Who owns
the trip when
the driver
disappears?

Autonomy removes the driver from the vehicle, but insurance, financing, cleaning, maintenance, remote assistance, and empty miles remain. A lower production cost becomes a demand catalyst when it translates into a lower customer fare.

For Uber, the decisive question is where that next ride is booked. If the rider opens Uber, the marketplace can choose between suppliers. If the rider opens an AV operator’s own app, Uber can be bypassed. A growing market helps Uber only to the extent it retains the relationship and earns attractive economics.

The question has therefore moved beyond whether a vehicle can complete a journey. It now includes whether riders return, how fares compare, what happens at busy times, and whether the service can expand without sacrificing economics.

Keep three categories separate throughout this case: an announced partnership, an operating service with a safety driver, and a paid driverless service. Each represents a different level of progress. A launch in a limited operating area is also different from reliable citywide coverage.

Local success can matter to Uber before autonomy becomes a large share of global trips. Losing valuable customers in a few dense areas could weaken future earnings even while the overall marketplace grows.

A private ride, without the small talk.

For some riders, the absence of a driver is part of the product: privacy, no small talk, and time to watch a video on their own phone, work, or simply be alone. Others may prefer a human driver for assistance or reassurance. These preferences can influence which service people choose, even when fares are similar.

This cuts both ways for Uber. A compelling Waymo or Tesla experience can build loyalty to a direct app. When a comparable autonomous ride is available through Uber, that same appeal can strengthen Uber’s offering. The advantage belongs to the app the rider chooses, not automatically to the company that built the vehicle.

The competition now includes purpose-built vehicles.

Tesla began limited Cybercab rides in Austin on September 3, 2026. The vehicle has no steering wheel or pedals. NHTSA opened a certification audit around the launch; an audit is not a finding that the vehicle violates safety rules. The limited rollout should not be confused with a fleet operating at Waymo’s scale. [9]

Amazon’s Zoox began paid rides in Las Vegas in August 2026. Chinese autonomy developers are also reaching overseas: Baidu’s Apollo Go has launched fully driverless service on Uber in Dubai, while Uber and Pony.ai have announced plans for more than 2,000 robotaxis in Europe. A launch and a future fleet commitment are different stages of progress. [10] [11] [12]

Waymo is already taking meaningful share inside its service area.

Yipit data reported in June 2025 showed Waymo ahead of Lyft in San Francisco within Waymo’s operating area. This comparison covers trips inside the service zone, not the whole Bay Area or every airport journey. It is evidence that riders will adopt a direct autonomous service when coverage works for them. [13]

The alignment is conditional. An operator may initially use Uber to fill vehicles, then prefer direct distribution once its own customer base is large enough. Uber may help create a market without retaining all the value it helped create.

For a supplier without a strong local consumer brand, Uber could shorten the route to customers. For a supplier that already attracts sufficient demand, Uber’s fee must be justified against the cost of acquiring and serving riders directly.

That is a continuum, not a permanent division between strong and weak technology companies. A partner can become a competitor. An operator with a strong brand can still use third-party distribution in a new geography.

Geography also changes the exposure. Uber’s FY2025 presentation attributes approximately 30% of U.S. Mobility bookings and 25% of its profits to its top 20 Mobility cities. Its definition counts trips beginning and ending within Census city limits. [3]

This supports diversification beyond the largest city cores. It does not measure exposure across entire metropolitan areas, and it does not make those cores expendable. The investment case needs both the resilience of the wider business and the potential importance of concentrated losses.

Waymo wants riders to book directly.

Waymo’s own app in established markets, alongside partnerships elsewhere, suggests it prefers the direct customer relationship where it has enough demand. That is an interpretation of its distribution choices, not a rule that every dense market must follow. Uber remains useful when the extra trips and operating support justify its fee.

Tesla can build cars faster than it can build a transport network.

Tesla’s manufacturing capability is a serious advantage. But a production line measures cars built, while a ride-hailing business depends on paid trips completed. Even very rapid Cybercab production would leave three separate tasks: operate the fleet, obtain permission to serve each market, and secure access to valuable pickup locations.

