Modeled estimate · Bipedal humanoids
Scenario range 31.4k–79.0k/Inspect assumptions
General-purpose bipeds, including research and internal fleets. The clock follows a production model; it is not live factory telemetry.
Draft model · Reviewed 2026-09-05
2020 → 2030 / Production, over time
From research fleets to a possible manufacturing scale-up. Explore reconstructed history and the range of paths to 2030.
Current estimate · Sep 5, 2026
≈47,875
Cumulative humanoids built
Assumption scenarios
31.4k–79.0k
Not statistical confidence limits
2026 estimated output
≈45,307/ year
27.8k–77.6k · includes forecast
2020–2026 · independent linear scale
Same model, closer view. Select a date here or use the shared date slider. The dashed section includes modeled production since July 2026.
Quarterly production follows the manufacturer's baseline, ramp delay, growth and execution/demand assumptions. Company ambitions remain separate. The reported sample is partial and may be disclosed retrospectively.
Unitree · ≈17,881 built in 2026
Includes forecast assumptions · 2023/2024 use reported production. 2025 uses the >6,500 company floor with inventory uncertainty. H1 2026 converts >7,000 shipments to 7,000/8,200/11,200 built; no repeated addition of shipments.
AGIBOT · ≈14,761 built in 2026
Includes forecast assumptions · Mixed milestones are never summed. 2024 biped share is uncertain. Central June 2026 cumulative biped estimate uses the earlier 71.76% family-share proxy; A2-W makes this an assumption, not a measured biped share.
UBTECH · ≈2,545 built in 2026
Includes forecast assumptions · Early Walker cohorts are editorial allocations, not measured totals. Later periods use shipment/sales proxies with biped-share and inventory uncertainty; toy and wheeled lines are excluded.
Leju Robotics · ≈1,744 built in 2026
Includes forecast assumptions · Kuavo-scale platforms only. Dedicated small education products are excluded. Earlier zeros are scope allocations; later figures use shipment proxies plus modeled inventory.
Figure · ≈1,411 built in 2026
Includes forecast assumptions · Prototype and older-generation cohorts are reconstructed. H1 2026 includes the >350 Figure 03 milestone and uncertain subsequent builds; a demonstrated hourly cycle does not establish 24/7 output.
EngineAI · ≈1,328 built in 2026
Includes forecast assumptions · General-purpose bipeds only. Shipment estimates are proxies; 2024 and H1 2026 allowances are explicitly modeled.
Fourier · ≈872 built in 2026
Includes forecast assumptions · GR-series bipeds only; rehabilitation devices excluded. Early cohorts and inventory are weakly evidenced reconstruction allowances.
Agility Robotics · ≈371 built in 2026
Includes forecast assumptions · Only complete Digits enter this row; leg-only Cassie robots excluded. Ford's two deliveries are one observation. Remaining early production is an uncertain research/pilot allowance.
Tesla · ≈604 built in 2026
Includes forecast assumptions · Actual internal fleet totals are not public. These are wide editorial prototype/pilot scenarios, not a claim that management's targets were achieved or missed.
Boston Dynamics / Atlas · ≈60 built in 2026
Includes forecast assumptions · Hydraulic and electric Atlas are distinct physical cohorts. Seven DARPA platforms form an old sample floor; later prototype and product cohorts are assumptions.
Legacy research programs · ≈183 built in 2026
Includes forecast assumptions · Disjoint legacy/research allowance for Honda, PAL, HUBO, HRP and comparable institutional bipeds, excluding every named maker. Program evidence supports existence, not these quantities. All quantities are editorial scenarios.
Other manufacturers · ≈3,541 built in 2026
Includes forecast assumptions · Disjoint emerging-maker cohort: Apptronik, 1X bipeds, newer Chinese, European, Indian and other programs. Excludes the named makers and legacy research row. EVE and other wheeled products are excluded. Missing coverage is modeled, never an invented reported total.
