Bolt Chooses Paris for 2028 Driverless Taxis With Stellantis
By Daniel Rao
Bolt wants to put driverless taxis on the streets of Paris — and it wants to do it with French-built vans, Chinese self-driving software, American chips, and the blessing of the Greater Paris regional government. If that sounds like a geopolitical jigsaw puzzle, welcome to the European robotaxi race, where the vehicles are local but almost nothing else is.
Here's what was actually announced. On Friday 23 October 2026, Estonian ride-hailing firm Bolt, carmaker Stellantis and the Métropole du Grand Paris said in a joint statement that they will launch an autonomous-driving pilot in the French capital in 2028, with the aim of starting commercial operations across Greater Paris by 2030. The pilot — officially "The Greater Paris Autonomous Driving Living Lab" — will use Stellantis vehicles, run on Bolt's app through its Bolt Autonomous Driving Solutions subsidiary, and be independently evaluated on safety, user acceptance and integration with public transport. The potential autonomy supplier? China's Pony.ai. Its role, the partners stress, is not confirmed.
And here's the twist that makes today genuinely remarkable: this was Stellantis's second autonomous-driving announcement in France on the same day. Hours apart, the carmaker also joined Wayve, NVIDIA and the French government at the Assises du Véhicule Autonome in Paris to build a driverless ecosystem on French soil. One company, two AV software partners, one country, one day — and both deals running, at the silicon layer, on NVIDIA hardware. Stellantis isn't betting on a horse. It's buying the racetrack.
The Greater Paris Autonomous Driving Living Lab
The details, such as they are, sketch a cautious, government-shaped programme rather than a Silicon Valley blitz. First deployments will cover up to five municipalities, mainly in the north-west of the metropolitan area — the partners didn't name them. The metro area itself spans Paris plus 129 surrounding communes with about 7.2 million residents, and the ambition is to reach around 30 municipalities by 2030, with a second phase assessing the conditions for wider deployment.
What will the vehicles actually do? The unglamorous, useful stuff: first- and last-mile trips, access to employment hubs, off-peak travel, and coverage for major events. The pilot is explicitly framed as a gap-finder — it "will assess gaps in existing transport services," the three parties said — which is a telling admission. Even with the Grand Paris Express metro expansion being tunnelled beneath the region as we speak, large parts of the suburbs remain badly served by public transport. Driverless shuttles are being pitched as the filler, not the replacement.
We looked for a fleet size, an investment figure, or the names of the five starting municipalities in every version of the announcement we could find. None of them are there. That's either discipline or evasiveness — with these three partners, probably a bit of both. What is specified is the evaluation: an independent assessment covering safety, user acceptance, integration with public transport, and the service's social, environmental and economic effects. In other words, this pilot is as much a policy instrument as a technology trial.
That policy framing comes straight from the top. "This experiment will also help inform public decision-making and define the conditions for potential wider deployment," said Patrick Ollier, President of the Métropole du Grand Paris — a man whose organisation, he noted, has run more than 70 transport experiments over the past decade. "Technological progress only matters when it delivers tangible benefits to our citizens," Ollier added. Seventy experiments in ten years: Paris doesn't do tech hype, it does pilot programmes. This is how the city learns — slowly, publicly, and with the paperwork to prove it.
The division of labour is clean. The Métropole coordinates with the municipalities and oversees the evaluation. Bolt handles service integration and deployment planning. Stellantis supplies vehicles built on its Level 4-ready platforms, plus the engineering, manufacturing and integration work. Ned Curic, Stellantis's Chief Engineering and Technology Officer, said the goal was to develop "a driverless mobility that complements public transport" — complements, note, not competes with. In France, you do not pick a fight with the public transport unions. You compliment them and ask for a lane.
The Chinese brains on American chips
Let's talk about the elephant in the announcement: Pony.ai. The Guangzhou-based company is one of the few autonomous-driving firms on the planet actually making money from robotaxis — its second-quarter 2026 results showed robotaxi-services revenue up 691% year-on-year to $12.1 million, on a global fleet of 1,975 vehicles at the end of June, with a target of more than 3,500 by the end of this year. Its seventh-generation system runs on production vehicles from BAIC, GAC and Toyota; its PonyPilot app has over 1.5 million registered users in China; and management says unit economics have turned positive in Guangzhou and Shenzhen. This is not a lab project. It's a business, albeit one still losing money overall.