Fleet operations. Charging, cleaning, repairs, insurance, remote assistance and customer support all affect how many hours a car earns revenue. Manufacturing scale does not by itself establish that Tesla can deliver reliable, profitable robotaxi service at comparable scale. Tesla’s own September 2026 support page describes Cybercab availability as limited areas of Austin. [17]

Permission to operate. California’s published DMV list includes Tesla Robotaxi under testing with a driver, but not driverless testing or deployment. Commercial driverless passenger service also requires CPUC authority. A supervised ride service therefore does not demonstrate approval for an unsupervised taxi business. The published lists show permissions granted, not every application that might be pending. [18] [19]

Access to the pickup. Permission to drive on a road does not automatically authorize airport passenger service. California explicitly requires the airport authority’s approval. SFO’s phased Waymo permit illustrates the additional operating agreement, safety conditions and reporting involved. Hotels and gated properties can create their own access arrangements too; these need to be assessed location by location. [19] [27]

Politics can slow expansion even when the technology works.

Washington, D.C. offers an example of the debate. An AV bill introduced in May 2026 prompted discussion of fleet limits, a per-mile charge and restrictions around the timing of paid service. Reporting on the July hearing describes a proposed 200-vehicle cap through January 2028 and a 15-cent-per-mile charge. These are proposals, not nationwide rules or proof that all commercial service must wait until 2028. [25] [26]

The investment implication is a rollout market by market, shaped by safety, congestion, jobs and local economics. A factory ramp cannot remove those negotiations. It also does not mean that a larger Tesla fleet would have nowhere profitable to operate.

The scale gap is large, but the measures are different.

Reuters reported 420 Tesla autonomous vehicles registered in Texas, including 45 Cybercabs, on September 3, 2026. Registration does not establish how many cars are actively serving passengers or how many paid trips they complete. [22]

For context, Uber reported more than 40 million daily trips globally at its Q4 2025 run rate. That includes Mobility and Delivery, rather than U.S. robotaxi rides alone. This shows the breadth of Uber’s existing marketplace; it is not a market-share comparison or a basis for calculating Tesla’s daily rides. [4]

A low-cost robotaxi may not remain a uniquely Tesla product.

Tesla has discussed a Cybercab selling price below $30,000. That is a price ambition, not a verified all-in operating cost. Chinese competitors have also announced inexpensive vehicles: Baidu’s RT6 was reported at a 204,600-yuan invoice price in May 2024, about $28,350 at the time. The specifications, commercial terms and operating costs differ, but low vehicle prices are not exclusive to Tesla. [23] [24]

The U.S. is also a protected market. USTR’s 2024 action raised Section 301 tariffs on Chinese EVs to 100%, while BIS restrictions cover certain Chinese- and Russian-linked connected vehicles and software, with key prohibitions beginning in model year 2027. Tariffs and technology restrictions are separate barriers. They limit competition; they do not establish that protection is the only source of Tesla’s potential cost advantage. [20] [21]

Where more suppliers can compete, the bull thesis is that autonomous driving becomes increasingly interchangeable from the rider’s perspective. That could weaken suppliers’ pricing power and strengthen the app that brings demand. This is an industry-structure scenario, not an established outcome: safety, reliability, brand and approvals can still differentiate operators.

Uber’s Autonomous Solutions offering includes mapping, venue management, regulatory support, fleet operations, and training data. These are company-described capabilities that extend the partnership proposition beyond booking distribution. [5]

Separate three potential sources of value. Marketplace information may improve positioning and matching. Local operations may make a service easier to launch and run. Driving data may help partners develop their technology.

The first two fit the existing marketplace logic. The third needs separate evidence of partner adoption, differentiated data quality, and monetization. A large dataset is not by itself proof of a profitable product or technical superiority.

Uber knows where the next ride will come from.

Uber sees where requests appear, when demand spikes, and how price and waiting time affect bookings. That knowledge helps decide where a vehicle should wait and which trips it should serve. Its local operations help translate the prediction into a completed ride.

That is a structural advantage for an aggregator. An AV developer with little local brand recognition may have excellent driving technology and still struggle to fill its cars. Uber offers an existing stream of customers and the operations to serve them. Waymo and Tesla may need less of that help where they can attract the rider themselves.

AV Labs is an additional opportunity, with a separate burden of proof.

