On 22 September 2026, Mercedes-Benz and the British startup Wayve signed a final production agreement to integrate the AI Driver system into the German group's production models. The first cars with this technology are expected to appear on the market within the next two years — another sign that leading automakers have begun a mass transition from rigidly programmed algorithms to trainable neural networks.
This is not about full autonomy, but about an advanced driver-assistance system for "point-to-point" driving in urban conditions and on highways. It is based on the AV2.0 approach: the neural network learns to drive on real data, without rigid high-definition maps and geofences. This allows it to adapt to new cities without retraining for a specific location — a fundamental difference from competing solutions that require costly map rework for each region.
The deal was a logical continuation of many years of cooperation. In February 2026, Mercedes-Benz invested in Wayve's Series D round, which ultimately raised 1,2 billion dollars, with additional funding commitments from Uber and other investors. At the same time, technology giants Microsoft and Nvidia, as well as Mercedes competitors Nissan and Stellantis, participated in the round, signaling a growing consensus around end-to-end AI as the foundational platform for autonomous driving. Now the software is integrated into the MB.OS architecture, uses Mercedes' standard hardware, and runs on Nvidia infrastructure via Microsoft Azure. Joint tests have already taken place in Stuttgart, London, and San Francisco, demonstrating the versatility of the approach.
For Mercedes, this is the first case of deploying end-to-end AI in the premium segment. The company is simultaneously developing several technology stacks: with Nvidia (based on the DRIVE AV platform for L4 robotaxis) and Momenta (for the Chinese and global markets). However, the choice of Wayve underscores a bet on scalable solutions that can work without detailed city maps and spread more quickly across various geographic regions. For Wayve, this partnership opens access to the engineering expertise and production capacity of one of the leading automakers, and also fits into its software licensing strategy — the company already works with Nissan and Stellantis on similar terms.
Buyers will get more natural and flexible driving assistance, especially in complex urban environments with dense traffic and unpredictable situations. At the same time, responsibility for driving remains with the human — the system does not claim level 4 or 5 and does not allow the driver to become fully distracted. For the industry, this is a signal: the traditional approach with predefined rules is giving way to systems capable of learning from real-world data and improving continuously.
In the long term, such a partnership could accelerate the emergence of truly universal solutions capable of operating in different countries and climate zones without costly adaptation and retraining. The main technical challenge for the industry now lies not so much in the algorithms themselves as in how quickly regulators in different regions will accept new approaches to certification and allow manufacturers to deploy such systems on a mass scale.
