Hot money is starting to pour into robot brains.

Just a couple of days ago, the gong sounded at the Hong Kong Stock Exchange for the "First Stock of Robot Brains" as SEER Robotics went public, surging over 38% at one point during its debut. What stood out were the numbers from the IPO phase: the public offering was nearly 6,000 times oversubscribed, with a one-lot allocation rate of just 5%. Eight cornerstone investors, including Hillhouse and Yuanbao Family Office, collectively poured in HK$462 million.

A company established just six years ago, with annual revenue of RMB 442 million and still in the red, has managed to garner such enthusiasm in the capital market. The core selling point isn't the robot itself but the controller known as the "robot brain." Data from CIC Consulting shows that by sales volume, SEER Robotics' robot controllers have captured 24.8% of the global market and 45.2% of the Chinese market, both ranking first.

SEER Robotics' IPO comes at a pivotal moment for the industry.

Over the past two years, the embodied AI sector has primarily focused on humanoid robot bodies—how steadily they walk, how flexibly they move, and how closely they resemble humans have always been the focal points of public discussion. But as 2026 begins, the winds have shifted.

Capital is now bypassing the "body" and heading straight for the "brain." Statistics from QuantumBit show that in the first half of 2026, domestic financing in the embodied AI sector totaled approximately RMB 43.8 billion, with over half flowing to companies focused on the "brain," while body manufacturers received less than 20%.

This marks a shift in the industrial value focus as embodied AI moves from concept validation to large-scale deployment. As hardware bodies mature and supply chains stabilize, the core variables determining what robots can do and how well they do it are shifting from mechanical structures to intelligent systems.

01

Bypassing the Body, Heading Straight for the Brain

Rewind to 2024, when humanoid robot companies were the hottest trend in the embodied AI sector.

UBTECH went public in Hong Kong, while companies like Zhipu, Unitree, and Fourier rolled out new products one after another, each launch drawing collective attention from the tech community. The narrative was straightforward: embodied AI is the next-generation computing platform, and humanoid robots are its ultimate form. Whoever mass-produces a body first gains the upper hand.

Yet, in just over a year, capital attention has shifted.

The financing structure in the first half of 2026 tells the story. Companies focused on the "brain" secured half of the funds, while full-stack, core component, and body-focused firms found themselves off-center. Viewed differently, companies related to the "brain" collectively took nearly 70% of the financing, while pure hardware body companies were pushed to the margins.

The pace of financing is also accelerating. Many brain companies complete a funding round every month, with the fastest two securing back-to-back rounds just two weeks apart. Such speed is uncommon in the hard tech sector, typically seen only in internet and pure software industries with such dense (intense) capital injections.

The logic behind this isn't complex. After two years of explosive growth, the technical barriers for robot bodies are rapidly declining. Unitree Technology announced a price drop for its bipedal humanoid robot Unitree R1 from RMB 39,900 to RMB 29,900, with immediate availability.

Previously, humanoid robots were generally seen as products costing tens of thousands of yuan. The price drop reflects supply chain maturity and the start of mass production—core hardware costs for joint modules, motors, and reducers are all falling rapidly, showcasing the advantages of Chinese manufacturing.

When the "body" is no longer scarce, the value of the "brain" stands out.

The reasoning is simple: the same robot hardware, equipped with different brains, can perform vastly different tasks. Low-end controllers allow robots to repeat preset movements, while high-end intelligent controllers enable them to perceive environmental changes, plan paths autonomously, coordinate multi-robot operations, and even understand natural language instructions through large models. The former sells hardware; the latter sells capabilities.

Gross margin differences best reflect this value stratification. SEER Robotics' prospectus shows a 79.8% gross margin for its controller business and 89.3% for software, compared to just 38.4% for complete robots and 15.7% for accessories. An 80% gross margin is rare in the hardware-dominated robotics industry, closer to the profitability levels of pure software companies.

