How Self-Driving Ubers Work, Safety, Availability, and What Riders Should Expect

Quick Answer

Uber's self-driving future relies on partnerships rather than in-house vehicle development. The company has abandoned its earlier effort to build its own autonomous vehicles and now works with a wide network of technology and automaker partners, including Nvidia, Lucid, Nuro, Momenta, Baidu, and others.

Riders should expect a phased rollout, with testing already underway in San Francisco and Munich, and a large-scale deployment of up to 100,000 self-driving vehicles targeted for 2027.

Key Facts

Uber shifted from building its own self-driving cars to partnering with companies like Nuro and Lucid for its robotaxi program.
Uber plans to deploy 100,000 self-driving vehicles by 2027, powered by Nvidia's Hyperion autonomous platform.
Autonomous testing is currently underway on public roads in the San Francisco Bay Area, with a public launch expected later this year.
Uber and Momenta plan to begin self-driving testing in Munich, Germany, in 2026.
Baidu will start testing Apollo Go robotaxis in London in the first half of 2026.
Lucid will supply 20,000 Gravity SUVs to Nuro, which will install its autonomous technology.
Uber has invested directly in Lucid Motors for a premium robotaxi service.
Stellantis has committed to delivering 5,000 vehicles for Uber's robotaxi program.
As of November 17, 2025, there were 5,202 reported autonomous vehicle incidents in the United States, with 451 injuries and fatalities combined.
Approximately 7.4% of autonomous vehicle accidents resulted in injury, and 1.2% resulted in a fatality.
Waymo has reported the most autonomous driving system (ADS) vehicle accidents, while Tesla has reported the most advanced driver assistance system (ADAS) accidents.

From Building Cars to Brokering Rides Uber's Strategic Pivot

Uber's journey to self-driving vehicles has been anything but linear. The company originally invested heavily in developing its own autonomous vehicle technology, a costly endeavor that included acquiring Otto, a self-driving truck company founded by former Google engineer Anthony Levandowski.

That in-house approach has been largely abandoned. Instead, Uber has repositioned itself as the orchestrator of a sprawling network of partnerships, effectively betting on multiple horses in the autonomous vehicle race rather than breeding its own.

This shift is significant for riders because it changes how and when self-driving Ubers will arrive. Rather than waiting for a single proprietary system to perfect, Uber is integrating autonomous technology from various sources into its existing ride-hail platform.

The company's strategy appears to be agnostic regarding the underlying technology—if a vehicle can drive itself safely and meet Uber's standards, it can join the network. The most ambitious public commitment to date involves Nvidia.

In October 2025, Uber and Nvidia announced a partnership to use Nvidia's Hyperion autonomous platform to accelerate Uber's robotaxi program. As part of that announcement, Stellantis committed to delivering vehicles for the effort.

The target is ambitious: 100,000 self-driving cars on the road starting in 2027. For context, that would represent a substantial portion of Uber's active driver fleet in major markets.

Uber has also put real money behind electric vehicle maker Lucid Motors. The investment supports a premium robotaxi service using robotaxi-ready vehicles from Lucid.

In a related arrangement, Lucid will supply 20,000 Gravity SUVs to Nuro, which will install its own autonomous driving system. This layering of companies—Uber providing the platform, Lucid building the vehicle, Nuro supplying the self-driving brain—illustrates the modular approach Uber now takes.

Why Partnerships Beat In-House Development

The economics of autonomous vehicle development are brutal. Building a self-driving system requires billions in research, testing, and regulatory compliance.

Uber's early forays into this territory proved costly and, at times, tragic. The company's decision to abandon in-house development in favor of partnerships is a recognition that other firms have made faster progress and that capital is better spent on expanding the network than on reinventing the sensor stack.

For riders, this partnership model has a practical implication: the experience of riding in a self-driving Uber will vary depending on which partner's vehicle is dispatched. A ride in a Lucid-based robotaxi will feel different from one in a Hyundai Ioniq 5 powered by Motional or an Apollo Go vehicle from Baidu.

This diversity is a feature of Uber's strategy, not a bug—it allows the company to test multiple approaches simultaneously and scale whichever proves most reliable and cost-effective.

