WEBVTT
00:00:00.000 --> 00:00:01.520
Aptronic Apollo vs.
00:00:01.679 --> 00:00:03.279
UbiTech Walker S2.
00:00:03.680 --> 00:00:07.360
Which 2026 Factory Humanoid offers the better business case?
00:00:07.599 --> 00:00:10.480
By Max Richter, Chief Futurist and Visionary.
00:00:10.640 --> 00:00:13.759
Evidence checked through October 28, 2026.
00:00:14.000 --> 00:00:21.519
The short answer: UbiTech's Walker S2 has the stronger business case for a factory that wants to run a humanoid pilot now.
00:00:21.679 --> 00:00:30.160
It has clearer published specifications, a battery swap system built for long shifts, and a much stronger public record of production and sales.
00:00:30.399 --> 00:00:36.560
Aptronix Apollo 2 is the more flexible platform, but it's still mainly a learning and pilot system.
00:00:36.719 --> 00:00:40.719
Its wield option may suit flat factory floors better than legs.
00:00:40.880 --> 00:00:54.399
Its ties to Google DeepMind and its work with Mercedes-Benz, GXO, and Jable are valuable, but buyers cannot yet judge Apollo 2 from a full public spec sheet, a list price, or disclosed uptime results.
00:00:54.560 --> 00:00:56.000
One date point matters.
00:00:56.240 --> 00:01:01.200
I found no verified second half 2026 hardware refresh for either robot.
00:01:01.439 --> 00:01:05.840
Aptronic unveiled Apollo 2 on June 30th, 2026.
00:01:06.159 --> 00:01:11.519
UbiTech's Walker S2 is its current industrial model, launched in 2025.
00:01:11.680 --> 00:01:20.480
The notable second half developments are about pilots, distribution, and production capacity, not a confirmed new Apollo or Walker S2 version.
00:01:20.640 --> 00:01:27.439
That makes this less a comparison of two new 2026 models and more a comparison of two different stages of commercialization.
00:01:27.680 --> 00:01:30.000
Aptronic is still proving and training.
00:01:30.239 --> 00:01:33.519
UbiTech is already reporting full-size robot sales.
00:01:33.760 --> 00:01:38.879
At a glance, for the full table, please open this article on robotcomparisons.com.
00:01:39.040 --> 00:01:41.200
The table shows the central trade-off.
00:01:41.439 --> 00:01:43.920
Apollo offers more choice in how the robot moves.
00:01:44.079 --> 00:01:49.120
Walker S2 offers more public evidence that the product is already being made and sold.
00:01:49.359 --> 00:01:53.840
The hardware, what buyers can and cannot compare dimensions and payload.
00:01:54.079 --> 00:01:56.640
UbiTech publishes more useful purchase data.
00:01:56.959 --> 00:02:03.040
Walker S2 stands 176 centimeters tall and weighs 70 kilograms without its dexterous hands.
00:02:03.280 --> 00:02:10.080
Its manual lists a 177 centimeter arm span and a maximum combined arm load of 15 kilograms.
00:02:10.319 --> 00:02:17.199
UbiTech says it can handle that load across a 0 to 1.8 meter workspace, which includes floor level and higher reaches.
00:02:17.439 --> 00:02:20.159
Apollo 2's public information is much thinner.
00:02:20.400 --> 00:02:28.400
Aptronic describes its current robot as a human-scale platform with dexterous object handling, perception systems, and modular mobility.
00:02:28.560 --> 00:02:36.719
But it does not publish Apollo 2's height, weight, rated payload, walking speed, or full degree of freedom count on its current product page.
00:02:36.960 --> 00:02:50.080
The familiar Apollo figures, about 1.7 meters tall, 72.6 kilograms, and a 25 kilogram lift, belong to the version announced in 2023, not a verified Apollo 2 specification.
00:02:50.240 --> 00:02:57.360
They should not be used in a 2026 purchase comparison unless Apptronic confirms them for the exact robot on offer.
00:02:57.520 --> 00:02:58.879
That gap matters.
00:02:59.120 --> 00:03:04.400
A buyer cannot size a robot for a task from human scale and dexterous alone.
00:03:04.639 --> 00:03:12.319
Ask for a payload versus reach chart, cycle limits, and hand configuration for the exact Apollo 2 unit being proposed.
00:03:12.639 --> 00:03:22.719
Dexterity and useful work Walker S2 has 52 degrees of freedom across the robot's body, according to UBTech's 2025 annual report.
00:03:22.960 --> 00:03:27.439
It also has binocular vision and force sensors in its published documentation.
00:03:27.759 --> 00:03:35.520
UBTech's product materials describe work such as sorting parts, handling totes, and reaching from floor level to higher shelves.
00:03:35.680 --> 00:03:45.680
Those are more useful clues than a bare joint count, but they still do not replace task-level results such as successful picks per hour or the number of human interventions.
00:03:46.000 --> 00:03:53.120
Aptronic says Apollo 2 can handle objects with dexterity, but does not publish comparable hand specifications.
00:03:53.360 --> 00:03:56.159
That makes the real dexterity race unresolved.
00:03:56.319 --> 00:04:03.520
A factory should compare the actual hand and tool setup on its own parts, not assume that a more human-like hand is automatically better.
00:04:03.759 --> 00:04:08.240
A simple gripper can be faster and easier to maintain for a fixed task.
00:04:08.479 --> 00:04:09.439
Locomotion.
00:04:09.680 --> 00:04:10.879
Wheels versus legs.
00:04:11.120 --> 00:04:15.039
Apollo 2 can be configured with wheels or bipedal legs.
00:04:15.199 --> 00:04:17.439
That is a meaningful practical advantage.
00:04:17.600 --> 00:04:23.120
On a smooth factory floor, wheels are usually the sensible choice for moving between stations.
00:04:23.279 --> 00:04:28.480
They avoid the balanced demands of walking and fit better with established mobile robot practices.
00:04:28.720 --> 00:04:35.439
Legs make more sense when the robot must handle steps, awkward floor layouts, or places that a wheeled base cannot reach.
00:04:35.680 --> 00:04:44.000
Aptronic presents the wheeled version as suited to current industrial mobile robot settings, and the bipedal version as its more adaptable form.
