In the first eight months of 2026, the sales profit margin of the automotive industry stood at merely 3.6%, a significant drop from 4.3% in 2024 and 4.1% in 2025, with total profits declining by 16% year-on-year. Under the multiple pressures of price wars and annual price reductions, some component suppliers would rather forfeit large orders than accept supply contracts destined for losses.
In fact, in recent years, as the core business has become increasingly difficult to generate profits, a group of component companies have turned their attention beyond automobiles. AI data centers, embodied intelligence, low-altitude economy, and energy storage—these new scenarios that seem far removed from the core automotive business at first glance—are now seeing an accelerating influx of component companies making strategic moves.
However, this cross-boundary expansion is not open to everyone. Those who can truly step out are mostly those whose technologies align well with the new tracks. Why are they crossing over? What do they see in it? And can they gain a firm foothold in this new territory?
These Tracks Are Now in the Crosshairs of Automotive Component Suppliers
AI data centers represent a sector with relatively higher certainty at present.
According to estimates by CICC, the global market size for liquid cooling in intelligent computing centers will reach CNY 114.7 billion in 2026, a year-on-year growth of 273%. Data from the China ICI Research Institute shows that the penetration rate of liquid-cooled servers in China is expected to surge from about 20% in 2025 to 37% in 2026, and will exceed 50% in 2027.
The most direct entry point is liquid cooling thermal management. As the power consumption of AI servers pushes air cooling to its limits, liquid cooling has shifted from an optional choice to a mandatory one. The capabilities required to precisely transport, distribute, and control the temperature of the coolant within the racks—such as sealing, fluid piping, heat exchange, and cold plate manufacturing—are exactly the skills that automotive thermal management suppliers have honed over decades.
Consequently, a group of companies that originally focused on engines and batteries have seized the opportunity to enter the field. Manufacturers of seals and fluid pipelines have extended their products into liquid cooling pipelines; chassis makers have started stamping liquid cooling cold plates; even giants in safety components are expanding into liquid cooling racks. Several leading thermal management companies have even listed data center liquid cooling as a strategic emerging business, setting up dedicated teams and production lines to cultivate it as a second growth curve.
Of course, data center thermal management is not a blank slate. Professional manufacturers such as Envicool, Shenling Environmental, and Goaland have been deeply engaged in this field for many years. For the newcomers, this is a piece of cake to be snatched from others, rather than unclaimed land.
However, thermal management is just one entry point. What automotive component companies bring is not just cooling capabilities; the expertise accumulated in power, connectivity, and structural design on vehicles is being repackaged and integrated into the new scenario of data centers.
The most typical example is Bosch. Its main focus in data centers is not cooling, but power supply and power conversion.
Bosch included data center power supplies as an application direction in 2023. At the 6th Intelligent Automotive Chip Ecosystem Conference in 2026, Wang Junyue, Senior Manager of Power Semiconductor Products and Head of Global Business for Silicon Carbide Data Centers at Bosch, mentioned that building on this, Bosch officially launched its global AI data center business on July 1 this year. The core is to provide SiC power semiconductor solutions for AI data centers, especially for Solid State Transformers (SST), which is almost an all-silicon carbide solution.
Connectivity and electrical architecture companies are also entering the fields of high-voltage, high-speed connectivity and power distribution, while other companies are manufacturing structural components such as racks and chassis.
Yang Xiaoming, President of Aptiv China and Asia Pacific, recently stated at GNEV 2026 that China's non-automotive market is bringing new growth opportunities, one of which is data centers. "Data centers are growing at a very fast pace, and the domestic market is expected to grow into a globally leading market in terms of scale."
Embodied intelligence is another highly imaginative track.
Humanoid robots are the most watched category among them. According to Wang Xianbin from Gasgoo Research Institute, global shipments of humanoid robots will reach 60,000 units this year, with the Chinese market accounting for about 50,000 units, an increase of more than twofold from 18,000 units in 2025.