The Testing Landscape Where Self-Driving Ubers Are Already Driving

Autonomous vehicle testing is no longer confined to closed courses or Silicon Valley back roads. Uber's robotaxis are already navigating real traffic in the San Francisco Bay Area, and the company has confirmed that autonomous testing is underway on public roads there.

This is a notable milestone because San Francisco's streets are notoriously challenging, with dense pedestrian traffic, unpredictable cyclists, and steep, winding roads. The vehicle that Uber recently unveiled for this purpose made its first appearance earlier this year, and the company has stated that a public launch is planned for later this year.

Riders in the Bay Area will likely be the first to hail a self-driving Uber as a standard option in the app. Internationally, Uber is expanding its testing footprint.

The company announced in September 2025 that it would partner with Beijing-based Momenta to begin testing robotaxis in Munich, Germany, in 2026. Munich's regulatory environment and infrastructure make it a sensible European testbed.

Similarly, Baidu will start testing its Apollo Go robotaxis in London in the first half of 2026. London presents unique challenges, including narrow streets, heavy congestion, and left-hand traffic, making it a rigorous test of any autonomous system.

What Testing Actually Looks Like

When Uber says it is "testing" self-driving vehicles, it typically means vehicles are operating on public roads with safety drivers behind the wheel, ready to take control if the system encounters a situation it cannot handle. The progression from safety-driver testing to fully driverless operation is gradual.

Companies typically start in geofenced areas—defined zones with mapped roads and known conditions—before expanding to wider regions. The reference material indicates that autonomous testing in specific zones is set to begin next year, suggesting Uber is taking a measured approach.

This is consistent with industry practice: regulators often require companies to demonstrate safety in limited areas before granting permission for broader deployment. Riders should understand that being matched with a self-driving Uber during the testing phase does not necessarily mean the vehicle will be driverless.

Early deployments may still include a safety operator. The fully autonomous experience—no one in the driver's seat—will arrive market by market, as regulators approve it.

Safety Data What the Numbers Actually Show

Safety is the central question for any rider considering a self-driving Uber. The available data paints a complex picture that defies simple conclusions.

Through November 17, 2025, there were 5,202 reported autonomous vehicle incidents in the United States. Of these, 451 involved injuries or fatalities.

The breakdown shows that the vast majority—4,751 incidents—involved unknown severity or no reported injury. Minor injuries accounted for 257 incidents, moderate injuries for 70, and serious injuries for 59.

Looking at the year-over-year trend, incidents have increased as autonomous vehicles have become more prevalent. There were 329 incidents in 2021, 742 in 2022, 921 in 2023, 1,384 in 2024, and 1,793 in 2025 through November 17.

This upward trajectory is expected—more autonomous miles driven means more opportunities for incidents—but it also reflects the growing scale of deployment.

Contextualizing the Numbers

The raw incident counts can be misleading without context. Approximately 7.4% of autonomous vehicle accidents resulted in injury, and 1.2% resulted in a fatality.

But these figures include both fully autonomous systems (ADS) and driver-assistance systems (ADAS), which are fundamentally different technologies. Tesla has reported the most ADAS-related accidents, while Waymo has reported the most ADS-related accidents.

These two categories should not be conflated. Notably, the reference material highlights that robotaxi services operated by companies like Waymo were running at a loss as of 2025.

The operating costs of autonomous fleets—including maintenance, remote monitoring, and insurance—currently exceed fare revenue. This is a business-model challenge, not a safety one, but it affects how quickly services can scale.

The data also shows that consumer hesitation remains a significant barrier. Annual surveys from the American Automobile Association have documented persistent fear of self-driving technology among the public.

Even as safety statistics improve, perception lags reality. For Uber, this means the company must not only perfect the technology but also convince riders to trust it.

Regulatory Frameworks Are Still Evolving

The regulatory landscape is catching up to the technology. The UK's Automated Vehicles Act of 2024 was a landmark step, formally defining liability and safety standards based on the SAE Levels of automation.

For the first time, it was written into law how responsibility is assigned when an autonomous vehicle is involved in a collision. Similar frameworks are emerging in other jurisdictions, though at different speeds.

The United States has a patchwork of state-level regulations, with some states welcoming autonomous testing and others imposing restrictions. The reference material notes that major improvements are still needed before autonomous technology has a significant impact on Uber's business model, and this includes regulatory harmonization.

