01 Trigger Event
In March 2026, SAIC Group's board approved a proposal to establish an intelligent-driving joint venture with registered capital of RMB 300 million.
The ownership structure is clear: Momenta subscribed for RMB 153 million and holds 51%; SAIC Group subscribed for RMB 75 million and holds 25%; and Yandi Technology subscribed for RMB 72 million and holds 24%. Together, SAIC Group and Yandi Technology retain a combined 49% stake.
This is the heart of the matter: SAIC did not make the intelligent-driving JV a wholly owned R&D project. Instead, it placed Momenta in control.
02 The Real Significance
On the surface, this is a capital partnership between SAIC and Momenta.
At a deeper level, SAIC is partially delegating the product definition, software iteration, and engineering cadence of autonomous driving to an algorithm company while retaining 49% of the economic interest.
This reflects a division of labor already emerging in the automotive industry: OEMs control brands, distribution, supply chains, vehicle platforms, and safety responsibility; algorithm companies handle perception, planning, end-to-end models, data loops, and software updates.
The issue is not who can write a driving model, but who can turn models into automotive products that can be delivered at scale, iterated continuously, and held accountable for safety.
The intelligent-driving capabilities of traditional automakers are often constrained by organizational silos. Vehicle development, supply chain, engineering validation, and software release remain disconnected, making it difficult for algorithm iterations to move at internet-product speed. Momenta's core asset may not be its models alone, but rather the engineering system connecting data to mass-produced vehicles.
Therefore, 51% means more than receiving 2 million additional shares. It gives Momenta control for consolidated financial reporting purposes, while SAIC captures long-term returns approaching half of the venture.
What will ultimately be valued is mass-production capability and the data loop, not the visual impact of a one-time demo.
For AI builders, this offers an important reminder: entering the automotive industry requires more than technical capability. Delivery cycles, the Safety case, hardware adaptation, OTA updates, supplier management, and after-sales responsibility will all determine commercialization speed.
I have not reviewed the product scope or governance agreement for the joint venture from the inside. Based solely on the ownership structure, it would be premature to conclude that SAIC has fully ceded control of its intelligent-driving strategy.
03 Historical Analogy / Structural Comparison
The arrangement resembles the restructuring of the smartphone industry around 2010: hardware companies continued to control manufacturing, distribution, and supply chains, while operating systems and core software came under the control of specialist companies.
BlackBerry and Nokia chose closed models and maintained their own ecosystems. Apple, by contrast, used iOS to control critical user experiences and enabled third-party developers to build an application economy around it.
The SAIC-Momenta combination falls somewhere between these approaches. SAIC and Yandi Technology jointly retain 49%, preserving substantial economic interests and industrial synergy, while Momenta's 51% stake gives it the organizational position required to advance product decisions.
There is also a structural parallel to the changing role of AWS after its emergence in 2006. IT capabilities gradually migrated from companies' internal data centers to externally delivered platforms capable of scaling. Likewise, automotive intelligent driving could evolve from an internal automaker project into a platform capability defined by a specialist software organization.
Yet cars have one critical difference from phones: a phone failure usually means a system crash, whereas an automotive failure can involve physical safety. Control cannot therefore be transferred solely according to internet software logic. Responsibility boundaries must be written into the governance structure in advance.
I may be misjudging this point. Momenta's 51% stake could primarily reflect its initial technology and team contributions rather than a permanent transfer of SAIC's control over future technical direction.
04 What It Means for AI Builders
First, reassess the technological center of gravity in automotive AI.
End-to-end driving models are important, but the truly scarce capabilities are the data loop, simulation validation, corner case management, and the deployment chain from models to vehicle OTA. For automakers as target customers, private deployment, inference optimization, and hardware adaptation may matter more than benchmark rankings beyond the model API.
Second, distinguish the platform provider from control.
Momenta may not control every vehicle model or customer relationship, but it could command the intelligent-driving software stack. The real moat may emerge when multiple OEMs use the same technical foundation, creating shared data scale and iteration speed.
Third, AI startups should take the joint-venture strategy seriously.
This model sits between acquisition and licensing: the algorithm company gains organizational control and expansion speed, while the OEM receives capital returns, faster product execution, and a degree of strategic autonomy. For Momenta, the joint venture can also reduce the risks of entering SAIC's supply chain, procurement system, and mass-production infrastructure.
Fourth, independent AI product founders should make a pragmatic distinction.
Do not look only for an API buyer; find a channel partner willing to share responsibility for delivery. Model capabilities can be integrated through APIs, but the closer a product gets to the physical world, the harder it becomes to close the commercial loop with a single model. Responsibility for deployment, validation, updates, and incident response may matter more than model elegance.
05 Counterargument / Risks
I may be overestimating the strategic significance of this transaction.
With registered capital of only RMB 300 million, the venture is not significant relative to SAIC's industrial scale. This could also be a limited capital arrangement: Momenta provides technology, SAIC provides vehicle introductions, and Yandi Technology supplies industrial resources. Without further product plans, vehicle commitments, or revenue guarantees, it is impossible to conclude that SAIC is comprehensively adopting Momenta's technical roadmap.
Moreover, a 51% equity stake does not necessarily translate into 51% product influence. Automotive projects involve safety, supply chains, and OEM brand boundaries, and technical control, vehicle definition rights, and intellectual property ownership may be divided differently.
The joint-venture model may also create new coordination costs. Momenta must serve shareholders, partners, and competitors, while SAIC must manage existing suppliers, technical teams, and internal organizational interests. Although the structure appears clear, decision-making is not automatically faster.
The greater risk lies in regulation and liability: if an algorithm company leads the intelligent-driving capabilities while the automaker bears vehicle responsibility, the two parties' interests regarding accident attribution, data use, and model updates are not naturally aligned.
I do not believe it is settled that automakers are abandoning in-house development. A more likely outcome is a small number of full-stack automakers alongside many hybrid organizations that rely on intelligent-driving platform companies. The key indicators to watch are not joint-venture announcements, but mass-produced models, delivery scale, OTA cadence, and whether the next OEM adopts the same structure.