Early years depend on an estimated pre-2020 opening balance and incomplete research-fleet disclosures. Use History to inspect the smaller numbers. Forecasts through the end of 2030 are assumption scenarios, not guaranteed bounds. Capacity, shipments and company targets remain separately labeled.
The 2030 outlook / Draft
≈1.29M
Cumulative built by Dec 31, 2030 · central scenario
The central scenario reaches approximately 1.29M cumulatively built, with 658.4k produced during 2030. The 163.0k–9.37M range reflects radically different execution and demand assumptions. It is not a probability interval.
≈386.2k
Cumulative built by 2030 · central. Growth 2×/year; ramp delay 0 quarters.
≈380.7k
Cumulative built by 2030 · central. Growth 1.8×/year; ramp delay 0 quarters.
≈165.0k
Cumulative built by 2030 · central. Growth 4.5×/year; ramp delay 2 quarters.
The largest contributions set the scale of this outlook. Ramp delays, annual growth and the combined capacity, utilization, yield, components and demand ceiling drive the differences between scenarios.
| Year | Built that year | Cumulative built |
|---|---|---|
| 2026 | ≈45,307 | ≈64,315 |
| 2027 | ≈80,525 | ≈144,840 |
| 2028 | ≈159,119 | ≈303,959 |
| 2029 | ≈324,233 | ≈628,192 |
| 2030 | ≈658,426 | ≈1,286,617 |
Quarterly output grows from each manufacturer's H1 2026 modeled daily baseline after a disclosed ramp delay. It is capped by nominal 2030 production potential × utilization × yield × component availability × demand share. All factors below are assumptions; capacity is not reported current output. Existing baseline output continues during the delay.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling. The eligible holdout selected constant-growth; its implied annual growth is capped at 2× in the central long-term case. The cap is editorial, not fitted to the holdout.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
All growth, delays, capacity paths, utilization, yield, component and demand factors are explicit editorial assumptions, informed by the linked disclosures. Capacity2030 is a scenario ceiling, not installed capacity. During ramp delays, baseline output remains subject to the same execution and demand ceiling.
Model assumptions reviewed 2026-09-05 · source dates below identify the underlying disclosures.
Annual output, capacity and supply-chain plans describe different quantities. They are not added to the cumulative estimate. A missing outcome is marked unverified, not automatically missed.
10,000 · annual output
Initial 2025 internal plan: roughly 10,000
Announced 2025-01-29 · horizon 2025-12-31 · revised · reviewed 2026-09-05
Musk qualified this as ambitious. March remarks focused on 5,000, with 10,000 as an upper ambition. No comparable actual full-year total has been established.
Our central 2025 reconstruction: ≈200 produced that year. This is modeled, not a disclosed outcome.
5,000 · annual output
March 2025 ambition: 5,000, with 10,000 at the upper end
Announced 2025-03-20 · horizon 2025-12-31 · unverified · reviewed 2026-09-05
No comparable company-reported 2025 total was found. July 2026 describes successor lines under construction; that does not quantify 2025 pilot output. Not scored as achieved or missed.
Our central 2025 reconstruction: ≈200 produced that year. This is modeled, not a disclosed outcome.
50,000 · annual output
2026 production ambition
Announced 2025-03-20 · horizon 2026-12-31 · future · reviewed 2026-09-05
Year still in progress; July 2026 filing describes line installation. Central model includes slower ramping and broad uncertainty.
Our central 2026 scenario: ≈604 produced that year. This is modeled, not a disclosed outcome.
1,000,000 · annual output
One million Optimus per year by 2030, potentially 2029
Announced 2025-04-22 · horizon 2030-12-31 · future · reviewed 2026-09-05
Management ambition, not a cumulative fleet estimate or a guaranteed production rate. The model uses explicit execution and demand assumptions.
