Day08 | Industrial robots: Why are record installations not lifting all profits?

This issue corresponds to 2026-11-2 in Beijing; the actual research cutoff is 2026-11-1 at 20:07:02 UTC. Every reference to “latest” is limited to that cutoff. The latest complete year for global volumes is 2025; listed companies' latest reporting periods differ. Amounts remain in their original currencies and are not aggregated across currencies. This is a research record, not investment advice.

1. One-sentence core judgment

Installation growth does not mean profits rise everywhere; the ability to turn equipment into reliable output deserves more scrutiny than shipments alone.

New industrial robot installations continue to rise globally, but supplier profits, regional demand and application categories are not moving together. Three questions must remain separate: how many more robots factories install, how much money suppliers retain, and how soon customers receive a cash return. This report supports divergence in the first two. The third remains an operating hypothesis requiring project-level evidence, not an already measured industry-wide migration of profits.[E01][E06][E08][E10][E26]

2. Industry boundaries and revisions

2.1 What this issue covers—and excludes

The scope is the industrial robot value chain: precision reducers, servos and controls, robot hardware, application integration, commissioning and acceptance, maintenance and upgrades. Under the ISO definition used by IFR, the core product is an automatically controlled, reprogrammable, multipurpose manipulator for industrial environments. Collaborative robots are an application/product category within industrial robots and must not be added separately to their total.[E20]

Humanoid fundraising, service robot shipments, warehouse AMR volumes, revenue across all industrial automation, and the AI model market are excluded from market sizing. Teradyne's Robotics segment includes collaborative arms and mobile robots and is used only as a company-level operating counterexample. Yaskawa's Motion Control includes non-robot applications and is used only as an adjacent-business comparison. Without sufficiently pure disclosures, pure industrial robot profits must not be invented.[E08][E10]

Unlike Day01 AI computing, Day02 battery storage, Day03 desalination, Day04 copper, Day05 semiconductor manufacturing equipment, Day06 power transformers and Day07 commercial satellite launch, this issue examines factory automation equipment and delivery economics. Semiconductors and automotive appear only as downstream demand, not as repeated value-chain studies.

2.2 Five measures that cannot be interchanged

1. Orders: customer commitments that may not yet be delivered; A3's North American data measure this stage.
2. New installations: equipment put into factory use during the year; IFR's global data measure this stage.
3. Operational stock: accumulated installations less retirements; it cannot be added to annual installations.
4. Revenue and profit: potentially different products, regions and recognition dates; group margins are not robot segment margins.
5. Forecasts and actuals: an industry report published in 2026 primarily measures 2025 and does not establish completed full-year 2026 results.[E01][E02][E09][E28]

This table identifies what each market statistic measures rather than ranking unlike quantities.

Forecast growth uses the institution's original rounding. Recalculating from already rounded absolute figures may not reproduce the same integer percentage precisely.

2.3 Revise the evidence before the thesis

Earlier industry forecasts are not treated as actuals, outdated financial guidance is not treated as current, and group data do not fill segment gaps. Nabtesco FY2026 H1 uses the 2026-09-30 revision. Its Q3 release was postponed to 2026-11-13, so the original scheduled release date does not mean results were published. ABB's historical robot segment data retain their FY2024 label and are not replaced with newer group margins.[E07][E12]

3. Verdicts on four falsifiable hypotheses

H1: An installation recovery will improve operating margins simultaneously for robot makers and core components

Verdict: reject the strong, universal and simultaneous version; retain the possibility of company-specific improvements. The latest earnings directions of Nabtesco CMP and Yaskawa Robotics differ. ERP implementation disruption and European restructuring also affected the latter, so not every decline can be attributed to price competition. Teradyne's revenue recovery likewise did not translate into a proportional rise in segment pre-tax profit.[E07][E08][E10]

The point is not to infer an industry average from a small sample, but to find counterexamples sufficient to disprove “every stage benefits.” The profit pool remains unmeasured. Gross profit, operating profit, pre-tax profit and Operational EBITA are not interchangeable metrics.