Of course, it is not just humanoid robots; the areas that automotive component companies are truly entering also include a broader range such as collaborative robots, autonomous mobile robots, and industrial and service robots.
Automotive component companies are entering the robotics sector through various paths.
For chassis and precision machining companies, the entry point is the joints of robots. Relying on their manufacturing experience in ball screws, reducers, and actuators, they have a relatively easy time entering the market and are currently mostly in the stage of small-batch delivery.
One of the entry points for automotive electronics companies is the brain and senses of robots. They are migrating their intelligent cockpit and intelligent driving domain control and perception capabilities to robots. However, the progress of domain control and perception is not consistent: domain control has already achieved mass production and delivery, while most perception products are still in the sampling, validation, or small-batch stage. Whether they can scale up depends on the pace of the OEMs, but Tier 1 suppliers are also securing design wins and building capacity in advance.
There is also a type of system supplier that chooses to use robots in its own factories first. Taking Aptiv as an example, according to its announced plan, by the end of the year, more than 100 embodied intelligence robots will be deployed on the production lines of its Shanghai factory to undertake actual production tasks. These robots come from TARS in Shanghai, and Aptiv's role is primarily to act as the scenario customer and validator by opening up its factory.
At the other end of the industry, it is also extending automotive electronics capabilities such as PULSE radar-vision fusion perception to collaborative robots and autonomous mobile robots, cooperating with robotics companies like Robust.AI and Comau to compete for supplier positions.
The low-altitude economy is the track with the farthest vision and the hardest threshold for implementation.
As eVTOLs gradually enter the certification and delivery stages, the low-altitude sector is regarded as another trillion-dollar space. Morgan Stanley predicted in its flying car market report several years ago that the global broad market could reach USD 300 billion by 2030.
The demand for power management, motor control, thermal management, and lightweight structures in aircraft is highly homologous to that of electric vehicles, which has allowed a group of three-electric system suppliers to see opportunities. Joyson Electronics has already developed a three-in-one product integrating battery management, voltage conversion, and power distribution for a foreign eVTOL company, secured the design win, and expects mass production in the first quarter of 2027.
Energy storage, on the other hand, is the track closest to the automotive technology base and relatively easier to generate revenue.
Data from the National Energy Administration shows that by the end of 2025, the cumulative installed capacity of new energy storage nationwide had reached 136 million kilowatts, a year-on-year increase of 84%, and will further rise to 153 million kilowatts by the end of June 2026.
Photovoltaic inverters, energy storage thermal management, BMS, and high-voltage connectors—these products originally supporting the three-electric system of electric vehicles—share a common technological base with energy storage power stations, making the transition relatively smooth. However, commonality does not mean direct copying. Long cycle life, large-capacity grouping, and grid safety all require companies to redevelop products and undergo recertification. The energy storage industry itself also has a considerable safety threshold and fierce price competition.
These tracks look bustling, but it needs to be reminded that there is still a considerable distance between securing design wins and contributing to profits. As for why these companies are so determined to cross over, we need to look one layer deeper.
The Reasons for Crossing Over Are More Realistic Than Imagined
This cross-boundary expansion is not a momentary impulse, but a choice made after the superposition of multiple pressures.
The most direct pressure comes from profits.
The automotive industry has a mechanism that has lasted for many years, called "annual price reduction." OEMs usually require suppliers to lower prices every year. In the past two years, as the price war for complete vehicles has intensified, this pressure has been transmitted layer by layer to the component end.
Huaguang Co., Ltd. (874806), which mainly produces automotive body components, saw the gross profit margin of its new energy supporting products drop from 13.95% in 2023 to 10.64% in 2024, and further to 8.21% in 2025, a decline of nearly 6 percentage points over three years.
What makes companies more anxious is that this round of low profits is not a cyclical trough that can be survived by enduring, but a structural problem caused by overcapacity and the long-term downward shift of the profit margin center.