Our central 2030 scenario: ≈127,949 produced that year. This is modeled, not a disclosed outcome.
12,000 · capacity
BotQ annual line design capacity
Announced 2025-03-15 · horizon not dated · future · reviewed 2026-09-05
Disclosure does not give a firm date for full utilization. April 2026 separately reports >350 builds; line capability is not completed production.
100,000 · supply chain capability
Supply chain able to scale to 100,000 robots over four years
Announced 2025-03-15 · horizon 2029-03-15 · future · reviewed 2026-09-05
Not 100,000 robots already built, orders, or an unconditional production commitment.
30,000 · capacity
Scalable system capable of 30,000 robots annually by 2028
Announced 2026-01-05 · horizon 2028-12-31 · future · reviewed 2026-09-05
Future robot-production system target; actual Atlas mix, commissioning, utilization and demand remain uncertain.
1 eligible holdout · mixed-form production · global accuracy not established
Only one eligible, mixed-form holdout has three earlier comparable dated milestones. It cannot validate a global biped forecast through 2030. Unitree's 2023/2024 production was published in 2026 and cannot be used to pretend a 2024 backtest was possible. Lower-bound shipment reports are not exact production outcomes.
AGIBOT holdout: forecast on 2026-03-30 → observed 2026-06-28. Actual cumulative mixed-form production: 15,000. Best short-horizon candidate: constant growth.
| Method | Prediction | Absolute error |
|---|---|---|
| constant output | 14,018 | 982 |
| constant growth | 15,290 | 290 |
| damped ramp | 14,597 | 403 |
Only sources published by the cutoff enter the inputs. Constant growth uses the change between the two preceding interval rates; the ramp comparator halves that log-growth rate. The preferred family informs AGIBOT's short-term direction. Long-term growth is capped at an explicitly assumed 2× annually and constrained by execution/demand ceilings. None of these results validates a 2030 global prediction.
Explore the documented factories powering the humanoid robotics industry, from active production lines to formally announced projects.
Explore global factory map
Editorial ledger
The registry retains 100 dossiers. Noetix Dongguan is excluded from qualifying counts and exports because the available sources do not establish complete biped manufacturing at this site. Historical audit milestones remain dated records.
The Fremont Optimus record now distinguishes reported initial production from Tesla's future one-million-unit design target.
A country-by-country verification pass expanded the map while retaining source, confidence, and public-location requirements.
Cumulative complete general-purpose bipedal humanoids built worldwide, including research prototypes and internal fleets. G1/X-series-scale robots qualify; small dedicated educational/toy humanoids do not.
Opening balance · before 2020
≈598
Modeled average production
+123.4 / day
Reported sample minimum
14,706+
Current estimate = pre-2020 opening balance + production accumulated across dated historical and forecast periods. The reported sample minimum comes from non-overlapping manufacturer groups and dated production/delivery disclosures. It is a partial historical floor, not a live global census.
At review on 2026-09-05, the central estimate was 47,875: 14,706 in the reported sample floor and approximately 33,169 in modeled history, inventory, missing coverage and subsequent production.
Low and high are explicit assumption scenarios, not statistical confidence limits or guaranteed bounds. Production can fall outside this range. One shared UTC model integrates each dated production segment. Forecast rates vary by quarter and stop at the end of 2030; this is not live factory telemetry. Historical within-period timing is interpolated, not a record of individual factory events.
At review · 2026-09-05
Open any row for the arithmetic, assumptions and original sources. All company totals below are modeled.
H/G/R bipedal platforms; wheeled and quadruped products excluded.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: 2023/2024 use reported production. 2025 uses the >6,500 company floor with inventory uncertainty. H1 2026 converts >7,000 shipments to 7,000/8,200/11,200 built; no repeated addition of shipments.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 36.3 / 48.8 / 69.7.
