The Landlord Who Can’t Afford a Winner

Uber sold its autonomous vehicle development unit in 2020. Uber does not build the technology. It builds the floor the technology stands on.

That distinction matters more than it looks. Because the floor Uber built is only valuable as long as the tenants need it. And the structural question that two years of partnership announcements have not answered is this: at what point does each partner’s need for Uber’s platform diminish, and what is Uber’s position when it does?

Uber brings 202 million monthly active customers and the fleet utilization density that number enables. For an AV operator trying to go direct, that density is the hardest problem.

Uber charges partners a platform take rate on every ride. The precise terms vary by deal and have not been publicly disclosed. The arithmetic case for partners remains the same: Uber’s demand density can improve fleet utilization enough to make the platform fee worth paying, before spending a dollar on customer acquisition.

The cold-start math is unforgiving. Going direct means absorbing billions in demand generation costs that Uber has already paid. And the AV cost curve, while declining, has not yet broken below the human-driven baseline at commercial scale. Uber’s platform is most valuable precisely during this window, before the curve crosses. Standalone operators that try to build direct demand before that crossover absorb both the technology risk and the customer acquisition cost simultaneously.

In February 2026, Uber formalized this into a named product suite. Uber Autonomous Solutions goes beyond marketplace access. It packages fleet management, depot operations, vehicle cleaning, real-time monitoring, product development support, customer support across the full trip lifecycle, fleet financing, AV-specific insurance, and on-road assistance when autonomous systems encounter edge cases.

For AV companies that have built strong technology but thin operational depth, Uber is offering to be the commercial operating layer. That is a real service. It solves a real problem.

The question is what it costs to accept it.

The Uber-Rivian partnership is a useful example of how this architecture actually works, because its terms are unusually transparent.

Uber plans to invest up to $1.25 billion in Rivian over five years. The initial tranche is $300 million. Bloomberg Opinion’s Liam Denning framed the deal as the $300 million being oughly what Rivian burns through in a month, but a sum Uber generates every ten days. The subsequent tranches are milestone tied to Rivian’s autonomous technology development which include autonomous vehicle deployment across 25 cities, launching in 2028, exclusively on the Uber platform.

Uber CEO Dara Khosrowshahi praised Rivian’s vertical integration, vehicle, compute, and software stack. But Uber controls the distribution.

The deal raises a structural question that the announcement does not answer. When Rivian’s autonomy milestones determine whether Uber releases subsequent capital tranches, the leverage in that relationship is with Uber. The milestone gates give Uber structured off-ramps if Rivian’s technology does not perform. Rivian is financially dependent on tranche release. Uber controls the timeline.

Yet, another reading is if Rivian’s technology develops ahead of schedule and alternative platform options mature. In that environment, Rivian’s need for Uber’s capital diminishes where only the exclusivity window becomes the binding constraint on Rivian.

There is another part of Uber’s strategy that is less obvious.

Uber is not simply partnering with AV companies because it wants to offer riders more choice. It is partnering with all of them because a fragmented AV market is the outcome that preserves Uber’s leverage.

If one AV player achieves dominance, that player gains pricing power over Uber. The dominant provider can renegotiate a better deal, demand exclusivity on its own terms, or bypass the platform entirely. Uber’s platform fee becomes a negotiating position rather than a settled arrangement.

If the market stays fragmented across a dozen competing providers, Uber becomes the essential marketplace. Access to the largest rider pool means no single partner can afford to walk away.

In January Uber launched AV Labs, a new division that will collect real-world training data for its partners. The data is currently offered without charge. Uber’s CTO told TechCrunch that the value of advancing the AV ecosystem outweighs near-term revenue from the data. Uber has said it plans to eventually charge for the data. That the offer is free today, fits the architecture of accelerating as many partners as possible to preserve competition among them, and keep Uber essential to all of them.

Uber currently handles approximately 1.5 million AV rides annually across partnerships with nearly 20 AV firms. The diversification is simultaneously a hedge against any single partner’s failure and a structural defense against dependency on any individual partner.

This strategy only works if none of them win decisively. That is not a flaw in the architecture. It is the architecture.

Waymo is not a case study in a partner outgrowing Uber. It is an AV operator that built its own demand base from the start.

Waymo currently operates commercial robotaxi service in ten U.S. metro areas. Eight of those are run entirely through its own Waymo app. Only Austin and Atlanta run exclusively through Uber's platform. The company has said it is on track to serve over one million rides per week by the end of 2026 and is laying groundwork for more than twenty cities, including international expansion to London and Tokyo. None of those announced cities involve Uber.

The Austin and Atlanta arrangement is worth examining for what Waymo actually gets from it. Uber provides demand access through its app and manages fleet operations. Waymo handles the autonomous technology, roadside assistance, and rider support functions.

The tradeoff is clear. Waymo gets access to Uber's rider base, but gives up the direct customer relationship.

An arrangement in Dallas shows what Waymo chose when it had the option to structure things differently. There, Waymo partnered with Avis for end-to-end fleet operations which includes infrastructure, vehicle readiness, maintenance, charging, and depot management. Riders hail a ride through the Waymo app.

Avis brings decades of fleet management experience. Avis is Waymo’s fleet vendor. In Atlanta and Austin, Uber is a platform that sits between Waymo and the rider. In Dallas, Waymo chose the vendor.

This does not mean Uber's partnership model is broken. It means the model's value proposition is clearest for partners that have not yet solved the demand problem on their own. Waymo is solving it. That is the structural signal useful for understanding Uber relative to the other AV firms.

Uber appears to understand the dynamic. The Rivian investment, the Lucid/Nuro deal, and the full Uber Autonomous Solutions suite all show Uber moving from pure distribution toward deeper structural integration - fleet financing that creates capital dependency, milestone-gated tranches that give Uber schedule control, exclusive deployment windows by market, and operational infrastructure that is difficult to replicate independently. These are mechanisms designed not for a partner like Waymo, but for partners that are not yet Waymo.

Uber's press releases describe every partnership as mutual. The economics suggest something more layered. For partners in early commercial deployment, Uber's offer solves a real problem: demand density, operational infrastructure, and a skip-the-line path past billions in customer acquisition costs. The dependency that comes with it is also real. Milestone-gated capital, exclusive platform commitments, and operational integration all create switching costs that compound over time.

The structural question is not whether Uber's platform is valuable today. It is whether the terms accepted today constrain the options available in three years, when the technology matures, when the cost curve crosses, and when the partners who accepted Uber's full suite of services discover what Waymo already demonstrated: that the demand problem, once solved, does not require a landlord.

If you have a perspective or disagreement, reply directly. I read every response.