When the gross profit of a set of components is reduced to single digits, a few companies that secured their positions early can still achieve a gross profit margin of around 27% for robot actuators. Liquid cooling and energy storage thermal management had even thicker gross margins in the early window of supply shortage in the past two years, but they are now declining as more players rush in. However, high gross margins are not a universal phenomenon; more companies are still subsidizing their new businesses to exchange for orders. Even so, where capacity and capital are heading is self-evident.
Beyond profits, what makes companies more uneasy is the ceiling of growth.
China's automotive market has shifted from incremental expansion to stock game. After the penetration rate of NEVs in the domestic market exceeds half, even the most storied electrification has entered close-quarters combat. When a supplier's share in the automotive sector reaches its limit, the ceiling is clearly visible.
Robotics and data centers, on the other hand, are markets from 0 to 1, where early entrants have the opportunity to participate in defining standards and securing positions with leading customers. The proposition of Guangyang Co., Ltd. is very representative: "Based on automotive, stepping out of automotive."
At a seminar in September, Zhang Yongwei, Chairman of the Board of the Chebaihui Research Institute, explicitly summarized the extension into new tracks such as embodied intelligence, low-altitude, and energy storage as the "second growth curve" for automakers and component companies, advocating expanding the automobile from a single means of transportation to a new hardware platform and value platform.
In his judgment, the automobile no longer needs to be treated as a closed transportation vehicle industry, but rather a hardware platform that can be opened to robotics, low-altitude, and energy storage. For suppliers, this means that the objects of service are no longer just limited OEMs, but a larger number and wider variety of new customers, thereby reopening the boundaries of growth.
Among these considerations, the most critical one is the leverage brought by technology reuse.
The R&D of an automotive-grade product often takes several years and requires an investment of hundreds of millions, along with repeated validations for safety and reliability. If this capability can only serve the automotive industry, the huge R&D costs can only be slowly amortized through vehicle sales. However, once it can be transferred to robotics, data centers, and low-altitude applications, the same technologies, code, and production lines can be repeatedly monetized across multiple markets, amortizing R&D costs, shortening cycles, and making it easier to improve yield rates.
Specifically in terms of products, the underlying capabilities of BMS, thermal management, sensors, and domain control can indeed be migrated among power batteries, energy storage cabinets, robots, aircraft, and servers. The overlap of technology stacks and supply chains between intelligent vehicles and embodied robots is generally around 60% to 70%. However, while underlying technologies can be reused, the products themselves cannot be directly copied.
In Aptiv's view, this is not a simple extension of business boundaries, but the migration and reuse of perception, computing, and connectivity capabilities validated by the stringent standards of the automotive industry into new industrial scenarios.
Ultimately, the new industries do not lack teams that can make demonstration samples; what they lack are those who can stably produce products in large volumes at low costs. And this is precisely the core competency honed by the automotive-grade system over decades.
There is another layer of consideration that is easily overlooked: hedging against cycles.
The automotive industry is a strongly cyclical one, following consumer confidence, purchase tax policies, and price wars. Data centers are driven by AI capital expenditures, robots by penetration rates and labor costs, energy storage by the pace of power construction, and the low-altitude sector by the progress of airworthiness certification. These forces are not synchronized. Betting on several tracks simultaneously is equivalent to using a multi-track portfolio to smooth revenue fluctuations; when one leg steps into the void, there are other legs to support it.
As for the capital market's willingness to give higher valuations to new businesses, many companies see it more as a byproduct after the aforementioned judgments hold true, rather than the original intention of opening up a second growth curve.
However, no matter how sufficient the reasoning is, it does not mean that new businesses will definitely make money.
Hot Concepts and Cold Revenues: Years Apart
The vision is ambitious, but the financial reports are stark.
The most illustrative aspect is the revenue structure. In the first half of 2026, Tuopu Group delivered robot actuators in small batches, generating a revenue of CNY 14.0477 million, a year-on-year increase of 83%. The numbers look good, but compared to its total revenue of CNY 14.199 billion in the same period, the proportion of this new business is only about 0.1%.