Reported sample: 12,500+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Bipedal A/X products only; modeled exclusion of G-series and A2-W.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Mixed milestones are never summed. 2024 biped share is uncertain. Central June 2026 cumulative biped estimate uses the earlier 71.76% family-share proxy; A2-W makes this an assumption, not a measured biped share.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 22.1 / 40.2 / 57.8.
Reported sample: 1,846+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
General-purpose Walker/industrial bipeds; company totals require a form-factor adjustment.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Early Walker cohorts are editorial allocations, not measured totals. Later periods use shipment/sales proxies with biped-share and inventory uncertainty; toy and wheeled lines are excluded.
Opening scenarios: 1 / 20 / 80. H1 2026 baseline daily rate scenarios: 4.4 / 6.9 / 10.8.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
General-purpose Kuavo bipeds; wheeled and small educational products excluded.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Kuavo-scale platforms only. Dedicated small education products are excluded. Earlier zeros are scope allocations; later figures use shipment proxies plus modeled inventory.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 3.6 / 4.8 / 6.8.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Figure 01/02/03, including internal research fleets.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Prototype and older-generation cohorts are reconstructed. H1 2026 includes the >350 Figure 03 milestone and uncertain subsequent builds; a demonstrated hourly cycle does not establish 24/7 output.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 2.5 / 3.9 / 8.5.
Reported sample: 350+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
General-purpose PM01/SE01/T800 bipeds; stationary lower-body-free platforms excluded.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: General-purpose bipeds only. Shipment estimates are proxies; 2024 and H1 2026 allowances are explicitly modeled.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 1.1 / 3.6 / 8.9.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
General-purpose GR/N bipedal platforms; rehabilitation devices excluded.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: GR-series bipeds only; rehabilitation devices excluded. Early cohorts and inventory are weakly evidenced reconstruction allowances.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 0.8 / 2.4 / 5.7.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Digit and predecessor research bipeds; internal fleets included.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Only complete Digits enter this row; leg-only Cassie robots excluded. Ford's two deliveries are one observation. Remaining early production is an uncertain research/pilot allowance.
Opening scenarios: 1 / 8 / 30. H1 2026 baseline daily rate scenarios: 0.4 / 1.0 / 2.5.
Reported sample: 2+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Optimus bipedal prototypes and internal fleets.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Actual internal fleet totals are not public. These are wide editorial prototype/pilot scenarios, not a claim that management's targets were achieved or missed.
Opening scenarios: 0 / 0 / 0. H1 2026 baseline daily rate scenarios: 0.3 / 1.7 / 10.7.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Complete hydraulic and electric Atlas platforms.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Hydraulic and electric Atlas are distinct physical cohorts. Seven DARPA platforms form an old sample floor; later prototype and product cohorts are assumptions.
Opening scenarios: 7 / 20 / 60. H1 2026 baseline daily rate scenarios: 0.0 / 0.2 / 0.7.
Reported sample: 7+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
Honda, PAL, HUBO/HRP and other institutional bipeds; excludes all other rows.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Disjoint legacy/research allowance for Honda, PAL, HUBO, HRP and comparable institutional bipeds, excluding every named maker. Program evidence supports existence, not these quantities. All quantities are editorial scenarios.
Opening scenarios: 100 / 500 / 2,000. H1 2026 baseline daily rate scenarios: 0.1 / 0.5 / 1.7.
Reported sample: 1+. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
All other general-purpose biped makers, including 1X, Apptronik, Boston Dynamics, Noetix and newer research fleets; excludes every named row above.
Opening balance: Pre-2020 opening balance is an explicit research-fleet scenario. It is retained cumulatively; retirement does not undo production.
Historical reconstruction: Disjoint emerging-maker cohort: Apptronik, 1X bipeds, newer Chinese, European, Indian and other programs. Excludes the named makers and legacy research row. EVE and other wheeled products are excluded. Missing coverage is modeled, never an invented reported total.
Opening scenarios: 0 / 50 / 300. H1 2026 baseline daily rate scenarios: 3.3 / 9.6 / 25.4.