00:04:44.319 --> 00:04:46.639
Walker S2 is a biped.
00:04:46.800 --> 00:04:49.439
That keeps its work range close to a person's.
00:04:49.600 --> 00:04:59.600
It can reach low, bend, and handle loads over a wide vertical space, but a two-legged robot also brings more moving parts and balance demands than a wheeled platform.
00:04:59.759 --> 00:05:06.639
If the job is only to carry items along a flat aisle, the legs may be solving a problem the factory does not have.
00:05:06.879 --> 00:05:12.480
Neither company provides enough comparable public data to rank the two on real factory speed.
00:05:12.720 --> 00:05:15.680
Top speed is not the same as task throughput.
00:05:15.839 --> 00:05:17.920
What matters is the complete cycle.
00:05:18.160 --> 00:05:23.839
Travel, pick, place, wait, recover from errors, and hand the next task to the robot.
00:05:24.160 --> 00:05:30.800
Runtime, battery swapping, and uptime, here, Walker S2 has the clearer design for long shifts.
00:05:31.040 --> 00:05:40.639
Ubitech says the robot can change its battery autonomously in about three minutes and claims 24-hour operation using its dual battery system and swap station.
00:05:40.879 --> 00:05:46.800
A reproduced Walker S2 user manual lists a 2.5-hour comprehensive range.
00:05:46.959 --> 00:05:49.120
Those figures describe different things.
00:05:49.360 --> 00:05:57.519
The manuals figure is an operating range for a charge, while the company's 24-hour claim depends on battery swapping and continued operation.
00:05:57.759 --> 00:06:03.199
Buyers should confirm actual runtime under their own walking, lifting, and idle time pattern.
00:06:03.439 --> 00:06:13.439
Aptronic says Apollo 2 uses swappable batteries to support 7 days by 22 hours operation, with opportunity charging and tethering also available.
00:06:13.600 --> 00:06:17.439
It does not publish a current Apollo 2 battery runtime per pack.
00:06:17.600 --> 00:06:22.000
The older Apollo's 4-hour figure is not a confirmed Apollo 2 specification.
00:06:22.240 --> 00:06:24.959
A battery swapping is not the same as uptime.
00:06:25.120 --> 00:06:33.279
A 3-minute SWAT is useful only if the robot can complete its task reliably, find the station, swap without help, and return to work.
00:06:33.519 --> 00:06:40.480
Neither company publishes enough independent, factory-level uptime data to show that it does this at production rates.
00:06:40.720 --> 00:06:49.839
Executives should ask both suppliers for logs of robot availability, completed cycles, remote interventions, unplanned stops, and time to repair.
00:06:50.160 --> 00:06:56.879
AI, Fleet Control, Safety, and Maintenance Apptronic divides the Apollo platform into three parts.
00:06:57.120 --> 00:07:02.800
Artemis coordinates perception, planning, control, safety, and task execution.
00:07:03.040 --> 00:07:08.879
Fleet Connect is designed to monitor robot status, assign work, and manage a group of robots.
00:07:09.199 --> 00:07:14.959
Apollo 2 also gathers training data through a mix of teleoperation and autonomous execution.
00:07:15.199 --> 00:07:20.639
Aptronic says that data supports its work with Google DeepMind on Gemini robotics models.
00:07:20.800 --> 00:07:22.800
That is a promising learning loop.
00:07:23.040 --> 00:07:26.879
But it does not mean every customer task is already autonomous.
00:07:27.199 --> 00:07:31.759
UbiTech software is aimed more directly at multi-robot factory work.
00:07:31.920 --> 00:07:33.199
It's BrainNet 2.
00:07:33.360 --> 00:07:40.879
Zero and Coagent Systems are designed to break tasks into parts, schedule work, share information, and coordinate several robots.
00:07:41.120 --> 00:07:45.759
UpTech also lists ROSA 2.0 and a robot management cloud platform.
00:07:46.160 --> 00:07:47.279
The idea is strong.
00:07:47.439 --> 00:07:51.680
A fleet should work like a connected team, not a row of isolated machines.
00:07:51.920 --> 00:07:59.920
But public descriptions do not establish how often operators must rescue robots or how well the system handles unexpected line changes.
00:08:00.079 --> 00:08:03.519
On safety, Apptronic describes hardware safety zones.
00:08:03.680 --> 00:08:13.360
Its product page says Apollo can pause when an object enters a defined impact zone and can adjust behavior around people or obstacles within a configurable perimeter.
00:08:13.680 --> 00:08:24.959
Walker S2's manual gives operating conditions, including trained operator requirements, an emergency stop, and warnings about slippery floors and certain hazardous or high interference areas.
00:08:25.120 --> 00:08:32.960
These are useful design and operating details, not proof of a specific safety certification or a record of safe operation at production scale.
00:08:33.200 --> 00:08:37.039
Buyers still need a site-specific risk assessment and acceptance test.
00:08:37.279 --> 00:08:49.279
Oh, for maintenance, Apptronics says its actuators are designed for high efficiency and simpler upkeep, and its partnership with Jable covers both Apollo manufacturing and trials in Jable's factories.
00:08:49.600 --> 00:08:58.960
Walker S2 has a three-minute battery swap feature, but its manual says operators need official training and restricts unauthorized changes to core settings.
00:08:59.200 --> 00:09:00.559
These are different strengths.
00:09:00.799 --> 00:09:04.559
Apptronic emphasizes modular design and a manufacturing partner.
00:09:04.799 --> 00:09:10.399
YouTech emphasizes a service and operating process built around its industrial robot platform.
00:09:10.559 --> 00:09:17.360
Neither company publishes the repair time and failure rate data needed to settle which will cost less to maintain.
00:09:17.679 --> 00:09:25.519
What the factory evidence actually says, automotive and electronics, UPTEC has a wider public footprint in industrial settings.
00:09:25.759 --> 00:09:31.120
Its customer list includes automotive, electronics, logistics, and semiconductor companies.
00:09:31.279 --> 00:09:37.360
Its published examples include work such as handling, assembly support, inspection, and material movement.