Falsification test: At the next annual disclosures, hold a comparable sample of robot/component segments fixed and test whether revenue-growing businesses also improve operating margins on the same basis. Persistent cases of rising revenue and falling margins would continue to invalidate the strong version. Broad, sustained improvement would warrant a stronger assessment of recovery breadth, but still would not reveal undisclosed global profit shares.

Confidence: high for divergence; no verdict on global profit shares by stage.

H2: Growth in China's market and domestic supplier volumes necessarily means further domestic share gains

Verdict: contradicted by 2025 actuals. Total Chinese installations grew faster than domestic suppliers' volumes. Domestic absolute volume rose, but share fell from 57% in 2024 to 55% in 2025. This measure follows supplier origin: it is not Chinese manufacturing content or the domestic share of critical components.[E04][E05]

Falsification test: Compare China's total installations, domestic supplier installations and domestic share using the same IFR definitions next year. Domestic sales growth announcements alone do not prove faster substitution. Local manufacturing by overseas companies is not automatically domestic-brand share.

Confidence: medium. The primary statistics are authoritative and internally explicable, but media largely repeat IFR rather than provide an independent census with matching definitions.

H3: Collaborative robots must be the fastest-growing industrial category and eliminate application safety and integration work

Verdict: neither strong claim holds. IFR reports approximately 66,000 global collaborative installations in 2025, an 11% share and +7% growth, below overall growth of +11%. This does not remove opportunities for collaborative products. A3 reports faster collaborative order growth in North America in 2026 Q3, showing how geography, period and measurement change the conclusion.[E03][E28]

On safety, ISO 10218-1 addresses the robot itself and ISO 10218-2 the complete application and cell. Manufacturers also explicitly require application risk assessments. Safety features on an arm do not automatically make its gripper, workpiece, speed and site layout safe.[E25][E26][E29]

Conflict record [conflict E03-vs-E28]: IFR's global annual installations and A3's North American quarterly orders show different growth patterns, not a contradiction within the same population. Retain both: North American orders do not overturn historical global installations, and historical global installations do not invalidate a new quarter's orders.

Falsification test: Compare collaborative and overall growth on the same geographic, temporal and measurement basis. For delivery complexity, examine risk assessments, commissioning hours and acceptance records. An out-of-the-box standardized application does not establish that every application needs no integration.

Confidence: high for safety responsibilities; medium for statistical divergence and future growth.

H4: Easier hardware availability automatically transfers profits to integrators and software service providers

Verdict: necessity is supported; profit migration is unproven. Application safety and commissioning cannot be skipped, and IFR identifies validation, safety and skills constraints. But a customer's need to perform an activity does not guarantee supplier pricing power. Fixed-price projects, rework, delayed acceptance and receivables can turn engineering workload into cost rather than profit.[E16][E26][E29]

Falsification test: Track comparable integrators' consistently defined gross margins, operating cash flow, acceptance cycles and aftermarket revenue—not merely order values. Only shorter commissioning, greater reuse, better collections and sustained profit gains together would justify upgrading “delivery capability merits attention” to “profits are migrating.” Public samples are insufficient; no software/integration profit-pool share is reported.

Confidence: medium for delivery constraints; no verdict on the direction or magnitude of profit migration.

4. Asset one: Profit map—where does money remain?

4.1 Observable profits, not an invented profit pool

This table asks what public accounts actually allow us to compare. All amounts retain their original currencies; JPY billion and US$ million mean billions of yen and millions of dollars respectively. Unless specified otherwise, margins are rounded company disclosures.

Calculation basis: Yaskawa Robotics = 5.974 / 118.404; Motion Control = 15.973 / 137.286; Teradyne = 0.345 / 95.828. Percentages are rounded to 2 decimal places. These are ratios of reported financial figures, not new industry statistics.[E08][E10]

4.2 Profit transmission pathways

Upstream: Precision, reliability and customer qualification create entry barriers. But new-factory depreciation, utilization, cost improvements and negotiation jointly determine operating profit. Nabtesco's new factory and differing utilization across sites caution against jumping from “technically important” to “all capacity is scarce.”[E06][E22]

Robot hardware: Product platforms, installed bases, channels and process expertise may support repeat purchases. Demand volatility, pricing, product mix and execution matter too. Yaskawa's guidance even combines an upward sales revision with a downward profit revision—another reason not to equate revenue and earnings.[E23]