This is why capital is chasing brain companies—compared to the asset-heavy, low-margin, long-cycle hardware bodies, controllers and intelligent systems have lower marginal costs, stronger scaling effects, and once an ecological barrier is established, profit potential becomes substantial.

Of course, the term "brain" now carries far richer connotations than a few years ago.

Early robot controllers were essentially motion control boards, directing motor movements and coordinating joint actions, with technical barriers mainly in real-time performance and stability. But today's "robot brain" has evolved into an intelligent system integrating perception, decision-making, and control—incorporating SLAM localization and navigation, visual semantic recognition, reinforcement learning, multi-robot scheduling, and even integrating large language models and world models.

SEER Robotics' prospectus shows that in the global robot controller market, suppliers of independent controllers increased their supply from 6,000 units in 2021 to nearly 50,000 units in 2025, with an expected surge to over 300,000 units by 2030. By revenue, the global robot controller market grew from RMB 700 million in 2021 to RMB 2.4 billion in 2025, projected to reach RMB 8.4 billion by 2030, with a CAGR of 28.8% from 2026 to 2030.

A professional industry report also states that in 2025, the robot brain controller market reached RMB 2.236 billion, while the robot motion control system market hit RMB 6.073 billion. When including AI-powered next-gen intelligent controllers and matching (supporting) software, algorithms, and cloud services, the market space becomes much larger.

A report by UK-based Futures Market Insights projects that the global physical AI market will grow from approximately $383 billion in 2026 to $3.26 trillion by 2040.

Physical AI, in essence, equips various physical entities with intelligent brains. Robots are just one of its most typical forms.

02

Cloud Providers Jump In to Become Robot Brain Suppliers

If the financing frenzy among startups is merely a sector-internal signal, the collective entry of internet giants and cloud providers means the robot brain war has escalated to the ecological level.

This year, Huawei, Tencent, Baidu, and Alibaba have all launched their embodied AI platform products, and all have chosen the "no body, just brain" route—not building robots themselves but providing intelligent systems, development tools, and cloud service infrastructure for robot manufacturers.

Huawei has moved the fastest and most decisively. At this month's Huawei Cloud INSPIRE Innovators Summit, Huawei Cloud officially unveiled CloudRobo, an embodied AI development platform positioned as the "world's first end-to-end one-stop embodied AI development platform." According to official statements, the platform covers the full chain from data synthesis, model development, simulation verification, to cloud-edge-device deployment, with built-in millions of data assets and over 20 Ascend-compatible models, enabling "robot cloud integration in hours and model deployment in minutes."

Huawei's strategy is clear: leverage cloud capabilities to lower the barrier to robot intelligence. Previously, robot manufacturers had to develop algorithms, train models, and build simulation environments themselves—costly and time-consuming. Now, with Huawei Cloud's platform, everything from data to training to deployment can be done in one place, allowing robot manufacturers to focus solely on hardware bodies and scenario deployment.

On the launch day, over 20 companies, including Youibot and Huayan Robotics, announced their initial adoption of the CloudRobo platform, with Huawei simultaneously launching the "Hundred Models, Thousand Forms: Cloud Convergence for Win-Win" ecological cooperation program. Combined with Huawei's layout (layout) in Ascend chips, industrial software, and 5G networks, CloudRobo is essentially building an "end-edge-cloud" collaborative embodied AI infrastructure.

Tencent has chosen a lighter entry point. Tencent Robotics X Laboratory made its debut with a complete technical matrix, launching the Tairos embodied AI open platform, open-sourcing the Hunyuan Embodied Large Model HY-Embodied series, and showcasing robot body interconnection technology RoboFusion.

Zhu Yajuan, Tairos product ecosystem lead at Tencent Robotics X, stated on-site that Tencent positions itself as an "indispensable titanium screw" in the robotics industry—not building robot bodies but focusing on software and cloud services. This phrasing is interesting—it acknowledges the boundary of not doing hardware while emphasizing core value at the system level.