00:09:37.519 --> 00:09:39.679
But the exact robot version matters.
00:09:39.919 --> 00:09:44.960
Some named projects used earlier Walker models, not necessarily Walker S2.
00:09:45.360 --> 00:09:50.799
The company's broader customer list should not be mistaken for a list of Walker S2 installations.
00:09:50.960 --> 00:10:00.480
For Walker S2 specifically, UbiTech reports industrial manufacturing tasks that include tote handling, sorting, and quality checks.
00:10:00.720 --> 00:10:09.679
It has also announced a Walker S2 validation project with Hitachi in China, including work at a Hitachi elevator manufacturing site.
00:10:09.919 --> 00:10:19.200
This is valuable factory evidence, but it remains a validation effort, not public proof of a large, independently measured production rollout.
00:10:19.519 --> 00:10:26.399
Apptronic has pilots or partner work involving Mercedes-Benz, GXO, and Jabil.
00:10:26.720 --> 00:10:30.559
Mercedes Benz has publicly said it is testing Apollo in production.
00:10:30.720 --> 00:10:40.320
At Jabil, the announced pilot includes tasks such as inspection, sorting, kiting, line side delivery, fixture placement, and subassembly.
00:10:40.639 --> 00:10:46.000
Apptronic says Apollo 2 robots also collect data at customer and partner sites.
00:10:46.240 --> 00:10:54.399
Yet neither Apptronic nor these partners has published comparable throughput, uptime, intervention, or labor savings figures.
00:10:54.879 --> 00:11:01.360
So, the factory evidence supports a real distinction, but not a simplistic claim that one robot has one.
00:11:01.919 --> 00:11:05.039
UBTech has more public commercial scale evidence.
00:11:05.279 --> 00:11:09.840
Apptronic has credible partner trials and a substantial data collection effort.
00:11:10.080 --> 00:11:17.679
Neither has publicly shown enough operational metrics for a plant manager to copy a mature deployment plan with confidence.
00:11:18.000 --> 00:11:20.480
Orders, production, and availability.
00:11:20.720 --> 00:11:26.559
UBTech began Walker S2 mass production and delivery in November 2025.
00:11:26.879 --> 00:11:33.279
The company reported more than RMB$800 million in Walker Series orders at that point.
00:11:33.440 --> 00:11:41.759
Its 2025 annual report describes a roughly 1,000 unit level of small-scale Walker S2 production and delivery.
00:11:42.000 --> 00:11:54.720
In September 2026, a new Liazo factory began operation with a planned annual capacity of 10,000 units and a designed line pace of one robot every 10 minutes.
00:11:55.039 --> 00:11:59.279
That capacity covers Walker S and Cruiser Series production.
00:11:59.440 --> 00:12:04.080
It is not proof that 10,000 Walker S2 units have been built or sold.
00:12:04.320 --> 00:12:17.840
UBTech's H1 2026 Interim Results Report 921 full-size humanoid sales and RMB 590.3 million in revenue from full-size humanoid products and services.
00:12:18.080 --> 00:12:22.799
Those are company-wide category figures, not Walker S2 only sales.
00:12:23.120 --> 00:12:25.840
Apptronic's reported figures are less complete.
00:12:26.159 --> 00:12:35.039
Reuters reported in June 2026 that the company had built hundreds of Apollo 2 robots, but Aptronic did not disclose deployment numbers.
00:12:35.279 --> 00:12:43.120
Its CEO said pilots would continue through 2026, with production versions expected in 2027 and beyond.
00:12:43.440 --> 00:12:50.639
For regional access, Aptronic is based in Texas and has customer and partner work in the United States and Germany.
00:12:50.879 --> 00:12:55.600
But Apollo 2 is not presented as a generally available production robot today.
00:12:56.080 --> 00:13:02.879
UbiTech has announced a European distribution partner, Terra Robotics, for Germany, Switzerland, and Austria.
00:13:03.120 --> 00:13:07.360
It has also announced a Walker S2 deal with Airbus and a partnership with U.S.
00:13:07.440 --> 00:13:09.679
semiconductor maker Texas Instruments.
00:13:10.000 --> 00:13:14.240
Those overseas deals matter, but they do not establish a broad U.S.
00:13:14.399 --> 00:13:17.440
sales, service, and parts network for every buyer.
00:13:17.759 --> 00:13:18.240
A U.S.
00:13:18.320 --> 00:13:24.320
factory should confirm supply, service response, software support, and data handling terms directly.
00:13:24.799 --> 00:13:27.919
Price, integration cost, and the business case.
00:13:28.159 --> 00:13:32.000
Neither manufacturer publishes a standard price for these robots.
00:13:32.320 --> 00:13:36.080
UBTech's H1 2026 numbers provide a rough reference.
00:13:36.399 --> 00:13:48.080
RMB 590.3 million in full-size humanoid product and service revenue divided by 921 units is about RMB$641,000 per unit equivalent.
00:13:48.320 --> 00:13:52.639
That is roughly$95,000 using a simple currency conversion.
00:13:52.799 --> 00:13:55.039
It is not a Walker S2 price.
00:13:55.279 --> 00:14:01.039
The category includes more than one product, and the revenue includes services as well as hardware.
00:14:01.200 --> 00:14:04.159
Treat it as a budget clue, not a quote.
00:14:04.480 --> 00:14:09.120
Apptronic has not published an Apollo 2 price or a customer rate card.
00:14:09.279 --> 00:14:20.399
Since Apptronic says production versions are expected in 2027 and beyond, a business case for buying Apollo 2 today cannot be completed from public pricing data.
00:14:20.559 --> 00:14:27.519
Buyers need a written pilot quote that breaks out hardware, software, integration, support, and any recurring fees.
00:14:27.759 --> 00:14:31.600
Neither company has published standard factory integration costs.
00:14:31.840 --> 00:14:42.799
Those costs can include workflow changes, safety work, fixtures, battery or charging equipment, software links to factory systems, operator training, and ongoing support.
00:14:43.039 --> 00:14:46.480
The right number depends heavily on the task and site.
00:14:46.720 --> 00:14:50.559
An illustrative payback test to show why shift coverage matters.