Sanhua Intelligent Controls has been strategically deployed in robot actuators for a long time, but the media bluntly stated in the interim report observation: robots have not yet brought much revenue contribution.
Guangyang Co., Ltd., which started with bearings, generated a robot business revenue of CNY 9.9941 million in the first half of the year, with a gross profit margin of negative 22.56%, indicating that the new business is still in the early investment stage.
Yang Xiaoming was also very candid at GNEV 2026. In his view, the implementation of embodied robots still faces many challenges at this stage, but it represents the future development direction. From dancing on the stage of the Spring Festival Gala to actually working in factories, the distance in between is longer than imagined.
This means that for most companies, these new businesses are more like a long-term option in the short term, rather than an immediate meal ticket.
However, the value of crossing over cannot be denied because of this. Component companies do hold hard currency in their hands: automotive-grade reliability requirements, yield control, cost control capabilities, and large-scale system integration experience. These are the capabilities that new industries such as robotics, low-altitude, and data centers lack the most and are the hardest to develop quickly.
However, automotive-grade capabilities are not a master key, and moving them over intact might also become a burden. The automotive industry is willing to pay for extreme working conditions for safety redundancy, while robotics and energy storage customers value cost-effectiveness and iteration speed more. Deciding what can be reused and what must be redone tests the trade-offs of enterprises.
Specifically regarding implementation, standards and certifications are the first unavoidable hurdle. Airworthiness certification for low-altitude aircraft takes years, requiring safety to be proven item by item; energy storage must pass grid connection and safety certifications; robots must pass system certifications such as functional safety; and data centers must also win long-term and stringent validations from power and operator customers. None of these can be directly covered by the automotive-grade system. General technologies may solve most problems, but the small part that truly determines life and death must be relearned.
The business model is another hurdle. After securing a design win in the automotive industry, there is a relatively stable and predictable long-cycle supply. Although there are annual price reductions, the volume and pace are very clear. In new industries, some are one-time engineering projects, and some are fragmented channel orders. The payment collection cycle, pricing methods, and customer relationships are completely different. Tier 1 suppliers, accustomed to dealing with OEMs, need to relearn how to do business with data center operators, robotics companies, and power customers.
A more realistic challenge is people and culture. The automotive industry emphasizes processes, standards, and zero defects, while new industries emphasize rapid iteration and a willingness to make mistakes. The two paces are not naturally in sync. Teams accustomed to the vehicle development cycle may not be able to keep up with the monthly or even weekly iterations of robotics and data center customers. To make up for this lesson, relying solely on internal transformation within the enterprise is often not enough; strategic cooperation and investment mergers and acquisitions are also needed to introduce new capabilities and genes from the outside.
Precisely because of this, Zhang Yongwei specifically reminded that deploying in new tracks does not equal blind cross-boundary expansion. In terms of difficulty, tracks like energy storage and industrial thermal management, which are closest to the automotive industry and have the most mature certifications, are the easiest to run through. Fields like low-altitude, which have high certification thresholds and undefined business models, require more patience. For small and medium-sized enterprises without the advantage of technological homology, this feast may not have an entry ticket for them from the very beginning.
Ultimately, the second growth curve is not about simply copying automotive capabilities, but translating the manufacturing and system capabilities accumulated over many years for new industries to understand, while acknowledging one's identity as a novice in the new field. Those who can truly succeed are the companies that can take away the hard skills of automotive-grade standards and are willing to use cooperation and mergers and acquisitions to make up for their shortcomings; those who only want to ride the hype and tell stories will eventually be exposed in front of the revenue statements.
The story of these companies has just begun. In the upcoming "Transformation of the Automotive Supply Chain," we will first step into the most bustling track: when automotive thermal management companies dive headfirst into data centers, what exactly do they see, and what will they collide with?