The Soochow Huang Xili team has given Momenta (06880) a "Buy" rating: Data + business closed loop leads to physical AI.

Recently, the Soochow Huang Xili team has initiated coverage of MOMENTA-W (06880) for the first time and given it a "Buy" rating. The research report points out that Momenta uses the R7 world model as the physical AI foundation, driving model iteration through a closed loop of mass-produced vehicle data and L4 operation. This constructs a physical AI platform that can be reused across different vehicle models, OEMs, and scenarios. The valuation significance of Momenta's world model should not be merely understood as an "upgrade of smart driving experience," but rather as the embryonic form of physical AI infrastructure. Traditional project-based smart driving delivery heavily relies on engineering adaptation, with human and marginal costs increasing as the number of projects rises; if the R7 world model becomes a unified foundation, then algorithm capabilities can be reused across different vehicle models, chips, and scenarios, meaning the addition of new OEMs and scenarios manifests more as adaptation costs rather than the need to rebuild an entire system. Mass-produced smart driving is reaching an outsourcing turning point, solidifying its leading position. The pattern of car companies relying on "self-developed 40% + outsourced 60%" is taking shape, highlighting the value of independent third-party suppliers. The report indicates that by 2026, the pace of mass production and targeted expansion will further accelerate. Momenta has become the "greatest common divisor" for mainstream car manufacturers in Japan, Europe, Korea, and the United States, maintaining first place in China's third-party urban NOA market with approximately 65% market share. The "data flywheel" effect is driving a qualitative shift towards a scalable software licensing platform, rapidly reducing marginal delivery costs, with R7 deployment entering a phase of scalability in the second half of 2026. Additionally, the company's urban NOA penetration rate is expected to rise quickly, with domestic L3 smart driving commercialization anticipated to formally land after 2027. Scalable Robo: The second growth curve driven by the world model foundation platform The Scalable Robo strategy supports the large-scale deployment of passenger cars, Robovans, Robotaxis, and Robotrucks using a single R7 large model foundation. Data collected across different scenarios flows back into a unified training system, achieving collaborative evolution of capabilities. This platform attribute is the fundamental reason for the company's valuation distinguishing it from single smart driving software suppliers: each newly operational scenario represents the reuse and amplification of the same foundational capabilities rather than repetitive investment, with decreasing marginal costs and increasing scenario value. Robovan: For the first time, a no-map solution has been introduced into the unmanned delivery industry, operating in Suzhou for several months and conducting regular operational verification in scenarios such as JD LOGISTICS parks, while exploring cooperation with express logistics clients. Robotaxi: Leveraging the "L2 data feeding back to L4" mechanism to create generational differences, testing or operations have begun in four domestic cities, as well as in Abu Dhabi and Munich, Germany. Plans are in place to deploy several hundred unmanned taxis domestically and internationally by the end of 2026, obtaining licenses in more than 10 cities worldwide. The Huang Xili team believes that the transfer of the mature R7 world model capabilities validated on the mass production side to L4 scenarios constitutes the most direct realization of the "L2 feeding back to L4" mechanism to date, and is the fundamental reason for the generation gap in iteration efficiency compared to pure L4 players. Moreover, strategic partnerships have been established successively with the world's largest mobility platform, Uber, Southeast Asia's leading mobility provider, Grab, SAIC's Enjoy Travel, and the UAE state-owned taxi company Lumo, creating a four-dimensional layout across "Europe and the United States, China, the Middle East, and Southeast Asia." This open ecosystem model is defined as "a key piece of the puzzle in achieving global scalable autonomous driving"where algorithms and vehicle manufacturers cooperate to provide vehicles, mobility platforms provide passenger flow, and Momenta provides the autonomous driving brain, forming a clear division of labor in the industry chain. The company also plans to expand into the Robotruck field by 2027, with projections that the Robotruck market will grow to approximately $33 billion globally and $16.5 billion in China by 2030. Additionally, the sister company New Chip Navigation is assisting the company in moving towards an integrated hardware-software model. The collaboration between algorithms and chips is accelerating the dissemination of the world model to mainstream price bands.