00:14:50.720 --> 00:14:53.440
Consider a planning example, not a vendor quote.
00:14:53.679 --> 00:15:02.000
Robot purchase price proxy,$95,000, based on UBTech's blended H1 revenue per unit calculation above.
00:15:02.240 --> 00:15:06.639
One-time integration and site setup,$50,000 assumed.
00:15:06.879 --> 00:15:10.960
Annual service and consumables,$15,000 assumed.
00:15:11.200 --> 00:15:14.559
Loaded labor cost,$30 per hour assumed.
00:15:14.720 --> 00:15:19.840
Robot completes useful work during 75% of scheduled hours assumed.
00:15:20.000 --> 00:15:24.320
Each useful robot hour fully replaces one paid labor hour.
00:15:24.559 --> 00:15:32.879
Under those assumptions, one 2,000-hour shift schedule produces$1,500 useful robot hours a year.
00:15:33.120 --> 00:15:36.639
That is$45,000 of gross labor value.
00:15:36.879 --> 00:15:47.200
After the assumed$15,000 annual service cost, simple payback on the$145,000 installed cost is about 4.8 years.
00:15:47.440 --> 00:15:51.679
At two shifts, the model produces 3,000 useful hours a year.
00:15:51.840 --> 00:15:59.120
That is$90,000 of gross labor value and$75,000 after the assumed service cost.
00:15:59.360 --> 00:16:02.960
Simple payback falls to about 1.9 years.
00:16:03.120 --> 00:16:05.200
This is deliberately a simple test.
00:16:05.360 --> 00:16:13.039
It excludes financing, software fees, taxes, and downtime beyond the assumed 25% of scheduled hours.
00:16:13.279 --> 00:16:16.960
Most important, it assumes the robot actually replaces labor.
00:16:17.120 --> 00:16:22.080
If workers still need to supervise it or handle frequent exceptions, the real savings fall.
00:16:22.320 --> 00:16:23.600
The lesson is plain.
00:16:23.840 --> 00:16:28.159
A humanoid that runs one easy shift may be an expensive experiment.
00:16:28.399 --> 00:16:34.879
A robot that reliably covers multiple shifts on a hard-to-automate task has a much better chance of paying back.
00:16:35.039 --> 00:16:37.120
Which robot should a factory choose?
00:16:37.360 --> 00:16:39.840
Choose Walker S2 for a pilot now.
00:16:40.080 --> 00:16:51.440
Walker S2 is the stronger choice if the factory wants a robot with published payload and reach figures, an autonomous battery swap system, and a company already reporting full-size robot sales.
00:16:51.679 --> 00:17:00.159
Its lead is clearest for buyers in China and for European buyers with access to UBTKS, announced DAHC distribution partner.
00:17:00.559 --> 00:17:08.880
Its main weakness is that public sales, order, and capacity figures do not prove Walker S2 only output or production grade uptime.
00:17:09.039 --> 00:17:11.920
Choose Apollo 2 for a US learning pilot.
00:17:12.160 --> 00:17:21.279
Apollo 2 is the better fit if the goal is to test a flexible platform, compare wheels with legs, or build experience with Aptronics AI and fleet system.
00:17:21.680 --> 00:17:36.000
Its ties to Google DeepMind, JABL, Mercedes Benz, and GXO give the project real industrial context, but the robot's missing public specifications and price make it a poor candidate for a purchase decision based on a spreadsheet alone.
00:17:36.240 --> 00:17:39.839
Do not choose either for a task that does not need a humanoid.
00:17:40.000 --> 00:17:46.960
If an AMR, conveyor, fixed robot arm, or simple fixture can do the work more cheaply and reliably, start there.
00:17:47.200 --> 00:17:53.200
Humanoids make the most sense when a task combines movement through a human built space with changing hands-on work.
00:17:53.440 --> 00:17:58.240
For example, moving mixed parts between stations or handling items at different heights.
00:17:58.480 --> 00:18:15.759
Before signing a pilot, require both vendors to report the same measures, completed cycles per hour, successful task rate, human interventions per shift, time lost defaults, battery and charging downtime, repair time, and measured labor hours actually removed or redeployed.
00:18:15.920 --> 00:18:16.720
Conclusion.
00:18:16.960 --> 00:18:23.920
Walker S2 offers the better 2026 business case for an executive who wants to test a factory humanoid now.
00:18:24.160 --> 00:18:28.000
Apollo 2 offers the more flexible bet on what comes next.
00:18:28.240 --> 00:18:33.200
That is not the same as saying Walker S2 is ready to replace a production workforce.
00:18:33.359 --> 00:18:34.079
It is not.
00:18:34.400 --> 00:18:46.079
UBTK's sales, customer announcements, and new factory make its commercial position stronger, but its public figures still leave key questions about S2 specific deliveries and dependable uptime.
00:18:46.319 --> 00:18:53.119
Uptronic's platform looks promising, especially where wheels, AI learning, and US-based partner work matter.
00:18:53.279 --> 00:19:00.000
But without current Apollo 2 specifications, price, and measured performance, it remains harder to underwrite.
00:19:00.160 --> 00:19:03.279
The winning robot will not be the one with the boldest demo.
00:19:03.440 --> 00:19:09.200
It will be the one that finishes the job, shift after shift, at a cost the factory can prove.
00:19:09.440 --> 00:19:13.279
All links to sources are available in the text version of this article.
00:19:13.440 --> 00:19:18.079
You can find the full article at robotcomparisons.com/slash blog.
00:19:18.240 --> 00:19:19.359
Thanks for listening.
00:19:19.599 --> 00:19:21.839
Thanks for listening to Robot Comparisons.
00:19:22.000 --> 00:19:26.960
If you found this episode valuable, make sure to subscribe so you never miss a new deep dive.
00:19:27.119 --> 00:19:34.559
For the full written breakdown, detailed spec sheets, and side by side comparisons, head over to robotcomparisons.com.
00:19:34.720 --> 00:19:40.319
That's R O B O T C O M P A R I S O N S dot com.
00:19:40.559 --> 00:19:42.880
New research drops multiple times a week.
00:19:43.039 --> 00:19:44.079
See you in the next one.
