Self-Driving Cars in France: Wayve, Stellantis and Nvidia
Wayve, Stellantis and Nvidia are partnering with France on self-driving cars. Explore the technology, regulatory hurdles and what a European rollout requires.
France's Self-Driving Car Plan: What Wayve, Stellantis and Nvidia Bring
Self-driving cars are moving into a new phase in Europe: fewer "look what we can do" demos, and more plans that tie vehicle makers, AI teams, computing suppliers, and transport regulators together. Wayve, Stellantis and Nvidia say they'll work with the French government to deploy autonomous vehicles in France and support a broader European rollout. Uber is also reportedly involved, which would connect the effort to an existing ride-hailing network.
Still, this isn't a launch schedule. The companies haven't named the first cities, confirmed a fleet size, or set a start date. And that matters, because getting to "robotaxi that works" is more than having a capable car and software stack. It also depends on regulatory approval, day-to-day safety operations, local transport agreements, and a service model that can actually run profitably.
Here's what the partnership does say, what's left undecided, and what you should look for if you want to judge whether this can move beyond carefully managed trials.
What the partnership actually announces
The companies made the announcement at the Assises du Véhicule Autonome in Paris. Their stated plan is to work with the French government on deploying autonomous vehicles in France, with the goal of accelerating adoption across the European Union.
Each partner has a different role:
- Stellantis will supply vehicles based on platforms designed to be ready for Level 4 driving.
- Wayve (a London-based autonomous-driving company) will provide its AI Driver software.
- Nvidia will provide its Hyperion computing and sensor system.
- Uber is reported to be part of the plan, contributing its ride-hailing network, according to Journal de l'Automobile, which cited Stellantis.
On paper, that breakdown is the practical logic of a robotaxi service: you need a vehicle platform, software that can interpret the driving environment and make decisions, and an operator who can run rides. These companies cover multiple layers rather than presenting a single end-to-end solution.
But the operational picture is thin. There's no confirmed launch date, fleet size, or list of cities. The companies also haven't spelled out the service model - whether the first vehicles will operate as public robotaxis, in a restricted area, or as part of a limited pilot.
Those distinctions matter. "Deploying" could mean everything from engineering tests to supervised public trials to a restricted commercial service and, later, wider availability. Until the partners publish where they'll operate and what the on-the-ground plan looks like, it's hard to treat the announcement as proof that driverless rides are about to become widely available in France.
Uber's role also needs specifics. The involvement was reported by a French trade publication, but the available announcement details don't lay out Uber's operating plan in France. Uber and Wayve also have a separate robotaxi collaboration in London. In that program, rides launched in September with a safety driver in every car, which shows supervised operations - not an unrestricted driverless service.
The original report from The Next Web has the initial details. The next update that will actually matter is more concrete: where the vehicles will operate, who will oversee them, and what level of automation the service will offer.
How the technology stack fits together
A self-driving vehicle isn't one software app. It's a full system: the car platform, sensors, onboard computing, driving software, operational procedures, and a way to manage the service.
Stellantis supplies the vehicle platform
Stellantis says it will provide platforms designed to be ready for Level 4 driving. In the usual automation framework, Level 4 means the vehicle can handle the driving task within defined conditions or an operational area without requiring a human to take over in normal operation. It doesn't mean it can drive anywhere, in any weather, on every road.
Those limits are part of the design. A Level 4 service might be constrained by geography, road type, weather, or other operating conditions. A system that works in a mapped urban area may still need more work for motorways or rural routes. In other words, the operating boundary matters as much as the "Level 4" label.
Stellantis has already worked with Wayve on demonstrations. Earlier in October, the companies showed a Fiat 500e and a Maserati Grecale driving hands-free with Wayve software in Turin. They also described a first launch in North America planned for 2028. That's evidence of collaboration, but a hands-free demo isn't the same as running a driverless service on public roads. It doesn't, by itself, establish a French schedule or approvals for public operation.
Wayve provides the driving software
Wayve's AI Driver is meant to interpret the environment and control how the vehicle drives. In practice, the system has to use sensor inputs to identify relevant road users and conditions, understand what's happening around the car, and choose actions that keep the vehicle moving safely.
The real test isn't whether a vehicle can pull off a polished demonstration. It's whether the system responds consistently to the "long tail" of everyday situations: temporary roadworks, delivery vehicles blocking lanes, confusing signage, cyclists filtering through traffic, emergency vehicles, and unfamiliar road layouts. A setup can look strong in one city and still require substantial evaluation elsewhere.
Wayve's involvement in both the French plan and the London collaboration with Uber could help the teams learn across programs. But it doesn't mean permissions, operating conditions, and system behavior transfer cleanly. A service in London can't automatically be assumed to work the same way in Paris or other parts of France without local validation.
Nvidia supplies computing and sensors
Nvidia's Hyperion system is described as the vehicle's computer and sensor platform - the hardware layer that supports perception and the driving software, including the processing of sensor data. In a real vehicle, that layer has to cope with heat, vibration, power draw, and component failures, not just clean lab conditions.
This is why you should think of the partnership as an integrated system. If a software update changes how the car reacts, the partners need to track which software version runs on which hardware and under which approved conditions. If a sensor needs replacement, you may need calibration and checks before the vehicle returns to service. Those processes affect fleet availability and cost, not just raw technical performance.
