Text by Xiao Luyu, Editor: Yang Xuran
Mech-Mind's first financial report since its IPO can be described as particularly crucial. Its data not only demonstrates the company's own operational changes but also reflects the development trends of the entire embodied AI industry.
From the perspective of its own operations, both growth and losses are quite pronounced. In the first half of this year, the company achieved a revenue of CNY 240 million, a year-on-year increase of 54.7%; the gross profit margin was 65.0%, up 3.6 percentage points year-on-year; overseas revenue reached nearly CNY 100 million, a year-on-year increase of 68.3%; newly signed orders amounted to CNY 335 million, a year-on-year increase of 75.3%, with a repurchase rate of 93.9%.
Meanwhile, the company's operating loss has expanded year-on-year to CNY 77.45 million, of which R&D expenses reached CNY 80.1 million, a year-on-year increase of 72.9%. The net loss was approximately CNY 100 million, with limited narrowing.
More noteworthy is its revenue structure. Currently, nearly 90% of the company's revenue still comes from industrial 3D cameras and vision software. The "eyes" remain the core business and absolute cash cow, while the "brain" and "hands" are still in the investment phase—which is somewhat inconsistent with the premium given by the capital market during the IPO.
Mech-Mind's price-to-sales ratio of nearly 20 times is far higher than the average of about 1.1 times for the machinery industry on the Hong Kong Stock Exchange and about 4.7 times for its peers. This valuation is clearly built on the market's expectation of the future volume growth of general-purpose robots.
When the traditional business can barely sustain cash flow, but the new stories cannot yet contribute scaled revenue, this first post-IPO financial report appears somewhat pressured in the face of high expectations.
This article is an in-depth value piece from the content team of "JuChao WAVE". Welcome to follow us across multiple platforms.
Technical Barriers
Out of Mech-Mind's total revenue of CNY 237.3 million in the first half of 2026, the embodied AI robot guidance business contributed CNY 216.3 million, a year-on-year increase of 49.9%, accounting for 91.1% of total revenue.
The core products corresponding to this business line are the Mech-Eye industrial 3D camera, along with the supporting Mech-Vision vision software and Mech-Viz robot programming software. These help robots identify the position and pose of workpieces, plan motion paths, and perform operations such as grasping, handling, and loading/unloading.
In terms of technical parameters, the Z-axis single-point repeatability accuracy of Mech-Eye can reach 0.3mm@2.0m, and the VDI/VDE measurement accuracy reaches 0.2mm@2.0m, which is far higher than the industry average benchmark of 0.5 to 1.0 mm in well-controlled environments.
In industrial scenarios, this accuracy means the camera can generate a complete 3D point cloud for thin-walled sheet metal parts with a thickness of less than 1mm, clearly presenting features such as tiny circular holes and warpage on the surface. The imaging speed can reach up to 0.08 seconds per frame capture, adapting to the takt time requirements of high-speed production lines such as express parcel supply, which is about 1,800 pieces/hour.
In other words, the technical barrier of this set of "eyes" genuinely exists, supporting Mech-Mind's market share of nearly 40% in China's 3D vision-guided robot market.
However, the ceiling of the "eyes" is also a real existence. It essentially solves the problem of how robots "see," identifying where objects are and in what pose they exist. As for what the robot should do after "seeing," why it should do it, and how to respond to unseen situations, these belong to the realm of the "brain."
In Mech-Mind's product architecture, the "brain" corresponds to the Mech-GPT multimodal large model. From a technical definition perspective, Mech-GPT is based on a self-developed vision-language-action multimodal architecture, integrating perception, language understanding, and motion planning capabilities, and has been trained on over 100,000 real objects and industrial scenarios.
Its goal is to enable robots to directly understand natural language instructions, synthesize visual inputs and 3D spatial data for reasoning, and independently determine the optimal solution to complete tasks without predefined programming.
However, it should be noted that Mech-Mind's "brain" is not essentially a "brain" in the sense of Artificial General Intelligence (AGI), but rather an intelligence enhancement technology at the industrial automation level.