What Riders Should Actually Expect

For the average Uber user, the arrival of self-driving vehicles will be gradual and, in some ways, unremarkable. The ride-hail app will not change dramatically.

A rider will request a trip, and occasionally a self-driving vehicle will arrive instead of a human-driven one. The price and experience may be similar, at least initially.

The Timeline for Public Availability

The most concrete public commitment is the Nvidia partnership targeting 100,000 self-driving vehicles by 2027. This is an ambitious goal, and it is important to understand that deployment will be phased.

The first vehicles will likely appear in specific cities where regulations permit and where testing has been successful. San Francisco is the current epicenter, with Munich and London following.

The 2027 target does not mean every Uber rider will have access to self-driving vehicles by then. It means the vehicles will be on the road in some markets, at some times, under certain conditions.

Geofencing, weather constraints, and operational design domains will limit where and when robotaxis operate. A self-driving Uber in a sunny Phoenix suburb is a different proposition from one navigating a snowstorm in Chicago.

Riders should also expect the vehicle mix to be diverse. The reference material lists partnerships with multiple companies across different segments.

Avride is deploying Hyundai Ioniq 5 vehicles for robotaxi and delivery services. Aurora is focused on autonomous trucking for freight.

Motional is operating Hyundai Ioniq 5 vehicles on Uber's network in Las Vegas. Mercedes-Benz is working with Nvidia on S-Class robotaxis for the Middle East.

Each of these partnerships serves a different market segment and timeline.

The Premium and Budget Ends of the Market

Uber's strategy appears to cover the full spectrum of the market. The Lucid partnership targets premium riders—Lucid's Gravity SUV is a luxury electric vehicle, and the service will presumably command higher fares.

At the same time, partnerships with companies like Nuro, which operates delivery robots and more utilitarian vehicles, suggest Uber is also preparing for lower-cost autonomous options. This dual-track approach makes sense.

Autonomous vehicles have high fixed costs but low marginal costs per mile. Once the technology is reliable, the marginal cost of an additional trip is primarily electricity and maintenance.

This economics could eventually push fares down, but that is unlikely to happen in the near term, given the capital intensity of deployment and the fact that all current robotaxi services operate at a loss.

Practical Advice for Riders

For riders who want to try a self-driving Uber, the best approach is to be patient and monitor availability in their city. During the testing phase, availability will be limited and unpredictable.

Riders in San Francisco have the best chance of encountering a robotaxi soon. Those in other markets should expect to wait until testing expands.

Riders should also understand the limitations of current autonomous systems. The reference material notes that cars need to be able to handle all conditions before autonomous technology impacts Uber's business model significantly.

This means there will be operational constraints—certain weather conditions, road types, or traffic situations where the system will decline to operate. This is not a failure of the technology; it is a safety feature.

Frequently Asked Questions

Are self-driving Ubers currently available to the public?

Autonomous testing is underway on public roads in the San Francisco Bay Area, and Uber has confirmed plans for a public launch later this year. Availability will be limited to specific areas at first.

Riders in other markets will need to wait for testing to expand.

Will there be a safety driver in the vehicle?

During the testing phase, safety drivers may be present in some vehicles. The progression to fully driverless operation happens gradually and depends on regulatory approval.

Riders should check the app and vehicle information when booking a trip.

How much will a self-driving Uber cost?

Pricing information for self-driving Ubers has not been specified in available sources. Currently, all major robotaxi services operate at a loss, suggesting that fares are not yet covering operating costs.

Pricing will likely be competitive with human-driven rides initially, with potential changes as the technology scales.

How safe are self-driving Ubers compared to human drivers?

As of November 17, 2025, there were 5,202 reported autonomous vehicle incidents in the United States, with 7.4% resulting in injury and 1.2% in fatalities. These figures include both fully autonomous and driver-assistance systems.

Studies comparing autonomous vehicle crash rates to human benchmarks are ongoing, and the safety picture continues to evolve as more data becomes available.

When will self-driving Ubers be widely available?

Uber and Nvidia have announced plans to deploy 100,000 self-driving vehicles starting in 2027. This deployment will be phased and market-specific.

Widespread availability across all Uber markets is likely years away.

Reference Notes

Information in this article is based on publicly available sources. Some details may change over time.

Verify with official sources before acting.