Integration and operations: Whoever takes responsibility for process fit, cycle time, quality, exceptions and acceptance is closer to the customer's economic problem. Yet bespoke engineering labor is not infinitely scalable, and bidding competition and collection terms can consume the value. This is a proposed mechanism to test, not a label of already-realized high profits for a particular stage.[E26][E29]

Integration trend: Nabtesco's investment in Alva shows exploration of motor, reducer and control synergies. It does not establish scaled revenue from new products; no usable investment amount is available for estimating a profit pool.[E24]

4.3 The attractive chart this issue will not draw

There is no pie chart assigning global profits to components, robot hardware and software. Consistent revenue, profit and intercompany-elimination data are missing. Summing listed-company sample profits would mix products, double-count and omit geographies. The more reliable map is observable segments plus explicit blanks, as above.

5. Asset two: Bottleneck urgency ranking

5.1 Global ranking: analytical priorities, not measured market scores

The table asks where project due diligence should start. The ranking follows standards, disclosed operating changes and market structure, without unsupported precise scores.

5.2 Regions require different explanations

• China: Large-scale installations and growing domestic volumes are established, but share changes and sharp application divergence require separating supply expansion from profit improvement. Prioritize customer concentration, collections, reliability and actual acceptance; domestic branding is not domestic component content.[E04][E05][E15]
• United States / North America: US installations and North American orders are not one market measure. Imported equipment, local engineering services and policy uncertainty must be assessed together. Recovering North American orders do not automatically establish ample US-made robot supply.[E13][E27][E28]
• European Union: Annual EU27 installation declines relate to industry demand; population aging alone cannot explain short-term orders. Examine automotive and metalworking budgets, restructuring and project delays first. The EU is not all of Europe.[E14][E10]
• Japan: Domestic installations and Japanese suppliers' global sales are different. Overseas business may offset a weakening domestic market. FANUC group profits cannot establish returns on new installations within Japan.[E09][E30]
• Newer markets, including the Gulf: This research did not obtain sufficiently granular, consistently defined industrial installation and integration-profit data. Factory investment news cannot establish a bottleneck ranking. Local service, industry structure and compliance belong on project verification lists, not in claims of proven shortages.

6. Asset three: Timeline of profit migration

6.1 What has actually happened

This table asks how the evidence changed over time, rather than stitching different vintages into a seemingly precise series.

Conflict record [conflict E19-vs-E18]: Preliminary and final global installations for the same year differ. This report uses the final figure for history and retains the preliminary figure to illustrate revision risk. They are neither averaged nor treated as additional installations.

Conflict record [conflict E11-vs-E01]: The old forecast for 2025 differs from subsequent actuals, and the forecast for 2026 has also been updated. This is a forecast-vintage change, not conflicting sets of realized facts. Forward assessment uses the forecast published on 2026-09-24 without guaranteeing realization.

6.2 A defensible sequence—and unproven stories

The business sequence is: validate customer needs and returns → place orders → supplier production → site integration and risk validation → acceptance → collection → maintenance and upgrades. Companies recognize revenue and bear inventory and collection risk at different stages. Rising total installations and a company's current profit therefore need not move together.

Unproven stories include “falling hardware prices automatically send all profit to software,” “easier AI programming eliminates all integration costs,” and “more installed robots necessarily mean higher-margin service revenue.” All lack longitudinal profit and cash-flow evidence for consistent samples and definitions. This report does not turn those stories into trend charts.

Next, observe whether standardized process packages genuinely reduce repeat commissioning, installed-base services produce verifiable renewals, and suppliers price delivery risk rather than absorb it for free. The verification window is the next reporting round and comparable project reviews—not an unsupported precise date for a profit inflection.

7. Asset four: Counterfactuals and three stress tests

All tests below are scenario assumptions and mechanical calculations, not actual outcomes, company guidance or industry forecasts. They identify sensitivities rather than value companies. Inputs, units, formulas and limitations are explicit. Apart from actual financial/installation baselines, parameters are set by this report.

Test one: With unchanged volume and a 5% price decline, what happens to component profit?