Its value is more reflected in making existing industrial robots smarter and more flexible in specific production line tasks. For example, when facing scenarios that are difficult for traditional automation to handle, such as randomly piled workpieces, reflective surfaces, and transparent materials, the model can independently complete recognition, decision-making, and motion planning, compressing the deployment cycle from months to days.
Moreover, even if it is not a general-purpose brain, making substantial technical progress and implementing it in business is not an easy task.
Mech-Mind's prospectus reveals that its "eyes" and "brain" have not yet reached an ideal coupling state at the technical level. However, the next-generation technological leap of the "eyes" precisely relies on the capability spillover of the "brain."
If large models with genuine general task understanding and planning capabilities emerge in the industry in the future, the moat currently built by Mech-Mind on "proprietary accuracy for industrial scenarios" will undoubtedly face challenges.
Once the "brain" is sufficiently general-purpose, will the "eyes" still need such a high degree of customization? Can a general-purpose brain paired with standardized vision sensors cover most of the scenarios currently served by Mech-Mind?
Obviously, this is a rather difficult question for Mech-Mind.
Customer Base
Summarizing current public information, Mech-Mind's unshakeable position in industrial scenarios perhaps stems more from being the first to enter the system integrator network and making integrators willing to relearn, readapt, and re-assume project risks for it. Compared to its technical capabilities, this is a higher and harder-to-break barrier.
The implementation of industrial automation projects has never been as simple as buying a camera, plugging it in, and using it. For a typical 3D vision-guided project, integrators need to complete a series of engineering processes, including camera calibration, hand-eye calibration, point cloud parameter tuning, grasping strategy configuration, and exception handling logic writing.
After years of iteration, Mech-Mind's Mech-Vision and Mech-Viz software have formed a relatively mature toolchain and developer ecosystem.
Since the integrators' engineers are already familiar with the operation logic and debugging methods of this interface, they are reluctant to relearn a new set of software, re-encounter engineering pitfalls, and re-explain to end customers why the project cycle has been extended, as there is a lot of engineering inertia and organizational cost involved.
From a data perspective, Mech-Mind's customer base is thickening: as of the end of March 2026, the cumulative number of paying customers reached 1,605, including over 100 Fortune Global 500 companies.
The customer repurchase rate reached 93.9% in the first half of 2026. The proportion of revenue contributed by existing customers rose from 61% in 2023 to 78% in 2025, and further reached 86% in the first quarter of 2026.
However, if interpreted from another perspective, this set of data can also be understood as Mech-Mind's revenue becoming increasingly reliant on continuous procurement and expanded deployment from existing customers. This can be a signal of improved customer stickiness, or a sign that the industrial "eyes" have completed market penetration and entered a stage of deep cultivation of the existing stock.
On this basis, the trend of rising customer concentration is worth being alert to—according to the prospectus and financial reports, the proportion of revenue from Mech-Mind's top five customers increased by 14.5 percentage points over two years, with the share of the single largest customer jumping from 6.4% to 15.0%.
This concentration itself is not extreme, but the particularity of Mech-Mind's customer structure lies in the fact that these customers are essentially system integrators rather than end-users, which makes the financial implications of the company's concentration more complex than they appear on the surface.
Integrators purchase Mech-Mind's products and then integrate them into the production lines of end customers (such as automakers and logistics companies). With a layer of integrators between Mech-Mind and the end customers, the perception of real demand changes, budget cycles, and technical preferences in end scenarios is indirect.
The risk of this structure lies in the fact that, in extreme cases, if core integrators switch to self-developed vision solutions or change suppliers, it will be very difficult for Mech-Mind to bypass the integrator and directly reach its end customers in the short term. Integrators in the field of industrial automation may not lack the motivation to do so, after all, the gross profit margin of vision-guided systems is high enough.
Another dangerous signal is that automakers have recently been making massive forays into the robotics industry. Automakers are not only the core industry where Mech-Mind's end customers are located, but also an important source of demand for the integrator network.
