How Can Optimus Generate Revenue Without Any Customers?

“

The ability to produce is no longer an issue, the ability to recall is the key

。

By the end of June 2026, Musk hosted a senior management review meeting, which approved the latest version of the third-generation Optimus.

After the review meeting, Tesla issued a clear directive to its supply chain, aiming for a weekly production of 1,000 units in September, 2,000 to 2,500 units by the end of the year, and the supply chain should be capable of providing 100,000 sets of parts annually by the end of the year.

This directive was interpreted by the market as a signal of accelerated production, and indeed it was. In early September, Tesla placed a large order for thousands of Optimus units with its core suppliers, which was the first batch-level order since the production guidance was launched in April. By the end of September, Tesla's team initiated a new round of production factory audits for the Chinese supply chain, with the focus on the consistency of robot parts and the ability to deliver mass production.

From the second half of 2026, production capacity is no longer an issue for Tesla.

In other words, the problem with Tesla's Optimus is no longer about being unable to produce it.

The Fremont production line was completed in 2026, and the停产 of Model S and Model X provided 46 days for the transformation. The new factory in Texas is under construction, with a long-term design capacity target of one million units annually. Manufacturing capacity is not a bottleneck.

The issue is where these robots go after they are produced.

01.

Production does not equal revenue

Tesla has not disclosed any operational metrics for Optimus. Online rate, cycle time, error rate, and throughput are the standard disclosure items for industrial equipment. A company that has been making cars for over a decade has remained silent on these four metrics.

And beyond these four, there is another key metric that is not disclosed, but the way it is disclosed itself indicates a problem.

A very straightforward description of the commercial status of Optimus in public materials is that there is currently no clear plan for external sales, and every unit produced is counted as capital expenditure, not immediately generating revenue.

We looked at Tesla's disclosed financial reports and found that Tesla's capital expenditures and operating revenues are two completely different lines in accounting.

When a robot is produced, it actually enters Tesla's fixed assets category, generating no revenue, only depreciation.

Depreciation eats into profit and is accounted for as a cost, adding a line of amortization to the financial statements.

It is not a product; it becomes an asset on the balance sheet.

So the production line of 2,000 units per week is drawn as fixed assets, not as a product line.

This distinction is not visible in Tesla's financial statements alone, because Optimus expenditures have never been disclosed separately.

In fact, it is buried in research and development expenses or amortized into factory equipment costs.

The outside world can only infer scale from production data, and cannot judge what these assets are doing through financial reports.

This is why the more production data there is, the more ambiguous people's judgment of Optimus becomes.

02.

The production of Chinese companies has become revenue

Switching the lens to Chinese peers, if they had followed Tesla's approach, they would have been discarded by capital long ago. However, at the same time, it is clear that they have redefined the word 'production' in a new way.

As of the end of March 2026, the cumulative production of general-purpose embodied robots by Zhixun Robotics exceeded 10,000 units.

This number was cited by several securities firms in their September tracking reports as evidence of accelerated mass production.

In 2025, the base model of Yu Shi Technology exceeded 6,500 units of production, and the招股书 disclosed that the shipment of humanoid robots exceeded 5,500 units, with revenue of 1.708 billion yuan and a non-GAAP net profit of 600 million yuan. The company's market value reached 60.993 billion yuan on the day of its listing on August 19, 2026.

The robots produced by these two Chinese companies enter the customer's production line, receive payment, are recorded as revenue, and are clearly written into the financial statements.

But Tesla's over 1,000 Optimus units enter its own production line, with tasks assigned by itself, data collected by itself, and acceptance judged by itself.

It can be said that there are currently no external customers disclosed for Optimus.

This is the result of Tesla's economic calculation.

According to a breakdown by Morgan Stanley in February 2026, the material cost of Optimus Gen2 is $46,000, and if the Chinese supply chain is removed, the cost would rise to $131,000. Musk described this situation as having no supply chain in his second-quarter earnings call on July 22, 2026.

On August 25, 2026, Bernstein released a calculation of the economic viability of humanoid robot applications, for the first time systematically quantifying the payback period for four types of scenarios.

The results differed significantly from industry expectations. Simply put, Tesla's Optimus is not expected to sell in the short term.

For example, in urban and industrial patrol scenarios, the payback period in the US is 1.0 year, and in China it is 1.5 years. In warehouse picking and handling, the payback period in the US is 2.9 years, and in China it is 4.3 years. In factory material handling and inspection, the payback period in the US is 2.7 years, and in China it is 6.7 years. In last-mile delivery, the payback period in the US is 5.0 years, and in China it is 4.2 years.