Specifically, the Hunyuan Embodied Large Model addresses robot "understanding and thinking," enabling them to interpret environments, follow instructions, and plan tasks autonomously; the Tairos platform improves development efficiency with standardized toolchains; and RoboFusion enables interoperability among different robots. Together, they form Tencent's robot brain solution.

Baidu follows a "data + model + infrastructure" approach. At May's Create 2026 Baidu AI Developer Conference, Baidu Smart Cloud clearly state (made it clear) it would increase investment in AI infrastructure, scenario linkage, and industry standard-setting. Prior to that, in April, Baidu had already launched the "Embodied AI Data Supermarket" with multiple robot companies, establishing a hierarchical data tagging system.

At the investment level, Baidu is also deeply involved. Its RMB 1 billion Series B in Zhipingfang and RMB 700 million Series A in the Beijing Humanoid Robot Innovation Center both feature Baidu's backing. Baidu's logic is to use its Wenxin Large Model capabilities to bind hardware companies, completing the AI loop from virtual to physical worlds.

Alibaba DAMO Academy released the RynnBrain embodied foundation model in February, open-sourcing seven models at once, including the industry's first 30B MoE architecture embodied model. Official materials state that this model, for the first time, equips robots with spatiotemporal memory and spatial reasoning abilities, setting new SOTA records on 16 embodied benchmarks and surpassing Google's Gemini Robotics ER 1.5.

ByteDance, while not launching a dedicated embodied AI platform, has made world models its top priority for 2026. At the Volcano Engine FORCE Conference on June 23, ByteDance mentioned that its Seedance video generation model could be applied to data synthesis in embodied AI. Given ByteDance's technical accumulation in multimodal large models and video generation, along with Volcano Engine's cloud service capabilities, its entry into the embodied AI brain track (sector) seems inevitable.

The collective choice by big tech to "do brains, not bodies" is fundamentally an economic calculation.

Hardware bodies, while tangible, are asset-heavy, low-margin, and supply-chain-intensive, likely ending up in price wars—a scenario repeatedly validated in the smartphone industry. Brain and platform layers, while requiring heavy upfront R&D investment, see rapidly declining marginal costs once an ecosystem is established, enabling Continuous charging (recurring revenue) through cloud services.

More importantly, cloud providers have natural advantages in building robot brains. Training embodied AI requires massive compute power, large-scale simulation environments, and multimodal large model foundations—all core strengths of cloud providers. Building an AI infrastructure from scratch is neither economical nor practical for robot body manufacturers.

From this perspective, the future embodied AI industry may resemble smartphones' layered structure: at the bottom are chips and compute power, in the middle are operating systems and intelligent brains, and at the top are various robot hardware forms and scenario applications. Cloud providers are vying for that critical middle layer.

03

Why Can't High-Margin Brains Support Revenue?

While the brain narrative is compelling, commercial realities persist.

SEER Robotics' prospectus offers much to ponder. On one hand, the 79.8% gross margin for controllers and 89.3% for software prove the commercial value of robot brains; on the other, controller revenue was just RMB 85 million in 2025, accounting for 19.3% of total revenue, while low-margin complete robots brought in RMB 300 million, or 67.9%.

In other words, the most profitable business is the smallest, while the revenue driver has low margins. This dilemma is shared by nearly all third-party controller manufacturers.

The reason is straightforward. Currently, robot manufacturers, especially industrial robot firms, prefer complete solutions over standalone controllers. On one hand, complete deliveries enable rapid deployment and immediate use, sparing customers integration efforts; on the other, controllers as core components require strong secondary development capabilities from customers, a high barrier for many traditional manufacturing clients.

Thus, controller manufacturers often need to bundle controllers with complete robots to make sales—first install controllers in their own robots, then sell to end customers. The brain is excellent but must attach to a body to sell.