Pony.ai's ambitions stretch beyond Paris: on 23 October it signed a memorandum of understanding with inDrive and NVIDIA to trial Gen-7 robotaxis in Astana, Kazakhstan in the first half of 2027, with NVIDIA supplying the same Hyperion platform and Halos safety system (TechCrunch; Dow Jones Newswires).
Pony.ai's European footprint is growing fast. It has more than 4,000 international vehicle commitments on the books, including over 2,000 robotaxis across five European cities tied to Uber — the two companies announced that plan in August. It already runs a paid service in Zagreb with local operator Verne and has testing planned in London. And it knows Stellantis hardware intimately: the two signed their own partnership in October 2025, starting with an eight-seat, fully electric Peugeot e-Traveller van for Level 4 testing in Luxembourg, with deployments in other European cities planned from 2026. Last June, Bolt joined that Luxembourg effort with a one-year trial — safety driver on board — running Pony.ai's seventh-generation system on a Stellantis mid-size van.
So the Paris pilot isn't a blind date. It's the third date, with the parents invited. And notice the technology stack the partners describe: if Pony.ai is chosen, its system would run on NVIDIA's Hyperion hardware platform with the Halos safety system. Chinese driving software, American compute, French-built van, Parisian regulator. Everybody gets a piece, which is precisely the point — and precisely the fragility. France's transport minister spent the same day demanding "data sovereignty" in autonomous vehicles. Squaring that circle with a Chinese AI stack processing Parisian street data is the kind of diplomatic engineering that will decide whether 2028 actually happens.
Madrid vs Paris: Bolt's two-stack strategy
The Paris announcement is only half of Bolt's week. On Wednesday, the company said it would start testing autonomous vehicles on the streets of Madrid in 2027 — a year earlier than Paris — using Lucid vehicles running NVIDIA's Alpamayo driving software on Hyperion hardware. That programme starts with "tens of vehicles" doing driving-data collection, and Madrid's regional government has already created a permit allowing limited commercial operations before full approval arrives. Paris, notably, has announced no equivalent framework; commercial service there depends on authorizations that don't exist yet.
Two cities, two stacks. Madrid gets American software (Alpamayo) on American hardware in American-designed, Saudi-backed Lucid vehicles. Paris gets potentially Chinese software on American hardware in Franco-Italian Stellantis vans. Bolt is running a controlled experiment on itself: which combination clears European regulators, wins riders and scales fastest?
The Madrid programme builds on Bolt's September agreement with Lucid for at least 25,000 robotaxis on the carmaker's upcoming Midsize platform — though it's worth remembering that agreement is a memorandum of understanding with no purchase commitment, and Lucid has only said the first batch could roll out as early as 2028. Bolt's stated endgame is enormous either way: 100,000 autonomous vehicles on its platform across Europe and the wider UNECE region by 2035. "Our ambition is not to bring autonomous driving to just a handful of cities," Jevgeni Kabanov, Bolt's President and Head of Autonomous Driving, said on Wednesday, naming Germany and France as key future markets.
Bolt's founder and CEO Markus Villig put the philosophy bluntly: "Autonomous vehicles will only matter if they solve real transport problems." It's a good line, and the Paris pilot is built to test exactly that thesis — first-mile gaps, employment hubs, off-peak deserts. But there's a less poetic reading too: Bolt, which operates ride-hailing in more than 50 countries, has realised it can't out-Waymo Waymo on technology, so it's trying to out-local everyone on deployment. Be the company that knows the mayor, not the company with the best neural net.
The sister deal signed the same day
Which returns us to the day's real story. While Bolt, Stellantis and the Métropole were announcing their living lab, Stellantis was also standing beside Wayve, NVIDIA and the French government at the Assises du Véhicule Autonome, committing its L4-Ready Platforms and manufacturing to a second, parallel driverless ecosystem for France — this one with Wayve's AI Driver running NVIDIA's Halos OS on Hyperion. (Full story: Stellantis Picks France for a Wayve-NVIDIA Driverless Alliance)
Read the two announcements together and Stellantis's strategy snaps into focus. The company has no intention of picking a winning autonomy software stack. Instead it's positioning itself as the indispensable vehicle supplier to all of them — Wayve's embodied AI in one deal, Pony.ai's system in the other — while NVIDIA collects the compute rent underneath both. CEO Antonio Filosa said back in December that Stellantis's AV-ready platforms "are designed for maximum flexibility," and expected the company's vehicles to play a major role in Bolt's plans. A double signing in Paris, eleven months later, is what that flexibility looks like when the cameras are on.