The biggest difference is not the difficulty of the scenarios, but the wages. Bernstein's calculation showed that the reason for the longest payback period for factory material handling in China is that the manufacturing wages are only 23% of those in the US, while the robot's price is close to that in the US. The same robot, when bought in different countries, has a payback period that differs by more than double.

Bernstein therefore proposed a 2.5-year threshold. If the payback period drops below this line, the logic of procurement decisions will shift from whether to try to why not to buy. To cross this line, the price of robots in the Chinese market needs to decrease by 30% to 50%, and in the US market by 5% to 40%.

In the same month, another report from IDTechEx gave even more extreme numbers. Under high utilization conditions, the payback period for humanoid robots can now be compressed to 6 months, and under medium utilization conditions, it would take 15 months.

The same robot, with two different operational intensities, differs by 9 months, and the cause of this 9-month difference is not hardware, but deployment methods.

These two calculations point to the same conclusion. The factor that determines the payback is not the price of the robot, but the utilization rate of the robot. And the utilization rate goes back to the oldest question: how many hours of effective work can the robot do in a day.

If the material cost of Optimus is $46,000, and the long-term retail price target is $20,000 to $30,000, there is a negative gross margin in between. This negative gross margin can either be spread out by increasing production volume or be internally absorbed.

Having the factory itself be the first customer is the only way Tesla can make the accounting work for now. The cost is that these robots cannot easily meet external acceptance standards in the short term.

03.

The utilization rate is determined by the bill of materials

Any factory operating globally decides to buy or not, and the cost table is not a single robot, but a list of items.

The robot itself, production line transformation, safety fence, charging facilities, deployment integration, maintenance spare parts, and software subscription. Any item on this list out of control will turn the previously calculated 6 months into 15 months.

Among these, the integration is the easiest to overlook. A 2026 industry calculation mentioned that the integration engineering cost for enterprise-level deployment may be between $100,000 and $500,000, and site renovation costs per site may add another $50,000 to $200,000. This money is not included in the robot's quotation, but it is definitely included in the CFO's financial model.

Therefore, to assess a robot company's commercialization capability, looking at the price it reports is not useful; instead, we should look at whether its bill of materials can be thinned out.

The most popular cost-cutting story in the industry over the past two years is the planetary roller screw. This component has high machining precision requirements, long service life, and high price, once considered a major cost-saving area for humanoid robots.

Tesla is now turning this direction around.

According to a report by Huafu Securities on September 21, 2026, with the continuous optimization of Tesla's robot solution, the usage of low-cost, easy-to-maintain roller screws is expected to increase, and the number of roller screws in the hand may also increase further. This securities firm also judged that the harmonic drive solution remains the main one, and RV and other solutions are expected to be alternatives.

Cost reduction is also happening in the area of dexterous hands. According to Counterpoint's estimation, the material cost of Tesla's Optimus in the second half of 2026 is still over $60,000, with the 22-degree-of-freedom dexterous hand accounting for about 20% of it. Musk's long-term retail price target of $20,000 to $30,000 is the common target range for the entire industry. Counterpoint judged that to reach this price, annual production needs to far exceed 100,000 units.

According to a tracking report by Open-Source Securities, the motor solution for the hand of Optimus may shift from the higher-cost hollow cup motor to the miniature frameless torque motor. The transmission solution for the third-generation dexterous hand is a composite structure of planetary gearboxes, miniature roller screws, and tendon cables, replacing the worm gear with the roller screw and the torsion spring with the tendon cable.

These changes are all in the same direction, replacing high-cost components with low-cost ones. Tesla is not incapable of cost reduction; it is actively reducing costs. And in this aspect, the position of the supply chain is more critical than the overall design.

There is a more detailed clue. On October 1, 2026, Musk announced on his social platform that the memory of AI5 chip was cut from 72GB to half, and the next day changed it back to 96GB. The memory of AI6 was reduced by a third to 144GB from the original plan.

According to data from SemiAnalysis, the contract price of LPDDR5 in early 2026 was around $10 per GB, three times that of early 2025, and the long-term contract price ceiling is higher. According to this price, the memory cost saved per unit is $500 to $1,000.

Saving $500 to $1,000 per robot is 1% to 2% of the material cost.

This is barely better, but it shows that Tesla is still fighting with performance and cost, rather than missing parts.

04.

The supply chain is no longer waiting

What is the price level of Chinese component manufacturers? From the perspective of overseas supply chain costs, it is already sufficient to meet the requirements, even for Chinese factories, they have already reached the ROI acceptable red line.

