Chinese carmakers are selling technology. Who can build a business around vehicle architecture?

From platform partnerships to the customer’s electronic architecture

Chinese automotive technology partnerships predate the current driver-assistance boom. In 2022, Geely and Renault announced cooperation in South Korea using CMA architecture and Geely hybrid technology, alongside an equity investment. This involved vehicle platforms and powertrains. It does not establish that Geely was independently licensing a complete E/E architecture.[4]

Cooperation moved further into intelligent vehicles in 2023. Volkswagen and XPeng agreed to develop models for China using XPeng’s G9 platform and relevant connectivity and driver-assistance software capabilities. SAIC and Audi also initiated cooperation on next-generation EVs. Chinese partners offered access to local digital development resources as well as competitive component costs.[5]

In 2024, the transactions became more specific: XPeng and Volkswagen agreed on joint E/E architecture development; NIO signed a technology licence with Forseven; and Leapmotor International began organising overseas operations through a joint venture. The AUDI E5 Sportback reached customer deliveries in 2025. In January 2026, Volkswagen announced production of the CEA architecture in the ID. UNYX 07.[1][6][7][8][9]

The progression matters. Cooperation over a vehicle platform is extending into responsibility for the vehicle’s software and control infrastructure. That infrastructure affects the cost of adding functions, integrating suppliers and updating vehicles throughout their lives.

E/E architecture defines how computation, communication and control are distributed: which processors run which functions, how sensors and actuators connect, how software with different safety requirements is separated, and what happens when a component fails. A zonal design connects nearby devices through local controllers to central computing. It can coexist with functional domains such as cockpit, body and powertrain control.

A 900-volt electrical platform, a driver-assistance chip and an infotainment system therefore demonstrate different capabilities. Each may be a saleable product; none alone proves the maturity of an entire E/E architecture. More useful evidence is reuse across models, acceptance into an external customer’s engineering process and economic returns after delivery.

Where the leading companies stand

The comparison below follows disclosed production and commercial evidence. Architecture names are not directly comparable technology rankings.

These are different strengths. External customer delivery, chip-and-operating-system reuse, and lower-cost integration all have commercial value. Choosing a supplier starts with the customer’s problem, rather than a ranking of peak computing power.

Fewer controllers may reduce wiring, connectors and some hardware costs. They also concentrate functions in fewer computing nodes. A buyer still needs to examine fault isolation, backup power, degraded operation and validation costs. Removing boxes is useful only when the resulting system delivers the required reliability and development economics.

The commercial models have different revenue sources

XPeng’s most significant achievement is entry into another automaker’s production architecture. CEA was jointly developed by Volkswagen China Technology Company, CARIAD China and XPeng. It had to fit Volkswagen’s vehicles, procurement and validation processes. This is a joint engineering programme, not a copy of an existing XPeng software installation.[1]

The financial evidence also matters. XPeng’s services and other revenue grew 93.9% year on year in the second quarter of 2026. It attributed growth to R&D service milestones and parts and accessories sales. The segment’s 75.1% gross margin exceeded the 12.1% vehicle margin, but its RMB2.70 billion revenue cannot all be treated as Volkswagen licence fees.[2]

The appeal is straightforward: technology initially funded for the manufacturer’s own products can earn additional revenue. The complication is that engineering revenue follows milestones. Programme delays and changing requirements affect its timing. The next meaningful proof would be a second customer adopting the technology with substantially less incremental engineering effort.

NIO provides a clearer example of licensing combined with engineering services. The original agreement provided for a non-exclusive worldwide licence, a fixed upfront fee and royalties linked to future licensed-product sales. NIO’s 2025 annual report explains that McLaren Automotive assumed the agreements and that separate paid services support vehicle development.[7][12]

This structure separates payment for initial IP and engineering work from potential returns as customer vehicles sell. NIO can spread technology investment across a wider product base; its customer gains access to an EV technology foundation. A signed licence and scaled customer vehicle sales remain distinct commercial milestones.