Uber describes AV Labs as a way to use real-world driving data to support partners’ training and evaluation, including unusual situations. That can reinforce its commercial relationships. It does not yet demonstrate a decisive training-data advantage over Waymo’s own autonomous miles, or establish data licensing as a large source of profit. The core case is the marketplace’s demand intelligence and operating capability; a major standalone data business would be additional upside. [16]

This gives the company resources to invest while the industry structure develops. It does not mean all reported free cash flow is available for repurchases. Equity investments, acquisitions, and some future contractual commitments can use cash outside the capital-expenditure deduction in the reported measure.

The partnership strategy should therefore be evaluated through deployment and returns: vehicles actually operating, incremental contribution, incentives paid, financing responsibility, and cash committed. Signed vehicle commitments are not equivalent to active supply.

Delivery provides a second business engine, but a large share of bookings is not the same as a large share of profit. Each segment needs its own economics.

Volatility makes Uber’s network more valuable.

An AV fleet is committed capital. Depreciation and financing costs continue when a vehicle waits in a depot. Buying enough cars for the rare peak can leave too much capacity during ordinary hours. Buying for recurring demand leaves trips uncovered when the city gets busy.

Human drivers provide the flexible layer. Surge pricing and incentives can make a busy period worth working, encouraging drivers to join or move toward demand. They can log off when the opportunity fades. Uber generally does not bear the depreciation of a driver-owned car when that driver is offline, although incentives, guarantees and platform overhead still cost money.

Combine the two and the advantage becomes clear: AVs serve recurring trips, while humans help absorb peaks and coverage gaps. Uber can offer broad availability without requiring its AV partners to purchase enough permanent capacity for every spike.

Buying more cars does not buy Uber’s demand history.

Uber has spent more than a decade learning where riders appear, how events change demand and how prices and waiting times affect bookings. Its dispatch systems use that history to position supply and complete trips. A fleet operator can buy vehicles quickly; building comparable customer relationships and operating knowledge takes time.

The bull case is that this combination becomes a durable competitive advantage: existing riders, demand intelligence, flexible human supply and multiple AV partners. Uber is unusually well positioned to run it at global scale. Other networks can pursue a hybrid model, so the case rests on Uber’s execution and reach rather than literal exclusivity.

More competing AV suppliers can strengthen the marketplace.

Uber has assembled a broad group of autonomy, vehicle and fleet partners. It can introduce different suppliers in different countries instead of relying on one technology. If several systems deliver a similarly good ride, availability and price become more important, and the marketplace can direct demand toward the best offer. [5] [11] [12]

That is the positive investment outcome: cheaper autonomous supply expands demand, Uber keeps the booking, and the flexible network turns more of those trips into cash. Direct services can grow too. Uber does not need every autonomous trip to pass through its app for this case to work.

Growth has to leave more money for shareholders.

More rides are useful only if Uber keeps enough of the revenue after paying partners and covering costs. If autonomous supply expands the service while improving those economics, the business can become more valuable.

The market also needs a reason to worry less.

Investors may fear that Waymo or Tesla will take Uber’s customers. Evidence that riders keep opening Uber, even as AVs scale, would weaken that concern. A more durable earnings outlook can make investors willing to pay more to own the business.

What would undermine the case?

If Uber loses valuable trips, accepts lower fees or spends heavily to secure supply, its cash generation could disappoint. Even strong performance may not lift the stock if investors already expected more. The positive case is better economics and greater confidence in their durability, not simply more autonomous vehicles on the road.

A gradual robotaxi rollout gives Uber time to keep growing.

The positive case is that Uber continues growing trips and cash generation while AV fleets expand through individual markets. Deployment takes vehicles, approvals, operating capacity and access agreements. That creates time for Uber to introduce partner supply, learn from pilots and improve its service.

If the business sustains double-digit growth and keeps enough of the resulting cash, shareholders can benefit without Tesla’s robotaxi ambitions failing. Tesla could build a valuable business over time while Uber also grows in a larger market. Growth in trips must still translate into cash after incentives and AV investment.

The fear can ease before competition disappears.

If investors currently expect rapid displacement, evidence of a slower transition and a successful Uber hybrid model could improve confidence in Uber’s earnings. That could support the share price alongside business growth. It does not establish that today’s stock is cheap, and the market may already anticipate part of that outcome.

Watch the deployment evidence: paid driverless trips, repeat customers, partner returns and Uber’s cash commitments. Factory production targets alone cannot settle the investment case.