The actual starting point is revised Nabtesco CMP FY2026 H1 revenue of JPY45.521 billion and operating profit of JPY4.841 billion.[E07] Assume unchanged volume, product mix and all costs, with average selling prices falling 5%. This is not a company-disclosed price sensitivity.

• Baseline costs = 45.521 − 4.841 = JPY40.680 billion.
• Scenario revenue = 45.521 × (1 − 5%) = JPY43.24495 billion.
• Scenario operating profit = 43.24495 − 40.680 = JPY2.56495 billion.
• Scenario operating margin = 2.56495 / 43.24495 = 5.93%.
• Profit decline = (4.841 − 2.56495) / 4.841 = 47.02%.

Transmission: lower prices → lower revenue → temporarily unchanged costs → amplified profit erosion. The counterfactual shows that technological barriers do not remove sensitivity. Procurement savings, higher volumes and expense cuts could offset the effect in practice; this profit figure is not a forecast.

Test two: Equipment is purchased, but integration overruns, savings disappoint and production starts late

Use a currency-free project index, not an actual factory project. Assume initial investment of 100 and annual net cash savings after commissioning of 40. Baseline simple payback = 100 / 40 = 2.50 years. Net savings already deduct ordinary operating expenses; do not deduct them again.

Stress assumptions: total investment overruns by 20%, steady-state annual net savings are 20% below plan, and commissioning is delayed by 0.50 years.

• Scenario investment = 100 × 1.20 = 120.
• Scenario annual net savings = 40 × 0.80 = 32.
• Simple payback after commissioning = 120 / 32 = 3.75 years.
• Approximate payback from initial expenditure = 0.50 + 3.75 = 4.25 years.

Transmission: poor site fit → extra engineering and waiting → higher costs and later benefits → pressure on customers' next investment round. This simplified model assumes concentrated initial spending and no benefits during delay. It excludes discounting, taxes, residual value, staged payments, capacity-expansion benefits and safety benefits, and cannot replace a full project cash-flow model. Its direction is consistent with standards requiring complete-application validation, but its parameters are not industry averages supplied by those standards.[E26][E29]

Test three: If Chinese installations fall 10% and other markets are unchanged, what happens globally?

The actual baseline uses IFR's rounded 2025 figures: 603,000 global installations and 354,000 in China.[E01][E04] A 10% Chinese decline with other markets unchanged is a stress assumption, not a forecast for 2026.

• Other-market baseline = 603,000 − 354,000 = 249,000 units.
• China scenario installations = 354,000 × 0.90 = 318,600 units.
• Global scenario installations = 318,600 + 249,000 = 567,600 units.
• Global change from baseline = (567,600 / 603,000 − 1) = −5.87%.

Transmission: Chinese demand changes → global volumes change → suppliers adjust utilization and mix. This does not imply an identical global revenue or profit decline: prices, regional shares, services and exchange rates are not modeled. The test illustrates concentration sensitivity, not an inevitable Chinese decline. It also must not be merged with IFR's 2026 forecast as though both describe the same scenario.

Combined operating implications

Price, delivery and regional concentration are distinct risk axes. Higher shipments may buffer price pressure but increase integration burdens. Better equipment performance may raise hardware prices while improving total customer returns through faster acceptance. Industry judgments should trace these chains, not select just one growth number.

8. Asset five: Evidence gaps and unresolved questions

1. Missing global profit pool. Public company coverage is incomplete, business scopes and accounting differ, and currencies cannot simply be added. Value-chain profit shares require consistent segment finances and intercompany eliminations.
2. Missing project-level returns. Too few independent factory samples were obtained to verify commissioning hours, acceptance delays, realized net savings and collection cycles. Delivery capability is a priority hypothesis, not a measured industry-wide top shortage.
3. Insufficient independent second sources on the same basis. IFR installation data are widely reproduced but remain one underlying statistical system. A3 is an independent orders survey that can corroborate demand direction, not precisely verify IFR global units.
4. Different financial scopes. CMP, Motion Control and Robotics vary in business purity. FANUC has no usable standalone robot segment profit in this research. ABB's historical anchor is not continuous with post-transaction disclosure.[E07][E09][E12]
5. Uneven recency. Complete industry years lag the current quarter. Nabtesco's latest verified actuals remain revised H1, with Q3 delayed. Undated static pages support definitions, not latest market size.[E07][E20]
6. Forecasts remain revisable. IFR forecasts, Nabtesco guidance and Yaskawa guidance differ in issuer and object. Market forecasts do not guarantee company orders; outdated guidance cannot chart current expected profits.[E01][E07][E23]
7. Policy remains unsettled. The 2026-10-28 US announcement opens negotiations and an update process; it does not itself impose a new uniform robot tariff. Landed costs still require product- and origin-specific verification. This is not legal or customs advice.[E27]
8. AI and new-form-factor commercialization. Demonstrations, investment and integration plans do not establish scaled profits. Without comparable task reliability, maintenance costs and customer payment data, humanoid and AI software revenue are excluded.[E16][E24]

