
By Liu Xinyi
Edited by Mao Shiyang
On July 8, Momenta listed on the main board of the Hong Kong Stock Exchange, stock code 6880.HK.
The offering price was 295.6 HKD per share. It surged over 6% at the opening, then the gain narrowed in the afternoon session, finally closing at 295.6 HKD. The total market value was approximately 69.6 billion HKD. This was the company with the largest market cap among the five companies ringing the opening bell at the HKEX that day.
More lively than the stock price was social media. On listing day, official accounts of nearly 20 vehicle partners simultaneously posted congratulatory articles and posters. The topic was unified as "Embracing the Future of Physical AI Together".
IM Motors, Mercedes-Benz, Toyota, and GM appeared together, described as "Half of the Global Automotive Industry's Social Circle".
This social media storm was not hastily arranged. Momenta was founded in 2016. Cao Xudong graduated from Tsinghua University and worked successively at Microsoft Research Asia and SenseTime. In ten years, it completed 20 financing rounds. The cost per share in Series A was $0.45, which has multiplied more than 80 times to today.
Behind it lies a rare list. SAIC Motor, holding about 20%, is the largest external shareholder. GM, Toyota, Mercedes-Benz, and Bosch are also listed. These entities are both investors and customers. Car manufacturers invest money while entrusting models to it. This "Customers as Shareholders" loop is the backdrop of the good news.
For an autonomous driving supplier, this scene itself is a narrative. It wants to tell the market that its customer list is the top tier of the global automotive industry.
This claim is supported by data. According to the Blue Book released by Frost & Sullivan in June, from March 2025 to February 2026, in the Chinese third-party urban NOA supplier market, Momenta's sales market share reached 65%, ranking first in the industry; among the top 10 global car manufacturers, nine cooperate with it.
The market enthusiasm was also high. The public offering received an oversubscription of about 414 times. The international offering received about 44 times before deducting anchor investors and greenshoe. Institutional orders exceeded 100 billion HKD. Unusually, this issuance did not set a price range, directly fixing the offering price. 14 anchor investors subscribed to nearly half the share.
In the list, the Singapore Government Investment Corporation (GIC), known for being cautious and rarely investing as an anchor in new shares, and Fidelity International each led with heavy investments of about USD 100 million. Franklin Templeton appeared as a Hong Kong anchor for the first time in history. More notably, Mercedes-Benz and BYD, two old shareholders and customers of Momenta, again bet as anchors at the key IPO node.
Cao Xudong said it took Mercedes-Benz exactly 8 years from the 2017 investment to mass production at the end of 2025. Once the breakthrough occurred, in 2024 Momenta obtained all of Mercedes-Benz's electric and fuel vehicle business.
The market is willing to pay this price because of the hardest data in the prospectus. From 2023 to 2025, Momenta's revenue increased from RMB 743 million to RMB 2.413 billion, sales costs remained almost unchanged, and gross margin rose from 17.5% to 71.6%.
Supporting this curve is licensing revenue. It multiplied 42 times in three years, rising from 3.1% to 40.1%. Licensing fees are charged per vehicle, with marginal costs close to zero. The higher the proportion, the higher the gross profit.
This means Momenta is transforming from a company that earns development fees based on projects to a company that collects licensing fees based on installation volume. A net loss of RMB 3.458 billion on the books looks scary, but RMB 2.843 billion of it is the change in fair value of preferred shares, a non-cash item that will disappear after listing conversion; stripping it out, adjusted loss narrowed from RMB 1.09 billion in 2023 to RMB 300 million in 2025, approaching break-even, and net cash outflow from operating activities dropped from RMB 1.069 billion to RMB 281 million.
Momenta's financial fundamentals are now those of a stable supplier.

Behind this curve lies the compound interest of engineering capabilities. It took Momenta two years to deliver the first 100,000 units; now it can complete 100,000 units in less than 40 days at the fastest. Model adaptation that required 400 people working for two years in 2022 can now be delivered by 10 to 50 people in three months.
Once the platform is standardized, new models no longer start from scratch. R&D personnel increased almost negligibly, and per capita output tripled in three years. This is also the reason it dares to say "The scale effect is stronger than chips". The marginal cost of software is zero; the more vehicles equipped, the thinner the cost spread, and the better the experience.
In terms of narrative, Momenta has also left rich imagination for the market. It is transforming from a "Smart Driving Tier 1" to a "Physical AI Company". With World Models as the foundation, it frames autonomous driving as the first scenario where Physical AI lands, extending to unmanned logistics, autonomous trucks, and even robots in the future.
Cao Xudong judges that eventually, this industry will leave only 2 to 3 companies in China and 3 to 4 globally, because autonomous driving is software with zero marginal cost, and scale effects are even stronger than chips.
This narrative has its basis. Autonomous driving is indeed one of the few physical AI scenarios that simultaneously possess real data, clear goals, scaled loops, and commercial revenue.
Viewed broadly, this is the first time a Chinese autonomous driving "third-party solution" has entered the global capital market with a volume of nearly USD 10 billion. Its pricing will become a reference for the entire industry chain.
Cao Xudong summarizes the core of Physical AI into two things: Data Scaling and Business Scaling, and the two feed each other positively. Autonomous driving is currently the only physical AI scenario that successfully implements both.
Momenta spun this flywheel with 900,000 mass-produced vehicles. What remains is time. It took AI ten years to climb to the GPT moment. Physical AI data is harder to obtain and the cycle is longer, but he believes that once it approaches human levels, transcendence will happen within one or two years.
The price the market gives today buys this time gap. Momenta has already proven itself to be a good business. What remains to be proven is whether it can convert the automotive smart driving supplier into a real Physical AI before the window closes.