00:00:00.000 --> 00:00:01.520
Aptronic Apollo vs.
00:00:01.679 --> 00:00:03.279
UbiTech Walker S2.
00:00:03.680 --> 00:00:07.360
Which 2026 Factory Humanoid offers the better business case?
00:00:07.599 --> 00:00:10.480
By Max Richter, Chief Futurist and Visionary.
00:00:10.640 --> 00:00:13.759
Evidence checked through October 28, 2026.
00:00:14.000 --> 00:00:21.519
The short answer: UbiTech's Walker S2 has the stronger business case for a factory that wants to run a humanoid pilot now.
00:00:21.679 --> 00:00:30.160
It has clearer published specifications, a battery swap system built for long shifts, and a much stronger public record of production and sales.
00:00:30.399 --> 00:00:36.560
Aptronix Apollo 2 is the more flexible platform, but it's still mainly a learning and pilot system.
00:00:36.719 --> 00:00:40.719
Its wield option may suit flat factory floors better than legs.
00:00:40.880 --> 00:00:54.399
Its ties to Google DeepMind and its work with Mercedes-Benz, GXO, and Jable are valuable, but buyers cannot yet judge Apollo 2 from a full public spec sheet, a list price, or disclosed uptime results.
00:00:54.560 --> 00:00:56.000
One date point matters.
00:00:56.240 --> 00:01:01.200
I found no verified second half 2026 hardware refresh for either robot.
00:01:01.439 --> 00:01:05.840
Aptronic unveiled Apollo 2 on June 30th, 2026.
00:01:06.159 --> 00:01:11.519
UbiTech's Walker S2 is its current industrial model, launched in 2025.
00:01:11.680 --> 00:01:20.480
The notable second half developments are about pilots, distribution, and production capacity, not a confirmed new Apollo or Walker S2 version.
00:01:20.640 --> 00:01:27.439
That makes this less a comparison of two new 2026 models and more a comparison of two different stages of commercialization.
00:01:27.680 --> 00:01:30.000
Aptronic is still proving and training.
00:01:30.239 --> 00:01:33.519
UbiTech is already reporting full-size robot sales.
00:01:33.760 --> 00:01:38.879
At a glance, for the full table, please open this article on robotcomparisons.com.
00:01:39.040 --> 00:01:41.200
The table shows the central trade-off.
00:01:41.439 --> 00:01:43.920
Apollo offers more choice in how the robot moves.
00:01:44.079 --> 00:01:49.120
Walker S2 offers more public evidence that the product is already being made and sold.
00:01:49.359 --> 00:01:53.840
The hardware, what buyers can and cannot compare dimensions and payload.
00:01:54.079 --> 00:01:56.640
UbiTech publishes more useful purchase data.
00:01:56.959 --> 00:02:03.040
Walker S2 stands 176 centimeters tall and weighs 70 kilograms without its dexterous hands.
00:02:03.280 --> 00:02:10.080
Its manual lists a 177 centimeter arm span and a maximum combined arm load of 15 kilograms.
00:02:10.319 --> 00:02:17.199
UbiTech says it can handle that load across a 0 to 1.8 meter workspace, which includes floor level and higher reaches.
00:02:17.439 --> 00:02:20.159
Apollo 2's public information is much thinner.
00:02:20.400 --> 00:02:28.400
Aptronic describes its current robot as a human-scale platform with dexterous object handling, perception systems, and modular mobility.
00:02:28.560 --> 00:02:36.719
But it does not publish Apollo 2's height, weight, rated payload, walking speed, or full degree of freedom count on its current product page.
00:02:36.960 --> 00:02:50.080
The familiar Apollo figures, about 1.7 meters tall, 72.6 kilograms, and a 25 kilogram lift, belong to the version announced in 2023, not a verified Apollo 2 specification.
00:02:50.240 --> 00:02:57.360
They should not be used in a 2026 purchase comparison unless Apptronic confirms them for the exact robot on offer.
00:02:57.520 --> 00:02:58.879
That gap matters.
00:02:59.120 --> 00:03:04.400
A buyer cannot size a robot for a task from human scale and dexterous alone.
00:03:04.639 --> 00:03:12.319
Ask for a payload versus reach chart, cycle limits, and hand configuration for the exact Apollo 2 unit being proposed.
00:03:12.639 --> 00:03:22.719
Dexterity and useful work Walker S2 has 52 degrees of freedom across the robot's body, according to UBTech's 2025 annual report.
00:03:22.960 --> 00:03:27.439
It also has binocular vision and force sensors in its published documentation.
00:03:27.759 --> 00:03:35.520
UBTech's product materials describe work such as sorting parts, handling totes, and reaching from floor level to higher shelves.
00:03:35.680 --> 00:03:45.680
Those are more useful clues than a bare joint count, but they still do not replace task-level results such as successful picks per hour or the number of human interventions.
00:03:46.000 --> 00:03:53.120
Aptronic says Apollo 2 can handle objects with dexterity, but does not publish comparable hand specifications.
00:03:53.360 --> 00:03:56.159
That makes the real dexterity race unresolved.
00:03:56.319 --> 00:04:03.520
A factory should compare the actual hand and tool setup on its own parts, not assume that a more human-like hand is automatically better.
00:04:03.759 --> 00:04:08.240
A simple gripper can be faster and easier to maintain for a fixed task.
00:04:08.479 --> 00:04:09.439
Locomotion.
00:04:09.680 --> 00:04:10.879
Wheels versus legs.
00:04:11.120 --> 00:04:15.039
Apollo 2 can be configured with wheels or bipedal legs.
00:04:15.199 --> 00:04:17.439
That is a meaningful practical advantage.
00:04:17.600 --> 00:04:23.120
On a smooth factory floor, wheels are usually the sensible choice for moving between stations.
00:04:23.279 --> 00:04:28.480
They avoid the balanced demands of walking and fit better with established mobile robot practices.
00:04:28.720 --> 00:04:35.439
Legs make more sense when the robot must handle steps, awkward floor layouts, or places that a wheeled base cannot reach.
00:04:35.680 --> 00:04:44.000
Aptronic presents the wheeled version as suited to current industrial mobile robot settings, and the bipedal version as its more adaptable form.