The next update needs more than “large market” material: comparable FY2026 full-year segment profits, installation-to-acceptance project samples, revisions to share statistics, and evidence that recovering orders become deliveries with cash returns.

9. Evidence ledger

The ledger separates facts, forecasts, guidance, definitions and analytical foundations. An independent second source must collect information independently; media repetition or another document from the same company does not qualify. High confidence means a document actually discloses a figure or rule, not that a future outcome is guaranteed. All links were checked or retrieved before the cutoff; limited access to full texts is disclosed.

Conflict record [conflict E04-vs-E30]: Specialist media call China's 59% a share of the global installed base; the original IFR figure measures annual new installations. This report follows the primary source's flow measure and rejects the operational-stock characterization. More reproductions do not equal better independent verification.

Conflict record [conflict E01-vs-E13]: The global news summary gives US installations as nearly 38,500, whereas the US-specific material says approximately 38,400. This report uses the dedicated release and retains the rounding difference. It does not claim precision to individual hundreds or interpret the discrepancy as a change in demand direction.

Currency conversion for checking Chinese cards: JPY79.3 billion = 793 hundred-million yen; JPY5.4 billion = 54 hundred-million yen; JPY67.6 billion = 676 hundred-million yen; JPY2.7 billion = 27 hundred-million yen. These are unit conversions, not additional estimates. E06 annual operating margins retain company disclosures rather than being replaced by recalculations from rounded amounts.

10. References and source tiers

Tier A: Company financial and regulatory disclosures

Core materials are Nabtesco FY2025 and revised FY2026 H1, Yaskawa FY2025 and FY2026 H1, Teradyne SEC 10-Q, FANUC FY2026 H1 and ABB's historical robot disclosures.[E06]–[E10][E12][E21]–[E23] Original links are in the ledger. Priority is revised financial statements → contemporary financial statements → management presentations. Management explanations and accounting figures are distinguished; an explanation does not prove that all other causes are excluded.

Tier B: Original industry association statistics and forecasts

IFR supplies global installations and operational stock; A3 supplies North American orders. They are independent but differ in survey objects, geography and time.[E01]–[E05][E11][E13]–[E20][E28] Original release dates distinguish forecast vintages. The complete databases were not purchased; this report does not imply examination of unobtained country/company-level underlying records.

Tier C: Specialist media

Drives & Controls helps check public narratives and identify mislabeled measures. It is not an independent global installation census.[E30] Reviewing secondary sources takes precedence over superficial agreement across multiple websites.

Tier D: Standards, government and company technical announcements

The two ISO safety standards, White House announcement, UR technical FAQ and Nabtesco strategic announcement establish responsibility, policy and technical boundaries, not market-size estimates.[E24]–[E27][E29] Standard summaries and previews are not full implementation manuals. Project safety design requires complete standards and qualified professional assessment.

Calculations, updates and the next issue

Ratios and the three stress scenarios can be recalculated formula by formula in sections 4 and 7. All scenario parameters are analyst-set and not mixed with actuals in the evidence ledger. Single sourcing is disclosed rather than disguised with more repetitions.

The conclusion remains restrained: data support a global installation recovery, company counterexamples establish profit divergence, and standards establish the need for engineering and safety validation of complete applications. There is still insufficient evidence that profits must migrate to a particular stage. The proposed next issue is Day09 | Civil aircraft engines, avoiding previously covered industries. That is an editorial plan, not an industry forecast.