When you finish watching the thirtieth short video in the car, the car has already driven thirty kilometers for you. This used to happen only in science fiction movies. Now someone has turned it into a business and even listed it.

Momenta, a company that sells "brains"

On July 8, 2026, Momenta was listed on the Hong Kong Stock Exchange, opening at HKD 314.8 and closing back at the issue price of HKD 295.6.

The high opening with a market value of HKD 70 billion and the subsequent drop outlines the market's complex mindset towards this "first stock of physical AI": willing to pay a premium for the narrative, yet reassessing the value of that premium within a single trading day.

What does Momenta do? It equips cars with "brains" and provides intelligent driving software solutions.

△Image source: Visual China

It does not manufacture radars, nor does it produce chips; it only focuses on algorithms. The advantage of this model is its extremely high gross margin. According to the prospectus data, the company's gross margin reached 71.6% in 2025, which is extremely rare in the automotive industry chain. Revenue tripled over three years, increasing from CNY 743 million in 2023 to CNY 2.413 billion in 2025, with an average annual compound growth rate exceeding 80%.

Another set of data is equally striking. From 2023 to 2025, the company's net profit attributable to shareholders was a loss of CNY 2.57 billion, CNY 3.206 billion, and CNY 3.458 billion, respectively, with a cumulative loss of nearly CNY 10 billion over three years. The losses mainly stemmed from changes in the fair value of preferred shares and high-intensity R&D investments, with a cumulative R&D expenditure of CNY 4.66 billion over the past three years. Founder Cao Xudong admitted after the listing: "This year may still see strategic losses, break-even next year, and profit the year after."

After the first day of flat closing, Huang Lichong, president of Huisheng International Capital, commented: "The overall performance is relatively flat, indicating that the capital market is not blindly chasing the physical AI story." Yang Delong, chief economist at Qianhai Kaiyuan, also pointed out that while funds recognize the scarcity of the track, there is a clear divergence regarding the company's high valuation and the timeline for realizing future narratives.

Physical AI is not the same as digital AI

It was often said that AI would disrupt the digital world, but now it is realized that the real revolution is that it is beginning to transform the physical world.

Unlike generative AI that processes text and images, physical AI requires systems to understand real-world rules such as gravity, collision, and spatial causality, forming a complete closed loop of "perception—decision—execution—feedback." In simple terms: digital AI excels at online creation, while physical AI excels at manipulating physical devices.

Implementing physical AI involves three steps: first, training models in a virtual environment, allowing AI to learn physical laws in a simulated world; second, using perception hardware such as LiDAR and sensors to "see" the real world; and finally, outputting decision instructions to execution terminals like cars and robots to make them move. Momenta is stuck in the middle stage. The R7 world model released in April this year learns physical laws from over 12 billion kilometers of real driving data, repeatedly simulating in the virtual world, and continuously optimizing decisions through reinforcement learning. The SAIC Volkswagen ID.ERA 9X has already been equipped with it As of the pre-IPO stage, the production vehicles equipped with the Momenta system have surpassed 1 million units, with over 210 models accumulated. Partnering automakers include Mercedes-Benz, BMW, Audi, Toyota, and BYD, with nine out of the top ten global automakers as its clients. In the Chinese third-party city NOA market, Momenta holds a market share of 65%, meaning that for every three vehicles equipped with third-party city NOA, two have its system installed. This technology is already on the road.

Cao Xudong stated that the past decade has been about teaching AI to drive, while the next decade aims to provide every household with robotic services such as "drivers, maids, doctors, and teachers." The underlying technology for autonomous driving and robotics is interconnected, both requiring an understanding of the laws governing the physical world; the only difference is that the carrier shifts from four wheels to bipedal or robotic arms. Following this logic, Momenta is extending its capabilities accumulated in autonomous driving into the robotics field.

However, the "next decade" timeframe seems somewhat lengthy for a publicly listed company that must present quarterly results to the market. The difficulty of technology transfer from autonomous driving to household robots and the commercialization cycle are both significant unknowns. For ordinary investors, Momenta provides a window to observe the physical AI track, but it also comes with a high barrier to entry.

The transition of AI from the digital world to the physical world is a major trend in industrial evolution.

In the automotive supply chain, compared to CATL and Hesai Technology, the roles of the three companies are clear: CATL provides the "heart" (batteries), Hesai Technology provides the "eyes" (LiDAR), and Momenta provides the "brain" (algorithms). The first two sell hardware, while Momenta sells software—where the marginal cost is nearly zero, and the fundamental reason for its gross margin of 71.6% lies here. All three have established overwhelming market positions in their respective tracks, following the same logic where the winner takes all after full competition.

Looking back at the reshuffling of the automotive industry over the past decade, survivors share a common set of rules: maintain clarity in a frenzied market, avoid chasing short-term trends, and put in "hard work" during long R&D cycles. When Momenta was founded, the industry was betting on L4, which allows systems to drive within a specific range without human operation. Cao Xudong chose a "less sexy" L2 mass production route: L2 is where humans drive, and the system occasionally lends a hand.

These companies also share a common point: all core technologies are self-developed, without relying on overseas supplies. Momenta's world model, CATL's battery materials, and Hesai's radar chips are all self-manufactured. Technological independence is a prerequisite for gaining the trust of global automakers and sovereign capital.

NVIDIA CEO Jensen Huang asserted at CES 2026 that the value of manufacturing and logistics industries that physical AI can reshape will reach $50 trillion. Whether this figure is accurate remains to be seen, but the transition of AI from the digital world to the physical world is a major trend in industrial evolution. Momenta's IPO is the first formal pricing of this trend in the capital market.

Cao Xudong expressed hope to "write the Silicon Valley legend of the East" together with all Chinese AI companies. Legends are never written in a day. Momenta has secured its entry ticket, but the real competition only begins from the moment the bell rings In the past, cars brought people closer together. Next, physical AI will redefine "on the road."

Xiaoxiang Morning Post reporter Hu Qiuxian Comprehensive report

Reporting and rights protection channel: Download the "Morning Video" client from the app market, search for "report" for one-click access, or call the hotline 0731-85571188; for cooperation, please call the business cooperation contact number 0731-85572288