Leapmotor’s approach is closer to an industrial partnership. Stellantis owns 51% of Leapmotor International and Leapmotor 49%. The venture covers sales and associated manufacturing outside Greater China. In May 2026, the partners announced an intention to expand cooperation, including components from the venture’s ecosystem for a proposed Opel electric SUV and broader joint purchasing.[8][18]

Returns can come from vehicles, components, procurement scale and joint-venture operations. A pure software-royalty valuation would miss that structure. The partner already has brands, factories and distribution, creating routes to deploy the technology. The corresponding costs are deeper dependence on the partner and more complex allocation of value through joint-venture governance.

SAIC and Audi illustrate another route: joint development with distinct brand responsibilities. Audi brings its product requirements and engineering standards, while the Chinese partner contributes local technology and development resources. The E5’s reported 30% development-time reduction belongs to the combined programme, not to one chip or controller.[3]

Why sell technology now—and why not every OEM will succeed

The first incentive is the mismatch between fixed development spending and the sales volume of an individual brand. Chips, foundational software and engineering tools need substantial teams even if their output supports only a relatively small vehicle fleet. External customers enlarge the base over which investment can earn a return. For an OEM with very large internal volumes, keeping technology primarily within its own products can still be the better economic choice, supporting both scale and differentiation.

Buyers are also purchasing time. Missing a product cycle costs more than the current year’s sales: it affects dealer confidence, brand position and the next development budget. Adopting a mature technical foundation allows engineering resources to focus on design, driving characteristics, quality and user experience. The buyer must retain the ability to validate and maintain the system. Otherwise, development risk is exchanged for dependency on a supplier.

OEMs selling technology also face established suppliers. Huawei reported RMB45.018 billion of intelligent automotive solution revenue for 2025, up 72.1%. This covers multiple products and services rather than pure licensing, but demonstrates the scale already achieved in automotive technology supply. Honda’s agreement with Momenta shows that a customer can procure driver-assistance development without taking an entire vehicle architecture.[19][20]

That purchasing flexibility favours modular offerings. A controller, algorithm or toolchain that can be delivered independently is easier to fit into an existing platform. A complete architecture is more suitable for a new vehicle generation, with larger organisational and switching costs. Both markets can grow, but their sales cycles, staffing requirements and revenue structures differ.

An OEM needs a customer-facing delivery organisation to sustain either business: stable interfaces, clear licensing boundaries, reusable validation evidence and teams that support older vehicle programmes. A sudden internal vehicle deadline cannot repeatedly displace contracted customer work. Technical strength alone will not persuade another manufacturer to entrust its next product generation to an unreliable delivery partner.

The next test is a second customer—and profit over the vehicle’s life

The most likely outcome is a limited number of Chinese OEMs becoming long-term partners with identifiable technical specialisms, while many others continue to prioritise internal deployment. Complete architecture agreements will remain tied to particular programmes and markets; component and software-module supply can spread more widely. Multinational manufacturers’ China programmes offer the most immediate entry point. Expansion elsewhere requires fresh checks on software, data, IP and supply-chain requirements.

For sellers, management should track customer production, cash collection and reuse rather than agreement counts. The code and engineering headcount required for a second customer reveal more about business quality than the prestige of the first. Pricing must also cover updates, security fixes and long-term supply after launch. Development-stage income is not a sufficient return if years of maintenance are left unfunded.

For buyers, contracts should settle concrete questions: how requirement changes are priced, who owns newly developed code, whether maintenance can continue if the supplier exits, how software migrates when chips change, and how responsibility is allocated between software defects and vehicle recalls. A low entry price is unhelpful if each subsequent update becomes another major development programme.

Chinese manufacturers have demonstrated that they can participate in the core development processes of multinational automakers. The next competitive test is repeatability across a second and third customer, with profits surviving years of support. XPeng needs to broaden its external customer base; NIO needs licensed technology to reach customer vehicles; Leapmotor needs to demonstrate returns from cross-brand component cooperation. Those outcomes deserve closer attention than another memorandum of understanding.