An institution calculated that in the first quarter of 2026, the single-unit cost of Chinese humanoid robots dropped to 100,000 yuan, a 33% decrease from 150,000 yuan in 2025. For example, the industrial version of Ubtech dropped to 128,000 yuan, the GR-3 of Fuyilai Intelligent reached 115,000 yuan, and the public procurement prices of various integrated versions of Yu Shi G1 have already fallen below 100,000 yuan.

The domestication path of harmonic reducers is the clearest footnote to this curve.

According to a report by Guotai Securities on September 20, 2026, comparing China and the US, there were four years of data. In 2017, the domestication rate was 16.41%, which increased to 55.15% in 2021, surpassing foreign capital for the first time, and further increased to 75.11% in 2024. At the same time, Japan's Harmonic Drive in 2025 had a global market share of 58%.

The simultaneous rise in domestication rate and foreign capital share indicates that the domestic replacement in this field is not a simple replacement, but market expansion, which can form scale benefits.

Tesla's September 2024 China supply chain audit, which focused on this location, is not about whether it can be manufactured, but about consistency. After passing the audit, those components that only crossed the 75% domestication rate in 2024 could potentially enter its bill of materials (BOM).

According to MinYin International's value decomposition of the supply chain as of September 14, 2026, the actuator assembly (motor, reducer, screw) of Optimus accounts for about 39% of the total BOM cost, with the screw accounting for about 13%, the reducer about 9%, and the motor about 17%. Sensors, bearings, and dexterous hands account for the remaining 31%. GuoTai HaiTong's same-day comparison study gave a figure of about 45% for the actuator in the total BOM cost, based on the Figure 03 sample.

The two brokerage reports differ in口径 and samples, but they point to the same conclusion: the actuator plus dexterous hand accounts for the majority of the total material cost.

This is the main battlefield for cost reduction. It is also what Tesla is really auditing during its factory audit. According to data cited by Jiemian News from industry estimates, the material cost of Optimus without relying on the Chinese supply chain is $131,000, and with reliance, it drops to $46,000. The nearly threefold difference comes from the domestic substitution of core components.

However, the pace of domestication varies across different components. The domestication rate of harmonic reducers has exceeded 72%, with the single-unit procurement price dropping from 1,280 yuan in 2023 to 410 yuan, a decrease of about 68%. For specialized servo motors, the domestication rate has risen from 23.1% to 69.5%, with the unit price about one-third of that of Japanese counterparts.

Tesla's position on this chain is somewhat special.

It has the most complete electric vehicle supply chain globally, but the supply chain for humanoid robots is another. Bernstein used an experience rule that is repeatedly cited in the industry: production volume doubles, and costs decrease by 15% to 18%, with a steeper curve in the small-batch phase.

This means that in the early stage of mass production, relying on an in-house supply chain could theoretically offer a steeper cost reduction. The prerequisite is volume. This is why its long-term production capacity targets have been repeatedly discussed.

In this context, Tesla's choice becomes clear. It deployed over 1,000 Optimus units entirely within its own factory, not for external sales. Its material cost is being pushed down to $20,000 to $30,000, not through external orders spreading the cost, but by using its own factory as the first customer.

This path can work only if the self-use scale is sufficiently large. The prerequisite is also the one everyone is waiting for: whether the factory is willing to use it.

05.

Who is paying for this mass production?

Looking at the timeline more recently, between October 1st and 2nd, Musk adjusted the memory chip scheme. His long-term goal is to produce 20,000 units of Optimus per week.

A week of 20,000 units, a year of 10 million units. This was the line he drew in 2024, and it still stands today.

From the actual red line of 2,000 to 2,500 units per week, to the long-term target of 20,000 units per week, there is a fourfold gap. This fourfold will not be achieved through capacity expansion, but will be determined by the thickness of the bill of materials. And the thickness of the bill of materials depends on whether the companies on the supply chain, not listed in this financial report, can push the quotations down a little more.

Therefore, the acceleration of Optimus production, the real beneficiaries are not entirely Tesla. After the round of audits, certifications, expansion, and price reductions, the first to benefit are the component suppliers, the first to be compressed is the gross margin of components, and the final result is borne by that yet-to-be-revealed material cost table.

Tesla has already achieved the capacity for producing 20,000 units of Optimus. But after the 20,000 units are off the line, each unit is a capital expenditure, not revenue.

From zero to 100,000 units, Chinese companies took three years, and during these three years, each unit was sold. From 1,000 units to 2,000 units, Tesla took half a year, and during these half a year, each unit was depreciation.