00:04:44.319 --> 00:04:46.639
Walker S2 is a biped.
00:04:46.800 --> 00:04:49.439
That keeps its work range close to a person's.
00:04:49.600 --> 00:04:59.600
It can reach low, bend, and handle loads over a wide vertical space, but a two-legged robot also brings more moving parts and balance demands than a wheeled platform.
00:04:59.759 --> 00:05:06.639
If the job is only to carry items along a flat aisle, the legs may be solving a problem the factory does not have.
00:05:06.879 --> 00:05:12.480
Neither company provides enough comparable public data to rank the two on real factory speed.
00:05:12.720 --> 00:05:15.680
Top speed is not the same as task throughput.
00:05:15.839 --> 00:05:17.920
What matters is the complete cycle.
00:05:18.160 --> 00:05:23.839
Travel, pick, place, wait, recover from errors, and hand the next task to the robot.
00:05:24.160 --> 00:05:30.800
Runtime, battery swapping, and uptime, here, Walker S2 has the clearer design for long shifts.
00:05:31.040 --> 00:05:40.639
Ubitech says the robot can change its battery autonomously in about three minutes and claims 24-hour operation using its dual battery system and swap station.
00:05:40.879 --> 00:05:46.800
A reproduced Walker S2 user manual lists a 2.5-hour comprehensive range.
00:05:46.959 --> 00:05:49.120
Those figures describe different things.
00:05:49.360 --> 00:05:57.519
The manuals figure is an operating range for a charge, while the company's 24-hour claim depends on battery swapping and continued operation.
00:05:57.759 --> 00:06:03.199
Buyers should confirm actual runtime under their own walking, lifting, and idle time pattern.
00:06:03.439 --> 00:06:13.439
Aptronic says Apollo 2 uses swappable batteries to support 7 days by 22 hours operation, with opportunity charging and tethering also available.
00:06:13.600 --> 00:06:17.439
It does not publish a current Apollo 2 battery runtime per pack.
00:06:17.600 --> 00:06:22.000
The older Apollo's 4-hour figure is not a confirmed Apollo 2 specification.
00:06:22.240 --> 00:06:24.959
A battery swapping is not the same as uptime.
00:06:25.120 --> 00:06:33.279
A 3-minute SWAT is useful only if the robot can complete its task reliably, find the station, swap without help, and return to work.
00:06:33.519 --> 00:06:40.480
Neither company publishes enough independent, factory-level uptime data to show that it does this at production rates.
00:06:40.720 --> 00:06:49.839
Executives should ask both suppliers for logs of robot availability, completed cycles, remote interventions, unplanned stops, and time to repair.
00:06:50.160 --> 00:06:56.879
AI, Fleet Control, Safety, and Maintenance Apptronic divides the Apollo platform into three parts.
00:06:57.120 --> 00:07:02.800
Artemis coordinates perception, planning, control, safety, and task execution.
00:07:03.040 --> 00:07:08.879
Fleet Connect is designed to monitor robot status, assign work, and manage a group of robots.
00:07:09.199 --> 00:07:14.959
Apollo 2 also gathers training data through a mix of teleoperation and autonomous execution.
00:07:15.199 --> 00:07:20.639
Aptronic says that data supports its work with Google DeepMind on Gemini robotics models.
00:07:20.800 --> 00:07:22.800
That is a promising learning loop.
00:07:23.040 --> 00:07:26.879
But it does not mean every customer task is already autonomous.
00:07:27.199 --> 00:07:31.759
UbiTech software is aimed more directly at multi-robot factory work.
00:07:31.920 --> 00:07:33.199
It's BrainNet 2.
00:07:33.360 --> 00:07:40.879
Zero and Coagent Systems are designed to break tasks into parts, schedule work, share information, and coordinate several robots.
00:07:41.120 --> 00:07:45.759
UpTech also lists ROSA 2.0 and a robot management cloud platform.
00:07:46.160 --> 00:07:47.279
The idea is strong.
00:07:47.439 --> 00:07:51.680
A fleet should work like a connected team, not a row of isolated machines.
00:07:51.920 --> 00:07:59.920
But public descriptions do not establish how often operators must rescue robots or how well the system handles unexpected line changes.
00:08:00.079 --> 00:08:03.519
On safety, Apptronic describes hardware safety zones.
00:08:03.680 --> 00:08:13.360
Its product page says Apollo can pause when an object enters a defined impact zone and can adjust behavior around people or obstacles within a configurable perimeter.
00:08:13.680 --> 00:08:24.959
Walker S2's manual gives operating conditions, including trained operator requirements, an emergency stop, and warnings about slippery floors and certain hazardous or high interference areas.
00:08:25.120 --> 00:08:32.960
These are useful design and operating details, not proof of a specific safety certification or a record of safe operation at production scale.
00:08:33.200 --> 00:08:37.039
Buyers still need a site-specific risk assessment and acceptance test.
00:08:37.279 --> 00:08:49.279
Oh, for maintenance, Apptronics says its actuators are designed for high efficiency and simpler upkeep, and its partnership with Jable covers both Apollo manufacturing and trials in Jable's factories.
00:08:49.600 --> 00:08:58.960
Walker S2 has a three-minute battery swap feature, but its manual says operators need official training and restricts unauthorized changes to core settings.
00:08:59.200 --> 00:09:00.559
These are different strengths.
00:09:00.799 --> 00:09:04.559
Apptronic emphasizes modular design and a manufacturing partner.
00:09:04.799 --> 00:09:10.399
YouTech emphasizes a service and operating process built around its industrial robot platform.
00:09:10.559 --> 00:09:17.360
Neither company publishes the repair time and failure rate data needed to settle which will cost less to maintain.
00:09:17.679 --> 00:09:25.519
What the factory evidence actually says, automotive and electronics, UPTEC has a wider public footprint in industrial settings.
00:09:25.759 --> 00:09:31.120
Its customer list includes automotive, electronics, logistics, and semiconductor companies.
00:09:31.279 --> 00:09:37.360
Its published examples include work such as handling, assembly support, inspection, and material movement.