When automakers start building their own robotics teams and developing their own vision-guided solutions, they may become new customers for Mech-Mind, but they may also become replacements for integrators, further compressing Mech-Mind's channel to reach end customers through integrators.
Here, the risk of the "brain" business progressing too slowly must also be superimposed. After all, no matter how solid the customer base of the "eyes" is, it is still the "brain" that truly determines long-term customer stickiness.
Valuation Pressure
How long a window period the market can still leave for Mech-Mind to prove itself is probably the most critical question.
Mech-Mind's current positioning in the capital market is as a company in the stage of building technical barriers and acquiring market share, where growth takes priority over profitability. However, placed at a time node when the profitability of the entire robotics industry is being re-examined, this self-positioning may also waver.
From the perspective of its loss structure, the operating loss in the first half of the year expanded year-on-year to CNY 77.45 million, while the adjusted loss for the period narrowed from CNY 55.8 million in the same period last year to CNY 52.5 million. The difference of approximately CNY 24.95 million is highly likely to come from share-based payment expenses, which also aligns with the statement that the company is still in a stage of strengthening R&D investment.
The adjusted EBITDA loss decreased from CNY 39.44 million to CNY 32.96 million. The main reason for the narrowing of the loss is that the revenue growth rate is faster than the expense growth rate, not that a qualitative change has occurred in the expense structure itself, and even less that there has been a substantial improvement in the company's fundamentals.
Mech-Mind's R&D expenses in the first half of the year were CNY 80.1 million, a year-on-year increase of 72.9%, accounting for about 33.8% of revenue. Although the sales expense ratio dropped from 102.7% in 2023 to 38.5%, it remains at a high level.
When the combined R&D and sales expenses approach CNY 180 million, while the gross profit is only CNY 154 million, the company's loss looks more structural than cyclical.
Its gross profit is certainly competitive; a comprehensive gross profit margin of 65.0% is already at a top-tier level in the industrial field, and the gross profit margin of overseas business is as high as 81.4%. However, without rapid growth in business scale, no matter how high the gross profit margin is, it is not safe—Mech-Mind's revenue of CNY 237 million in the first half of the year is too low for a company with a market capitalization at the CNY10 billion level.
And the expectation of a decline in Mech-Mind's high gross profit margin is also an issue that cannot be avoided.
The current trend in the robotics industry is to accelerate the decline in the price of complete machines in order to achieve popularization. The unit price of Unitree's humanoid robots has dropped by about 70% over two years. The Global Physical AI Report released by Goldman Sachs in August this year also predicts that the average price of humanoid robot complete machines will drop by about49% over the next decade.
The price reduction of complete machines will inevitably transmit upstream along the industrial chain. If Mech-Mind wants to be a foundational enterprise in the robotics industry, it cannot remain unaffected in an environment where overall gross profit margins will be under pressure in the future.
Compared to the compression of gross profit margins brought by the price reduction of complete machines, what Mech-Mind needs to worry about more is the various complex situations that may arise in the robotics industry: the disappointment of expectations caused by delayed popularization, implementation, and hindered commercialization.
At that time, the company will either have to cut advanced investments such as R&D, or continue to burn cash while waiting for a market turning point. However, the financing tolerance of the entire market for continuously loss-making robotics companies will rapidly decline.
Despite a significant plunge in its stock price after the IPO, Mech-Mind still maintains a price-to-sales ratio of nearly 20 times. This pricing is obviously not based on the company's current revenue scale and profit level, but rather bets on the irreplaceable ecological niche of the "eyes, brain, and hands" platform in the era of general-purpose robots.
Similar to the relationship between the NEV (New Energy Vehicle) industry and CATL (Contemporary Amperex Technology Co., Limited), the relationship between the robotics industry and Mech-Mind is one of sharing weal and woe; they prosper together and suffer together.
If the robotics industry cannot truly achieve volume growth, the company's technical premium and valuation narrative will lose their foundation. If priced according to the logic of machinery and industry, it is obviously something the management is very unwilling to see.