Reported sample: Not established. A disjoint reported sample at the latest historical cutoff; consult dated observations. Unknown does not mean zero.
These source observations retain their original measures. They are not an additive list.
Unitree · >6,500 · Calendar 2025
production · bipedal · Historical production floor. Not added to shipments of the same robots.
Observation date 2025-12-31 · reviewed with model 2026-09-05.
Unitree Robotics · published 2026-01-22 ↗Unitree · >5,500 · Calendar 2025
shipments · bipedal · Can be combined with H1 2026 deliveries because the reporting periods are disjoint; production and inventory stay separate.
Observation date 2025-12-31 · reviewed with model 2026-09-05.
Unitree Robotics · published 2026-01-22 ↗Unitree · >7,000 · H1 2026
shipments · bipedal · Delivery proxy and cumulative-build floor, not measured production during H1.
Observation date 2026-06-30 · reviewed with model 2026-09-05.
Counterpoint Research · published 2026-08-20 ↗AGIBOT · 1,846 · Cumulative X-series
cumulative production · bipedal · Directly identified biped family; included in the reported sample floor.
Observation date 2025-12-08 · reviewed with model 2026-09-05.
AGIBOT · published 2025-12-08 ↗AGIBOT · 10,000 · Cumulative mixed portfolio
cumulative production · mixed · Used only with the later milestone to derive a mixed-output interval.
Observation date 2026-03-30 · reviewed with model 2026-09-05.
AGIBOT · published 2026-03-30 ↗AGIBOT · 15,000 · Cumulative mixed portfolio
cumulative production · mixed · Apply a disclosed model-specific biped-share assumption; never count all 15,000 as bipeds.
Observation date 2026-06-28 · reviewed with model 2026-09-05.
AGIBOT · published 2026-06-28 ↗Figure · >350 · Cumulative Figure 03
cumulative production · bipedal · Direct completed-robot floor; extrapolation is modeled separately.
Observation date 2026-04-29 · reviewed with model 2026-09-05.
Figure · published 2026-04-29 ↗UBTECH · 1,079 · Calendar 2025
sales · mixed · Sales proxy under the company definition; unknown biped mix prevents a direct biped floor.
Observation date 2025-12-31 · reviewed with model 2026-09-05.
UBTECH · published 2026-07-13 ↗Leju Robotics · 650 · H1 2026
shipments · mixed · Model scope adjustment; not automatically treated as biped production.
Observation date 2026-06-30 · reviewed with model 2026-09-05.
Counterpoint Research · published 2026-08-20 ↗Unitree · 9 · Calendar 2023
production · bipedal · Exact annual production from the March 2026 filing. Sales are separate; this annual cohort is included once.
Observation date 2023-12-31 · reviewed with model 2026-09-05.
Unitree · prospectus (Chinese) · published 2026-03-20 ↗Unitree · 545 · Calendar 2024
production · bipedal · Exact annual production from the March 2026 filing. Sales are separate; this annual cohort is included once.
Observation date 2024-12-31 · reviewed with model 2026-09-05.
Unitree · prospectus (Chinese) · published 2026-03-20 ↗Unitree · 3,701 · January–September 2025
production · bipedal · Contained in the later full-year production disclosure; never add both periods.
Observation date 2025-09-30 · reviewed with model 2026-09-05.
Unitree · prospectus (Chinese) · published 2026-03-20 ↗Agility Robotics · 2 · First Ford delivery
shipments · bipedal · Ford's launch announcement and sustainability report describe the same two robots. Count once.
Observation date 2020-01-06 · reviewed with model 2026-09-05.
Ford · published 2020-01-06 ↗Boston Dynamics · 7 · DARPA Atlas platforms
shipments · bipedal · Already in the pre-2020 opening balance. Not an additional 2020 shipment.
