercial Optimus revenue in a FY2027 quarter — customer shipments, not internal training units — while FANUC Robot-division orders fall year over year in both halves of FY2027. One quarter of Robot softness is China capex. Two halves plus customer humanoids is the cell being skipped.
- **The joint.** Harmonic Drive disclosing AI-robot orders as more than half of consolidated orders for two consecutive halves while industrial-robot demand is disclosed as flat or down. That would show the tape was right to pay for the option in the gear rather than the arm.
- **Not the test.** A FANUC FA-order quarter that includes humanoid-component CNC. A Yaskawa AC-servo quarter at +65% while Robotics sales stay +2%. A Teradyne robotics print that stays in the $100 million area while Semiconductor Test sets records. An ABB SoftBank close. An Optimus factory photograph. A Keyence Asia print above 30% local currency. Those are the tape. They do not move the lock.

> **Status as of September 19, 2026** — Holding. FANUC's Q1 order spike is still a FA print, and the company said so. Yaskawa's Q1 order spike is still servo, and Robotics operating profit collapsed on ERP and Europe. ABB's Q2 still presents Robotics as discontinued operations awaiting a SoftBank close. Tesla's manufacturing table still says Construction for Optimus. Harmonic Drive's FY March 2026 filing still calls AI robots nascent. Nabtesco's H1 Component Solutions bounce is still industrial-arm restocking. Teradyne Robotics is still $100 million in a $1.33 billion quarter. No subsequent print has shown customer Optimus revenue or a Robot-division order decline against a humanoid mix shift.

The lesson that generalises is narrower than "avoid robotics." Own the name whose unit is already qualified — the cell, the peer cell, the sensor in the recipe, the procedure, the plant OS, the grid lineup — on a print that is that unit. Do not own a CNC recovery, a servo cycle, an ATE print, a robot-OEM sale, or a construction line because the word robot appears in the same sentence as AI. The basket already has a ticker. It is the one that cannot tell a yellow arm from a parts list.

## Sources

1. [FANUC Financial Results for the Three Months Ended June 30, 2026](https://www.fanuc.co.jp/en/ir/announce/pdf/2026/reference202606_e.pdf) — FANUC Corporation, July 31, 2026 (press-release)
2. [Q&A Summary of the Telephone Conference for FY2026 1Q](https://www.fanuc.co.jp/en/ir/announce/pdf/2026/qasummary202606_e.pdf) — FANUC Corporation, July 31, 2026 (transcript)
3. [Yaskawa Consolidated Results for the First Quarter of Fiscal Year Ending February 28, 2027](https://www.yaskawa-global.com/wp-content/uploads/2026/07/20260710_en.pdf) — Yaskawa Electric Corporation, July 10, 2026 (press-release)
4. [Yaskawa FY2026 1Q Financial Results Briefing Materials](https://www.yaskawa-global.com/wp-content/uploads/2026/07/20260710_haifu_en.pdf) — Yaskawa Electric Corporation, July 10, 2026 (company-site)
5. [Q&A for FY2026 1Q Results Briefing (Summary)](https://www.yaskawa-global.com/wp-content/uploads/2026/07/261Q_QA_EN.pdf) — Yaskawa Electric Corporation, July 10, 2026 (transcript)
6. [KEYENCE First Quarter of FY2026 Financial Results Materials](https://www.keyence.co.jp/pdf/FinancialResults_202607_en.pdf) — KEYENCE Corporation, July 28, 2026 (press-release)
7. [ABB Q2 2026 results](https://new.abb.com/news/detail/137496/q2-2026-results) — ABB Ltd, July 16, 2026 (press-release)
8. [ABB Q2 2026 Group results presentation](https://library.e.abb.com/public/47f0bfb2ecaf494b99846882bffa5409/ABB-Q2-2026-Group-results.pdf) — ABB Ltd, July 16, 2026 (company-site)
9. [ABB Q2 2026 Financial Information — discontinued operations (Robotics)](https://www.schweiz.biz/wp-content/uploads/ABB-Q2-2026-Financial_Information.pdf) — ABB Ltd (Financial Information booklet), July 16, 2026 (third-party)
10. [ABB to divest Robotics division to SoftBank Group](https://new.abb.com/news/detail/129685/abb-to-divest-robotics-division-to-softbank-group) — ABB Ltd, October 8, 2025 (press-release)
11. [Nabtesco Summary of Consolidated Financial Statements for FY2026 Q2](https://www.nabtesco.com/cms/wp-content/uploads/Summary_of_Consolidated_Financial_Statements_for_FY2026_Q2_e.pdf) — Nabtesco Corporation, July 31, 2026 (filing)
12. [Nabtesco Results Briefing Material for FY2026 Q2 (script)](https://www.nabtesco.com/cms/wp-content/uploads/Results_Briefing_Material_for_FY2026_Q2e_script.pdf) — Nabtesco Corporation, July 31, 2026 (transcript)
13. [Intuitive Announces Second Quarter Earnings (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/1035267/000103526726000047/q226ex-991earningsrelease.htm) — Intuitive Surgical (SEC EDGAR), July 16, 2026 (filing)
14. [Rockwell Automation Reports Second Quarter 2026 Results (Exhibit 99)](https://www.sec.gov/Archives/edgar/data/1024478/000102447826000020/q2fy26ex99.htm) — Rockwell Automation (SEC EDGAR), May 5, 2026 (filing)
15. [Teradyne Reports Second Quarter 2026 Results (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/97210/000119312526321933/ter-ex99_1.htm) — Teradyne (SEC EDGAR), July 28, 2026 (filing)
16. [Tesla Q2 2026 Update (Exhibit 99.1)](https://www.sec.gov/Archives/edgar/data/1318605/000162828026049213/exhibit991.htm) — Tesla (SEC EDGAR), July 22, 2026 (filing)
17. [Harmonic Drive Systems Consolidated Financial Results for FY March 2026 (Japanese GAAP)](https://www.nikkei.com/markets/ir/irftp/data/tdnr/tdnetg3/20260513/fx6gjx/140120260511521357.pdf) — Harmonic Drive Systems (TDnet), May 13, 2026 (filing)

---

Scores are computed deterministically from each asset's data file: composite = Moat^0.40 × Growth^0.30 × Valuation^0.30, scaled to 0–100.

Full analyses for every name above: https://investmoat.com/stocks. Site index for agents: https://investmoat.com/llms.txt

InvestMoat is an open-source research and education framework. Nothing here is financial advice.