The project has also been submitted to the European Commission's ADACities initiative, launched in June to help selected cities deploy autonomous mobility, with participating cities targeting fleets of 100 or more autonomous vehicles by 2030. Brussels wants robotaxis; Paris wants conditions; Stellantis wants volume. Everyone's incentives point the same direction, which is rarer than it sounds.
Our read: the politics are the product
Here's the uncomfortable truth about the Paris announcement: the technology is the least interesting part of it. Pony.ai's stack works — 1,975 robotaxis don't lie. Stellantis can build vans in its sleep. The genuinely hard problems are all political: a French regulatory framework that doesn't exist yet, data-sovereignty demands sitting awkwardly next to a Chinese software supplier, unions that can stall anything, and a 2030 commercial target that requires roughly four years of flawless execution across dozens of municipalities. Madrid understood this — it wrote the permit first and invited the cars second. Paris is doing it the French way: announce the ambition, then build the paperwork. That approach has a name in Paris. It's called normal.
And yet — dismissing this as press-release theatre would be the shoddy take. Look at what actually got assembled in one week: two European robotaxi programmes, two software stacks, a 100,000-vehicle 2035 ambition, a regional government co-signing as a partner rather than a spectator, and an EU initiative (ADACities) waiting at the end of the road with a 100-vehicle-per-city target. Bolt is doing something its American and Chinese rivals aren't: it is building the political machine first and bolting the robots onto it later. Waymo conquered Phoenix by being the best technology. In Europe, the winner might just be whoever is best at committees. Nobody does committees like the French — and Bolt just hired the whole country as a co-founder.
The race for Europe's streets is properly on now. Baidu's Apollo Go is coming with Lyft and Uber starting in London. WeRide and Uber have already taken Spain's first Level 4 permit for a 20-vehicle Madrid fleet. Waymo's driverless taxi launch in Tokyo shows the American leader's global intentions, the company has begun testing in Munich ahead of a planned 2027 launch, and Tesla's paid Cybercab rides in Austin prove the vertically integrated model can charge real fares. Against that field, a 2028 pilot with no named municipalities and no fleet size looks modest. Modest, though, is how Paris does everything — right up until the moment it doesn't.
Frequently asked questions
When will Bolt's autonomous driving programme launch in Greater Paris?
The pilot is planned for 2028, with commercial operations targeted by 2030 — subject to obtaining the necessary authorizations, which don't exist yet. Paris has no equivalent of Madrid's pre-approval commercial permit.
Which companies are involved in the Paris robotaxi experiment?
Bolt, Stellantis and the Métropole du Grand Paris (the Greater Paris regional authority). Bolt leads service integration through its Bolt Autonomous Driving Solutions subsidiary, Stellantis supplies the Level 4-ready vehicles, and the Métropole coordinates with municipalities and oversees evaluation. China's Pony.ai is a potential — not confirmed — autonomous-driving technology supplier.
Where exactly will the pilot run?
First deployments will cover up to five municipalities, mainly in the north-west of the metropolitan area — the partners haven't named them. The metro area covers Paris plus 129 surrounding communes (about 7.2 million residents), with a target of around 30 municipalities by 2030.
How does the Paris plan differ from Bolt's Madrid programme?
Madrid starts earlier (testing in 2027), uses Lucid vehicles running NVIDIA's Alpamayo software, begins with tens of vehicles collecting driving data, and already has a regional permit for limited commercial operations. Paris starts in 2028, uses Stellantis vehicles with Pony.ai as a possible software supplier, and still needs its regulatory framework built.
Sources
Reuters (23 October 2026, via 979weve; reporting by Danny Callaghan, editing by Milla Nissi-Prussak); eletric-vehicles.com; Pony.ai Q2 2026 results (GlobeNewswire, via Insider Monkey, Manila Times and aichina.news); Stellantis media and Wayve press releases (23 October 2026); globalbankingandfinance.com (Reuters reprint with FAQ).
About the author: Daniel Rao covers autonomous vehicles and AI infrastructure for TechPulse USA. He has reported on the robotaxi industry across three continents and still doesn't trust a demo video he hasn't seen with his own eyes.
Comments
Post a Comment