Xu Zhaoyun, a partner at Lihan Capital, analyzed for the media that domestic humanoid robot manufacturers are generally in the early stages of mass production scaling, with a core focus on controlling hardware BOM costs. Third-party controller manufacturers can only bind clients through low-price, high-volume strategies, with software value bundled and discounted by the hardware carrier. This creates an awkward situation: controllers have high technical barriers and margins but struggle to sell alone; complete robots have lower barriers and margins but dominate revenue.

SEER Robotics hasn't escaped this pattern.

From 2023 to 2025, revenue grew from RMB 249 million to RMB 442 million, a three-year CAGR of 33.2%, but net losses were RMB 47.7 million, RMB 42.3 million, and RMB 47.07 million, respectively, totaling approximately RMB 137 million over three years. Adjusted net losses narrowed—from RMB 20.91 million in 2023 to RMB 2.87 million in 2025—but true profitability remains distant.

This isn't unique to SEER Robotics but a stage-specific industry trait.

Embodied AI is still in early deployment, with clients more willing to pay for "visible, tangible" hardware while software, algorithms, and intelligent systems lack established willingness to pay. Much like early smartphones, where consumers paid for hardware first and software/service value emerged only after the mobile internet ecosystem matured.

Another challenge comes from big tech's in-house R&D. As Huawei, Tencent, Baidu, and other cloud providers launch their embodied AI platforms, will third-party independent controller manufacturers see their living space (living space) squeezed?

Currently, their positioning differs. Cloud platforms lean toward general-purpose large models, training tools, and cloud-based scheduling, while companies like SEER Robotics focus on bottom layer (low-level) real-time motion control and field execution—complementary rather than substitutive. But whether boundaries blur long-term remains to be seen.

In the international market, the traditional "Big Four" industrial robot companies—Fanuc, ABB, Yaskawa, and KUKA—have all developed their controllers and robot bodies in an integrated manner. They never sell controllers separately, keeping their core technologies firmly in-house. This is also a key reason they can maintain high profit margins. For Chinese robot manufacturers to move upmarket, they will most likely end up developing their own controllers.

So, the real opportunity for independent third-party controller manufacturers may not lie in selling to leading robot body manufacturers, but in serving the vast number of small and medium-sized customers and long-tail scenarios.

SEER's "Nebula" open system comes preloaded with over 1,000 robot models, covering forklifts, bin-picking robots, wheeled humanoids, robot dogs, and other categories, allowing customers to customize them modularly. This "LEGO-style" approach is highly attractive to SMEs that don't want to start from scratch but need rapid deployment.

By the end of 2025, SEER had served over 2,000 customers, with products sold in 35 countries and regions. Its end customers include Tesla, BYD, Foxconn, Siemens, and other enterprises. With a sufficiently large customer base and broad industry coverage, this forms the foundation for independent controller manufacturers to establish ecological barriers.

SEER's IPO is a landmark event. It signifies that capital markets are beginning to recognize the independent value of the "robot brain" and that the industrial division of labor in embodied AI is becoming more refined.

In the past few years, the industry's focus has always been on whether robots "can do it"—can they walk, run, or grasp objects. Now, these fundamental questions are gradually being answered, and the industry is entering the second half, which focuses on "how smart robots can be." What will determine the outcome in this phase is no longer who has the more sophisticated mechanical structure but whose brain is more intelligent, whose system is more open, and whose ecosystem is more vibrant.

Capital is flowing toward brains, major companies are building platforms, and startups are diving deep into vertical scenarios—these three trends are advancing in tandem. In the short term, hardware robot bodies will remain the main revenue driver, with the value of the "brain" Mainly reflected in technological barriers and differentiated competitiveness ; but in the long run, when robot hardware becomes commoditized, intelligent systems will be the true profit center and ecological gateway.

Of course, it's still too early to talk about the final outcome. Embodied AI is still in a very early stage. According to Omdia, global shipments of general-purpose embodied AI robots were only 3,000 units in 2024 and are not expected to reach 2.6 million until 2035. There's still a long road ahead, from 3,000 to 2.6 million.

But the direction is already clear. In the second half of embodied AI, the brain will call the shots.

— END —