Observation date 2013-07-11 · reviewed with model 2026-09-05.
DARPA · published 2013-07-11 ↗AGIBOT · 962 · Cumulative mixed portfolio
cumulative production · mixed · Mixed milestone used for the dated backtest. Included in later milestones, never summed.
Observation date 2024-12-15 · reviewed with model 2026-09-05.
China Mobile Robot Alliance · published 2024-12-22 ↗Unitree Robotics · 2026-01-22
2025 humanoid production exceeded 6,500; deliveries exceeded 5,500. Wheeled robots are explicitly excluded.
Both values are thresholds, not exact totals. Their ratio does not establish an inventory ratio.
AGIBOT · 2025-12-08
Cumulative output: A-series 1,742; X-series 1,846; G-series 1,412.
G-series is wheeled. A-series also has an A2-W variant; its share is not disclosed. Only X-series enters the direct biped floor.
AGIBOT · 2026-03-30
Cumulative mixed-portfolio production reached 10,000.
Includes wheeled robots; not a biped-only count.
AGIBOT · 2026-06-28
Cumulative mixed-portfolio production reached 15,000; the milestone robot was a G2.
The 5,000 additions over 90 days measure mixed output. The biped share after December is undisclosed.
Counterpoint Research · 2026-08-20
H1 shipments: Unitree >7,000; UBTECH around 1,000; Leju 650. Global total >22,000 includes wheeled models.
Shipments do not directly measure production. The page says August 19; its body dateline says August 20.
Figure · 2026-04-29
Over 350 Figure 03 robots delivered; a one-robot-per-hour cycle demonstrated.
A demonstrated cycle is not 24/7 realized output. Older Figure generations and subsequent output are not quantified here.
UBTECH · 2026-07-13
2025 full-sized humanoid sales totaled 1,079 under the company definition.
Sales, rather than production; the statement does not establish an exclusively bipedal product mix.
Omdia · 2026-01
Analyst estimates of manufacturer shipments in 2024 and 2025 provide historical scale for smaller manufacturers.
Mixed coverage and estimates, not audited builds. Its Unitree 2025 estimate is superseded by Unitree’s later disclosure. Hosted by AGIBOT; table visually checked.
Tesla / SEC · 2026-07-22
Describes the Optimus manufacturing ramp and planned line.
No current robot output total is established; design capacity is excluded from the count.
Interact Analysis · 2026-06-11
Global production exceeded 20,000 in 2025 under the publisher’s coverage.
Used as context only. It is neither added to manufacturer rows nor silently treated as biped-only.
Unitree · prospectus (Chinese) · 2026-03-20
The company reports 9 humanoids produced in 2023, 545 in 2024 and 3,701 in January–September 2025; sales were 5, 410 and 3,551 respectively.
Company filing reproduced by a financial-document host. This dated application is not the final prospectus; later full-year disclosure replaces the partial 2025 period. Production and revenue-recognized sales have different scopes.
Ford · 2020-01-06
Ford was the first customer for two commercial Digit robots in 2020.
A customer delivery minimum, not total Agility production. The old newsroom URL may redirect; Ford's 2020 sustainability report also documents the acquisition.
Ford · 2020
Ford describes acquiring two upright Digit robots for warehousing and delivery research.
Same two robots as the launch announcement; never add these records.
Honda (Japanese) · 2000-11-20
Documents a complete research humanoid program before the chart starts.
Establishes existence, not an all-time fleet count. Opening-balance quantities are editorial assumptions.
PAL Robotics · 2016-10-04
European full-size research humanoid program; first platform intended for LAAS-CNRS.
Announcement and intended delivery do not establish yearly production totals.
KAIST (Korean) · 2021-03-12
KAIST identifies the founders of Rainbow Robotics with its bipedal HUBO program.
Qualitative program evidence; no robot quantity is supplied. Quadrupeds and robot arms are excluded.
