Uber Announces Deal with NVIDIA to Deliver 100,000 Robotaxis from 2027

At NVIDIA GTC Washington, D.C., Uber announced a partnership with NVIDIA that will see the ride-hailing giant leveraging the Hyperion 10 Autonomous Vehicle (AV) development platform, along with software assets from the NVIDIA DRIVE AV software stack to put 100,000 driverless robotaxis on the road from 2027 onward. For manufacturing, Uber and NVIDIA will rely on automotive Original Equipment Manufacturers (OEMs), many of which are already leveraging Hyperion as the basis of their passenger vehicle Level 4 automation, including Stellantis, Lucid, and Nero, to construct the actual vehicles. The Hyperion reference platform in tandem with NVIDIA development tools such as the Cosmos world foundation model gives Uber and other vendors leveraging Hyperion access to an ecosystem of AV software development partners, including Wayve, Pony.AI, Momenta, Nuro, and WeRide, among others.

With low-cost mobility powered by shared driverless vehicles long touted as the killer use case for autonomous driving, but with few such vehicles deployed to date, a commitment by the world's most recognizable mobility brand to roll out 100,000 robotaxis backed by an ecosystem of driverless technology developers must be seen as a watershed moment for the autonomous driving market, although the exact pace at which these vehicles will be rolled out remains to be seen. However, given the importance of removing the cost of human drivers to enabling sustainable pricing from mobility services below that of conventional car ownership, it is expected that any initial success in folding robotaxis into Uber’s mix of ride fulfillment options will quickly lead to widespread rollout.

As much as ride-hailing companies need driverless platforms to put their operations on a sustainable footing, driverless platform developers need ride-hailing operator partners to introduce that technology to market and begin the long-awaited monetization of their multi-year investments.

When a robotaxi platform developer wants to deploy their technology in a shared mobility service, they have two options. The first is to create their own mobility service, developing their own platform to address demand management, dispatch, routing etc., and establishing a new brand for point-to-point mobility consumption. The second option is to partner with an existing supplier of shared mobility services, taking advantage of their experience in ride fulfillment, and instead operating as a fulfillment layer, powering the low-cost and high-utilization driverless vehicle fleet.

The partnership approach has clear advantages over the go-it-alone approach. While it means the driverless technology developer cannot benefit from the higher fulfillment layers and its related revenue opportunity, it gives them access to an established mobility brand, with a large user base, critical experience in rapidly and efficiently fulfilling mobility demand/rides, and somewhat paradoxically, a large fleet of manually-driven vehicles.

Consider the example of a new driverless mobility service, consisting of a fleet of 100 or so robotaxis entering a market already serviced by a conventional ride-hailing operation, from the perspective of a consumer that simply wants to get from A to B as quickly and cheaply as possible. Depending on the size of the area served, the standalone robotaxi service may not be able to rapidly fulfill a requested ride, with no cheap or easy ways to add more vehicles to the fleet to accommodate higher demand. Conversely, a conventional ride-hailing service, through mechanisms such as price surging, can quickly encourage more contract drivers to fulfill requested rides, leading to lower waiting times.

Similarly, a new robotaxi service with Operational Design Domain (ODD) constraints that prevent it from operating on certain road types or in certain areas of a city may be required to take more circuitous routes than manually-driven vehicles, leading to longer journey times for the pure robotaxi service, at least until ODD constraints are diminished. Finally, a pure robotaxi service, looking to amortize its costs across a comparatively small fleet of expensive vehicles delivering a modest number of rides can easily end up offering a cost-per-ride that is unfavorable compared to the conventional ride-hailing alternative.

This has been the typical consumer experience in the limited deployment of pure robotaxi services to date. It has been reported that wait times for a driverless Waymo service have been around 10 to 20 minutes longer than the equivalent conventional ride-hailing service from Uber or Lyft, with the same journey sometimes proving 20% to 50% more expensive, depending on the surge pricing in effect and the amount of tip paid to the driver. Once the initial novelty of a driverless experience has worn off, and consumers simply seek the most cost-effective and time-efficient way to get from A to B, it is clear which alternative the majority will choose.

By partnering with an established ride-hailing brand, a robotaxi platform supplier has a feasible path to market, introducing their technology into a mixed fleet that can spread the costs across an already high volume of rides, with consumer demand satisfied by price surging and manually-driven options, while the installed base of robotaxis slowly expands.

Uber has not been shy about partnering with third-party, fully-autonomous platform suppliers. Indeed, this is the latest in a slew of 2025 announcements, many of which were formed with the same software development partners announced as part of the Hyperion 10 ecosystem at NVIDIA GTC Washington, D.C. In July 2025, Uber announced partnerships with both Lucid/Nuro and Baidu Apollo, following the May announcements detailing partnerships with Pony.ai and Momenta, and a partnership with Volkswagen MOIA/Mobileye announced in April. The year 2025 also saw announcements with Avride and WeRide, as well as the expansion of a partnership with Waymo that began in 2024.

To an extent, this represents a hedging of Uber’s bets. Among the software development suppliers listed above, there are a variety of approaches to architectures (End-to-End (E2E) versus modular Artificial Intelligence (AI)), safety verification, and map use. Furthermore, the geographic variety of these partners can help Uber, a global mobility brand, to secure its position as a driverless mobility supplier in all key regions.

More fundamentally, this approach is a close reflection of how Uber operates today. Drivers are sourced from all backgrounds and levels of experience. Ultimately, it is how well these drivers perform that dictates whether they continue on the platform, fulfilling rides on behalf of customers. Therefore, Uber’s myriad software partners should consider themselves to be entering a trial period, with expansion of their system dependent on the ability of their platform to deliver on the Key Performance Indicators (KPIs) of safety, uptime, and breadth of ODD.

Further evidence of Uber’s current DNA in its future autonomous mobility plans is the variety of OEM partners. With both Stellantis and Lucid featured as manufacturing partners leveraging NVIDIA’s Hyperion, Uber can expect to have both premium and mass market vehicles in their future autonomous fleets, reflecting the UberX and Uber Black service levels that exist today. For Uber, this is made much easier, due to the fact that Stellantis and Lucid are already leveraging the NVIDIA DIVE AGX Hyperion platform in their Level 4 passenger vehicles. Therefore, autonomous vehicle platform suppliers should double down on Level 4 partnerships with passenger vehicle OEMs to maximize their reach into the robotaxi market, building partnerships with mass-market and premium OEMs to provide a prepackaged platform for a variety of mobility services.