The roles are clearer than the responsibilities. The announcement doesn't yet explain how the companies will split responsibility for service operations, incident response, software updates, or vehicle maintenance. Those details matter more as the program moves from an announcement to real-world testing.
What has to happen before a French rollout
A rollout needs several things to line up: permission to operate, evidence that the system is safe within its intended limits, agreement with where it will run, and a service model that can be maintained. The French government's participation matters because national rules and local implementation can determine whether a pilot can progress.
Regulation must enable trials without lowering the bar
France's transport minister, Philippe Tabarot, said the country plans to change its rules to speed up self-driving cars while keeping strict safety and data-sovereignty requirements. The balance is critical. If the process is too slow, testing doesn't happen when it could. If it's too loose, road users can face risks that haven't been assessed properly.
The source report says the EU approval route for fully self-driving vehicles is currently capped at 1,500 vehicles per year. That cap affects how manufacturers plan for bigger deployments. It doesn't mean the French partnership has already been approved for a specific number of vehicles, and the announcement doesn't specify which approval path its operators would use.
Data sovereignty is also practical, not theoretical. Autonomous vehicles collect and process information about roads and surrounding activity. The partners will need to explain how data is handled, where relevant processing occurs, who can access information, and how privacy and security responsibilities are divided. The announcement doesn't provide those operational details yet.
Local coordination is part of the engineering work
The partners say local transport partners will consult unions, city authorities, and other stakeholders before a rollout. That isn't just a communications step - it can shape the service itself: where vehicles pick up passengers, how they interact with bus and taxi infrastructure, which roads are suitable, and what support exists if a vehicle stops or needs help.
For example, a vehicle might be technically capable of navigating a busy city center, but the service can still fail if pickup points block traffic or if there's no agreed plan for a blocked roadway. The city's transport system becomes part of the deployment environment. Ignore it and a promising test can turn into an operational headache.
Union consultation is especially relevant when autonomous services overlap with existing driving and transport jobs. The public information doesn't say what agreements will be reached, which is another reason to separate the general intention to deploy from a confirmed service plan.
Safety operations need to be explicit
A robotaxi service isn't just a vehicle driving itself. It also requires processes for monitoring, responding to incidents, handling technical faults, and helping passengers. If the vehicle runs into a situation it can't manage, the operator needs a defined response. If a passenger needs assistance, there has to be a clear support channel.
The London rides between Wayve and Uber offer a useful comparison because each car has a safety driver. That approach can support supervised testing and show how the service behaves with real passengers. It's not the same as a fully driverless service, and it doesn't tell you what the French deployment will require.
Before assuming a rollout is imminent, look for specifics: the permitted operating area, whether a safety driver is present, how incidents are reported, and what happens when conditions fall outside the system's limits. Without those details, a launch date alone tells you very little about autonomy or real maturity.
Why the European race matters
Europe has lagged the United States and China in expanding robotaxi services, according to the source report. That gap doesn't come from one cause. A company might have capable technology but still face complicated approvals, differences between cities, and the cost of adapting operations to local conditions. The French initiative aims to coordinate major players and government around that deployment challenge.
There are already other efforts in the region. Uber's first European self-driving rides started in Zagreb in August, again with a safety driver in the car. Pony.ai's software is used for what the report describes as Europe's first fully driverless rides, also in Zagreb. Pony.ai and Uber plan to deploy more than 2,000 robotaxis on European roads, according to the same report.
These examples are why "like with like" matters. A supervised ride, a fully driverless operation, and a plan to expand a fleet aren't the same kind of milestone. A headline fleet target doesn't mean every vehicle is operating on public roads right now, and an early public ride doesn't automatically translate into citywide availability.
For the French partnership, the competitive question isn't only who reaches the road first. It's whether the partners can make the service repeatable across locations while meeting each market's safety and data requirements. A model that relies heavily on human support or works only in a tiny test zone may struggle to deliver real commercial value. On the other hand, tight limits can be a sensible way to build reliable operations before expanding.
The business side matters too. Stellantis can supply vehicles, Wayve can provide software, and Nvidia can supply computing and sensors - but those contributions don't automatically produce a ride service. Operators still need charging or refuelling, maintenance, cleaning, passenger support, insurance, and managing vehicle downtime. The announcement doesn't say how those costs and responsibilities will be handled.
So the next milestones should be judged by evidence, not ambition. A named pilot area, a clear safety-driver policy, a published approval pathway, and stated operating conditions would make the plan more concrete. Measured results on reliability and service availability would matter even more. The partnership has assembled the pieces - now the hard part is showing that they can work together under real local constraints.
Conclusion
Wayve, Stellantis and Nvidia have put forward a coherent set of technical roles for a French self-driving vehicle program, with Uber reportedly connected to ride-hailing operations. If France follows through on its plan to update its rules while keeping safety and data requirements clear, it could help move the work forward.
But this is still a partnership announcement, not a confirmed robotaxi launch. The key details - cities, fleet size, timing, safety-driver policy, and operating approvals - aren't public yet. For now, the best way to track progress is to watch for concrete deployment conditions and evidence from real-world operations, not just claims about autonomy.
For more context, read The Next Web's report on the partnership, then look for the partners' next update on where and how the first French vehicles will run.