DARPA · 2013-07-11
Seven Atlas platforms were provided to qualifying DARPA Robotics Challenge teams.
A historical sample, not Boston Dynamics' total production. Subsequent reuse does not create additional robots.
China Mobile Robot Alliance · 2024-12-22
Reports 962 cumulative robots as of December 15, 2024, covering both bipedal and wheeled models.
Mixed-form total cannot enter the bipedal floor. The article's anticipated 200 biped deliveries are a target, not a completed delivery count.
Tesla · 2025-03-20
Original recording associated with Musk's 2025 and 2026 production ambitions.
Video source reference; numerical statements were cross-checked against contemporary reporting. Targets do not establish achieved output.
Not a Tesla App · 2025-03-21
Contemporary recap reports ambitions of 5,000 robots in 2025 and 50,000 in 2026.
Secondary reporting of management ambitions; actual annual output remains unverified.
Tesla · 2025-04-22
Original webcast for Musk's longer-term Optimus production discussion.
The linked contemporaneous report provides the transcribed numerical statement; this is not achieved output.
The Register · 2025-04-23
Reports Musk's expectation of one million Optimus units annually by 2030, potentially 2029, while discussing component constraints.
An annual-output ambition, not cumulative robots built; no precise ramp probability can be inferred.
Hyundai Motor Group · 2026-01-05
Targets a production system capable of 30,000 robot units annually by 2028 and phased Atlas deployment.
A future production-system target; the release's robot scope and timing do not prove 30,000 current Atlas units.
Figure · 2025-03-15
Describes a 12,000-unit annual production-line design and supply-chain capability for 100,000 robots over four years.
Capacity and supply-chain capability are separate from actual completed production or contracted demand.
Apptronik · 2023-08-23
Documents the Apollo program's public introduction in 2023.
No production total disclosed; represented within the disjoint emerging-maker scenario.
Interact Analysis · 2026-05-21
Analyst outlook distinguishes future demand from limited present commercial workforce adoption.
Market shipments and applications differ from all-time bipedal production. Used as a constraint/context check, not an added count.
Omdia · 2024-07-25
A historical forecast put 2030 annual humanoid shipments at 38,000.
A dated forecast with an earlier market definition; not a current estimate or a comparable bipedal-production outcome.
Associated Press · 2026-06
Reporting examines Chinese manufacturing scale, customer demand and differing global forecasts.
Not an additive production ledger. Quantities attributed to analysts retain their shipment definitions.
Tesla · 2025-01-29
Original webcast associated with the early 2025 Optimus internal plan.
Numerical statement cross-checked in contemporary reporting; an ambition, not a completed output disclosure.
Electrek · 2025-01-31
Reports a 10,000-unit internal 2025 plan, with Musk expecting several thousand rather than full achievement.
Contemporaneous reporting of a management target. Outcome is not verified by this article.
Version 2026-09-05-draft-2 · draft for review. Download the complete model JSON
Model revision · 2026-09-05 · draft
At the same review time, the estimate changed by +2,837 as historical coverage and production assumptions were revised.
Previous counter · at the review date
≈45,037
2026-09-05-draft-1 · constant 122.2/day.
Rebuilt model · at the same instant
≈47,875
Manufacturer ledger, bipedal scope, 123.4 modeled additions/day.
The difference is a change in assumptions and coverage, not robots disappearing. The earlier preview’s H1 shipment figure measured a different quantity. That evidence remains available in its original edition.
The weekly briefing
One source-linked update each week: production disclosures, factory developments and changes to the estimate.
Sample · September 5, 2026 · draft
The draft model now reconstructs production from 2020 and includes quarterly scenarios through 2030. Early research fleets remain incompletely documented.
Company ambitions are shown separately from the forecast. Capacity announcements establish potential, rather than completed robots.
Inspect the outlook and original sources ↗Get one concise, source-linked industry update each week.
Your email is used for the briefing. Privacy details.
