Why Volkswagen May Stop Owning Every Line of Driving Code

A partnership designed for scale

When Bosch and Volkswagen’s software unit Cariad announced the Automated Driving Alliance in 2022, the pairing made sense. Bosch knew automotive components, validation and mass production. Volkswagen had a huge vehicle portfolio and the prospect of collecting data from millions of cars. More than 1,000 specialists were expected to work on assistance systems, including hands-free Level 2 features and Level 3 highway driving.

Reuters, citing Bild, now reports that Volkswagen plans to end the partnership as part of a wider effort to cut costs and improve competitiveness. The companies have not yet confirmed the reported breakup or claims that the system fell behind rivals. Even so, the report points to a familiar problem for established automakers: a software program can absorb years of spending without giving customers a visible advantage on schedule.

Four years is a long time in autonomy

Traditional vehicle development rewards patience. A platform may stay in production for most of a decade, and engineers validate safety-critical hardware through slow, tightly controlled processes. Machine-learning systems move faster. Progress depends on data pipelines, computing capacity, frequent iteration and the ability to update cars after delivery.

That difference in pace can make a sensible alliance hard to manage. Two large companies have to agree on sensors, computing hardware, software interfaces and validation. They also have to decide who is responsible when something fails and which brands get each feature. Extra approvals can reduce risk, but they can also delay learning.

Comparisons with competitors need care. Tesla’s Full Self-Driving in customer cars is still a supervised assistance system, not unrestricted autonomy. A demonstration, a consumer Level 2 product and a future Level 3 system are not the same thing. Customers may not follow those categories closely, but they notice whether updates arrive and whether the software feels consistent. That perception can affect sales well before the technical definitions line up.

The cost of owning the whole stack

Volkswagen is considering this change while restructuring other parts of the group. Reports have also described possible job cuts, factory closures and lower investment. Autonomy therefore competes for money with battery plants, upcoming models, factory upgrades and regional vehicle programs.

Building an automated-driving stack internally gives an automaker control over intellectual property, integration and future margins. It also brings large fixed costs. Fleets have to gather useful data, teams must label unusual events, simulations need constant maintenance and every release has to be checked across vehicles and markets. A system that makes financial sense in one premium model may not work across a group that also sells small cars, vans and sports cars.

Buying speed without giving up control

Volkswagen is already more willing to use outside software. Its joint venture with Rivian is developing a zonal electrical architecture and software platform for future vehicles, and Volkswagen plans to invest up to $5.8 billion in Rivian and the venture. That work does not replace an automated-driving stack, but the strategy is clear: use outside technology to shorten development, then apply Volkswagen’s production scale.

A new partner can create a new dependency. If a supplier owns the most important software, Volkswagen may find it harder to set its cars apart or control future costs. Integration can also become brittle when the electrical architecture, infotainment and driving system come from separate organizations.

Selective ownership may be the practical answer. Volkswagen could keep control of vehicle data, safety cases, the user experience and core interfaces while buying specialized perception or planning technology. It is a less dramatic story than owning every line of code, but it may produce a car sooner.

What Volkswagen has to prove next

Ending a partnership does not mean a better product is waiting. Volkswagen still has to choose a replacement, integrate it into production cars and show an improvement in timing, cost or performance. Drivers will judge the result in ordinary situations: whether the car stays centered, intervenes predictably, gives clear warnings and receives useful updates.

The unanswered question is whether Volkswagen can organize itself around software that is never truly finished. It does not need to own every component. It does need a firm view of which parts define the product and which parts can come from a partner.