00:09:37.519 --> 00:09:39.679
But the exact robot version matters.
00:09:39.919 --> 00:09:44.960
Some named projects used earlier Walker models, not necessarily Walker S2.
00:09:45.360 --> 00:09:50.799
The company's broader customer list should not be mistaken for a list of Walker S2 installations.
00:09:50.960 --> 00:10:00.480
For Walker S2 specifically, UbiTech reports industrial manufacturing tasks that include tote handling, sorting, and quality checks.
00:10:00.720 --> 00:10:09.679
It has also announced a Walker S2 validation project with Hitachi in China, including work at a Hitachi elevator manufacturing site.
00:10:09.919 --> 00:10:19.200
This is valuable factory evidence, but it remains a validation effort, not public proof of a large, independently measured production rollout.
00:10:19.519 --> 00:10:26.399
Apptronic has pilots or partner work involving Mercedes-Benz, GXO, and Jabil.
00:10:26.720 --> 00:10:30.559
Mercedes Benz has publicly said it is testing Apollo in production.
00:10:30.720 --> 00:10:40.320
At Jabil, the announced pilot includes tasks such as inspection, sorting, kiting, line side delivery, fixture placement, and subassembly.
00:10:40.639 --> 00:10:46.000
Apptronic says Apollo 2 robots also collect data at customer and partner sites.
00:10:46.240 --> 00:10:54.399
Yet neither Apptronic nor these partners has published comparable throughput, uptime, intervention, or labor savings figures.
00:10:54.879 --> 00:11:01.360
So, the factory evidence supports a real distinction, but not a simplistic claim that one robot has one.
00:11:01.919 --> 00:11:05.039
UBTech has more public commercial scale evidence.
00:11:05.279 --> 00:11:09.840
Apptronic has credible partner trials and a substantial data collection effort.
00:11:10.080 --> 00:11:17.679
Neither has publicly shown enough operational metrics for a plant manager to copy a mature deployment plan with confidence.
00:11:18.000 --> 00:11:20.480
Orders, production, and availability.
00:11:20.720 --> 00:11:26.559
UBTech began Walker S2 mass production and delivery in November 2025.
00:11:26.879 --> 00:11:33.279
The company reported more than RMB$800 million in Walker Series orders at that point.
00:11:33.440 --> 00:11:41.759
Its 2025 annual report describes a roughly 1,000 unit level of small-scale Walker S2 production and delivery.
00:11:42.000 --> 00:11:54.720
In September 2026, a new Liazo factory began operation with a planned annual capacity of 10,000 units and a designed line pace of one robot every 10 minutes.
00:11:55.039 --> 00:11:59.279
That capacity covers Walker S and Cruiser Series production.
00:11:59.440 --> 00:12:04.080
It is not proof that 10,000 Walker S2 units have been built or sold.
00:12:04.320 --> 00:12:17.840
UBTech's H1 2026 Interim Results Report 921 full-size humanoid sales and RMB 590.3 million in revenue from full-size humanoid products and services.
00:12:18.080 --> 00:12:22.799
Those are company-wide category figures, not Walker S2 only sales.
00:12:23.120 --> 00:12:25.840
Apptronic's reported figures are less complete.
00:12:26.159 --> 00:12:35.039
Reuters reported in June 2026 that the company had built hundreds of Apollo 2 robots, but Aptronic did not disclose deployment numbers.
00:12:35.279 --> 00:12:43.120
Its CEO said pilots would continue through 2026, with production versions expected in 2027 and beyond.
00:12:43.440 --> 00:12:50.639
For regional access, Aptronic is based in Texas and has customer and partner work in the United States and Germany.
00:12:50.879 --> 00:12:55.600
But Apollo 2 is not presented as a generally available production robot today.
00:12:56.080 --> 00:13:02.879
UbiTech has announced a European distribution partner, Terra Robotics, for Germany, Switzerland, and Austria.
00:13:03.120 --> 00:13:07.360
It has also announced a Walker S2 deal with Airbus and a partnership with U.S.
00:13:07.440 --> 00:13:09.679
semiconductor maker Texas Instruments.
00:13:10.000 --> 00:13:14.240
Those overseas deals matter, but they do not establish a broad U.S.
00:13:14.399 --> 00:13:17.440
sales, service, and parts network for every buyer.
00:13:17.759 --> 00:13:18.240
A U.S.
00:13:18.320 --> 00:13:24.320
factory should confirm supply, service response, software support, and data handling terms directly.
00:13:24.799 --> 00:13:27.919
Price, integration cost, and the business case.
00:13:28.159 --> 00:13:32.000
Neither manufacturer publishes a standard price for these robots.
00:13:32.320 --> 00:13:36.080
UBTech's H1 2026 numbers provide a rough reference.
00:13:36.399 --> 00:13:48.080
RMB 590.3 million in full-size humanoid product and service revenue divided by 921 units is about RMB$641,000 per unit equivalent.
00:13:48.320 --> 00:13:52.639
That is roughly$95,000 using a simple currency conversion.
00:13:52.799 --> 00:13:55.039
It is not a Walker S2 price.
00:13:55.279 --> 00:14:01.039
The category includes more than one product, and the revenue includes services as well as hardware.
00:14:01.200 --> 00:14:04.159
Treat it as a budget clue, not a quote.
00:14:04.480 --> 00:14:09.120
Apptronic has not published an Apollo 2 price or a customer rate card.
00:14:09.279 --> 00:14:20.399
Since Apptronic says production versions are expected in 2027 and beyond, a business case for buying Apollo 2 today cannot be completed from public pricing data.
00:14:20.559 --> 00:14:27.519
Buyers need a written pilot quote that breaks out hardware, software, integration, support, and any recurring fees.
00:14:27.759 --> 00:14:31.600
Neither company has published standard factory integration costs.
00:14:31.840 --> 00:14:42.799
Those costs can include workflow changes, safety work, fixtures, battery or charging equipment, software links to factory systems, operator training, and ongoing support.
00:14:43.039 --> 00:14:46.480
The right number depends heavily on the task and site.
00:14:46.720 --> 00:14:50.559
An illustrative payback test to show why shift coverage matters.