These two curves seem to be racing, but they are actually not on the same track.

What is truly worth watching next is not the production volume, but the day when the first Optimus is recorded in revenue rather than fixed assets.

Before that day arrives, the production numbers are just announcements of capacity, not the beginning of the product.

The indicators we can track are three.

First, Optimus officially publishes operational metrics.

Online rate, cycle time, error rate, any one of these being disclosed means it has moved from an internal R&D project to a deliverable product. If these four numbers remain silent, then even if the production volume is high, it is still an R&D project.

Second, whether the core components' domestic substitution has clearly entered Tesla's supply chain.

This determines whether the $60,000 can really be reduced to $20,000.

Third, the first external sale.

Any Optimus sold to a non-Tesla entity, whether through sale, lease, or hourly billing, will prove that its bill of materials can withstand an external contract.

The industry has already written the acceptance criteria clearly. Bernstein said that if the payback period drops below 2.5 years, the procurement decision will shift from whether to try to why not buy.

For Optimus, this threshold corresponds not to a production number, but to how much water is still in the bill of materials.

END

Industrial Robot Companies

Eston Automation

|

Efte Robot

|

Fao Robot

|

Yuejiang Robot

|

Jieka Robot

|

Songling Robot

|

Luoshi Robot

|

Atongmu Robot

|

Jizhi Jia

|

Hikvision Robot

|

Yifei Technology

|

Ailite Robot

Service and Specialized Robot Companies

Yijiahe

|

Jingpin Tezhuang

|

七腾机器人

|

史河机器人

|

普渡机器人

|

施罗德机器人

|

库犸科技MAMMOTION

人形机器人企业

优必选科技

|

宇树

|

云深处

|

星动纪元

|

伟景机器人

|

逐际动力

|

乐聚机器人

|

大象机器人

|

魔法原子

|

众擎机器人

|

帕西尼感知

|

赛博格机器人

|

数字华夏

|

傅利叶智能

|

天链机器人

|

开普勒人形机器人

|

灵宝CASBOT

|

清宝机器人

|

浙江人形机器人创新中心

|

动易科技

|

智身科技

|

PNDbotics

|

卓益得机器人

|

擎朗智能

|

伽利略GALILEO

|

松延动力

|

天机智能

|

卧安机器人

|

理工华汇

|

加速进化

具身智能企业

跨维智能

|

银河通用

|

千寻智能

|

灵心巧手

|

睿尔曼智能

|

微亿智造

|

推行科技

|

中科硅纪

|

枢途科技

|

灵巧智能

|

星尘智能

|

穹彻智能

|

方舟无限

| 科大讯飞 |

北京人形机器人创新中心

|

国地共建人形机器人创新中心

|

戴盟机器人

|

视比特机器人

|

星海图

|

月泉仿生

|

零次方机器人

|

中科深谷

|

智平方

|

大咖机器人

|

灏存科技

|

具识智能

|

Xynova曦诺未来

|

非夕科技

|

未来动力

|

博登智能

|

千诀科技

|

灵生科技

|

集萃智造

|

欣佰特科技

|

晨昏线科技

|

Dexmal 原力灵机

|

优理奇

|

自变量

|

睿研智控灵巧手

|

启物科技

|

RoboScience机器科学

|

中科第五纪

|

临界点

|

当虹科技

|

桥介数物

|

Vbot维他动力

|

他山科技

|

具脑磐石

|

优艾智合机器人

|

智行腱

|

阿米奥机器人

|

鹿明机器人

|

量子智能

|

若愚科技

|

钛虎机器人

|

智无际

医疗机器人企业

元化智能

|

天智航

|

思哲睿智能医疗

|

精锋医疗

|

佗道医疗

|

真易达

|

术锐®机器人

|

罗森博特

|

水木东方

｜

康诺思腾

|

迪视医疗

上游产业链企业

绿的谐波

|

因时机器人

|

坤维科技

|

脉塔智能

|

青瞳视觉

|

本末科技

|

蓝点触控

| 鑫精诚传感器

|

BrainCo强脑科技

|

宇立仪器

|

极亚精机

|

思岚科技

|

神源生

|

非普导航科技

|

因克斯

|

巨蟹智能驱动

|

凌云光 元客视界

|

璇玑动力

|

意优科技

| 瑞源精密 |

灵足时代

|

HIT华威科

|

星汇传感

|

凌迪科技

|

泉智博

|

CubeMars机器人动力

|

旺龙机器人乘梯

|

航凯微电

|

悟通感控