00:14:50.720 --> 00:14:53.440
Consider a planning example, not a vendor quote.
00:14:53.679 --> 00:15:02.000
Robot purchase price proxy,$95,000, based on UBTech's blended H1 revenue per unit calculation above.
00:15:02.240 --> 00:15:06.639
One-time integration and site setup,$50,000 assumed.
00:15:06.879 --> 00:15:10.960
Annual service and consumables,$15,000 assumed.
00:15:11.200 --> 00:15:14.559
Loaded labor cost,$30 per hour assumed.
00:15:14.720 --> 00:15:19.840
Robot completes useful work during 75% of scheduled hours assumed.
00:15:20.000 --> 00:15:24.320
Each useful robot hour fully replaces one paid labor hour.
00:15:24.559 --> 00:15:32.879
Under those assumptions, one 2,000-hour shift schedule produces$1,500 useful robot hours a year.
00:15:33.120 --> 00:15:36.639
That is$45,000 of gross labor value.
00:15:36.879 --> 00:15:47.200
After the assumed$15,000 annual service cost, simple payback on the$145,000 installed cost is about 4.8 years.
00:15:47.440 --> 00:15:51.679
At two shifts, the model produces 3,000 useful hours a year.
00:15:51.840 --> 00:15:59.120
That is$90,000 of gross labor value and$75,000 after the assumed service cost.
00:15:59.360 --> 00:16:02.960
Simple payback falls to about 1.9 years.
00:16:03.120 --> 00:16:05.200
This is deliberately a simple test.
00:16:05.360 --> 00:16:13.039
It excludes financing, software fees, taxes, and downtime beyond the assumed 25% of scheduled hours.
00:16:13.279 --> 00:16:16.960
Most important, it assumes the robot actually replaces labor.
00:16:17.120 --> 00:16:22.080
If workers still need to supervise it or handle frequent exceptions, the real savings fall.
00:16:22.320 --> 00:16:23.600
The lesson is plain.
00:16:23.840 --> 00:16:28.159
A humanoid that runs one easy shift may be an expensive experiment.
00:16:28.399 --> 00:16:34.879
A robot that reliably covers multiple shifts on a hard-to-automate task has a much better chance of paying back.
00:16:35.039 --> 00:16:37.120
Which robot should a factory choose?
00:16:37.360 --> 00:16:39.840
Choose Walker S2 for a pilot now.
00:16:40.080 --> 00:16:51.440
Walker S2 is the stronger choice if the factory wants a robot with published payload and reach figures, an autonomous battery swap system, and a company already reporting full-size robot sales.
00:16:51.679 --> 00:17:00.159
Its lead is clearest for buyers in China and for European buyers with access to UBTKS, announced DAHC distribution partner.
00:17:00.559 --> 00:17:08.880
Its main weakness is that public sales, order, and capacity figures do not prove Walker S2 only output or production grade uptime.
00:17:09.039 --> 00:17:11.920
Choose Apollo 2 for a US learning pilot.
00:17:12.160 --> 00:17:21.279
Apollo 2 is the better fit if the goal is to test a flexible platform, compare wheels with legs, or build experience with Aptronics AI and fleet system.
00:17:21.680 --> 00:17:36.000
Its ties to Google DeepMind, JABL, Mercedes Benz, and GXO give the project real industrial context, but the robot's missing public specifications and price make it a poor candidate for a purchase decision based on a spreadsheet alone.
00:17:36.240 --> 00:17:39.839
Do not choose either for a task that does not need a humanoid.
00:17:40.000 --> 00:17:46.960
If an AMR, conveyor, fixed robot arm, or simple fixture can do the work more cheaply and reliably, start there.
00:17:47.200 --> 00:17:53.200
Humanoids make the most sense when a task combines movement through a human built space with changing hands-on work.
00:17:53.440 --> 00:17:58.240
For example, moving mixed parts between stations or handling items at different heights.
00:17:58.480 --> 00:18:15.759
Before signing a pilot, require both vendors to report the same measures, completed cycles per hour, successful task rate, human interventions per shift, time lost defaults, battery and charging downtime, repair time, and measured labor hours actually removed or redeployed.
00:18:15.920 --> 00:18:16.720
Conclusion.
00:18:16.960 --> 00:18:23.920
Walker S2 offers the better 2026 business case for an executive who wants to test a factory humanoid now.
00:18:24.160 --> 00:18:28.000
Apollo 2 offers the more flexible bet on what comes next.
00:18:28.240 --> 00:18:33.200
That is not the same as saying Walker S2 is ready to replace a production workforce.
00:18:33.359 --> 00:18:34.079
It is not.
00:18:34.400 --> 00:18:46.079
UBTK's sales, customer announcements, and new factory make its commercial position stronger, but its public figures still leave key questions about S2 specific deliveries and dependable uptime.
00:18:46.319 --> 00:18:53.119
Uptronic's platform looks promising, especially where wheels, AI learning, and US-based partner work matter.
00:18:53.279 --> 00:19:00.000
But without current Apollo 2 specifications, price, and measured performance, it remains harder to underwrite.
00:19:00.160 --> 00:19:03.279
The winning robot will not be the one with the boldest demo.
00:19:03.440 --> 00:19:09.200
It will be the one that finishes the job, shift after shift, at a cost the factory can prove.
00:19:09.440 --> 00:19:13.279
All links to sources are available in the text version of this article.
00:19:13.440 --> 00:19:18.079
You can find the full article at robotcomparisons.com/slash blog.
00:19:18.240 --> 00:19:19.359
Thanks for listening.
00:19:19.599 --> 00:19:21.839
Thanks for listening to Robot Comparisons.
00:19:22.000 --> 00:19:26.960
If you found this episode valuable, make sure to subscribe so you never miss a new deep dive.
00:19:27.119 --> 00:19:34.559
For the full written breakdown, detailed spec sheets, and side by side comparisons, head over to robotcomparisons.com.
00:19:34.720 --> 00:19:40.319
That's R O B O T C O M P A R I S O N S dot com.
00:19:40.559 --> 00:19:42.880
New research drops multiple times a week.
00:19:43.039 --> 00:19:44.079
See you in the next one.
