Deep Dive: Uber ($UBER)

Uber cabs are so commonplace now that it’s easy to forget how revolutionary their launch was in the early 2010s. The founders, Travis Kalanick and Garrett Camp, could not find a cab in Paris after a technology conference in 2008. Recognizing this as a business opportunity, they came up with the idea that riders should be able to connect with cab drivers online. That is how Uber started, offering the convenience of on-demand cabs. 15 years later, the platform handles more than 40 million trips a day.

A little theory behind platform companies (which is what Uber is). There’s a lot of research behind aggregation platforms, but we will keep it simple. When supply is generic (meaning supply is more or less similar), a platform can aggregate it and offer it to customers. There is a natural tailwind for such a platform to scale, as it doesn’t need to spend to add suppliers, and both suppliers (seeking to meet high demand) and demand (seeking lower prices) benefit from this increase in scale. Hence, in such businesses, one or two platforms can scale and capture the whole market.

This is what Uber did. It aggregates the generic supply of cabs (there isn’t much difference in service between cabs) and then offers them to users. This basic framework is important, as we will use it to assess the impact of autonomous vehicles on Uber. More on that later.

Uber generated $52B in revenue in FY25 with ~202 million monthly active platform consumers (MAPCs) at the end of FY25, both up 18% YoY. It had gross bookings of $193B (up 19%) and free cash flow of ~$10B in FY25.

In Q2’26, gross bookings grew 24% to $58B, with Mobility up 22% to $29B and Delivery up 26% to $27.5B. Monthly active consumers hit 208 million.

Uber operates across 3 major segments:

The 2 big segments are complementary. 40% of Uber’s consumers use more than one product. Customers who use both products use the app more often and stay longer than those who use only one product. Hence, acquiring a rider is also a cheaper way to acquire an Uber Eats customer, and vice versa. Uber One, the paid loyalty tier that bundles both, has crossed 50 million members.

Another key business driver is advertising, which runs on top of the delivery platform. It is a ~$2B business, growing at >50% YoY in 2025.

Autonomous vehicles can potentially disrupt the Mobility segment. Delivery has a separate issue, which we will address later.

Uber traded just above $100 per share in Sep’25 and is now trading at ~$75. The drawdown is not due to weak results, as bookings grew 24% last quarter (revenue growth was lower but due to an accounting change) and buybacks have reduced the share count by ~4% in the last year.

Mobility: Uber’s moat is deepest in this segment. Here, they own the customer (rider) relationship across thousands of local markets as the app that is the default verb for getting a ride. It has ~75% of US rideshare spending (Bloomberg Second Measure, 2024), and no one has a rival demand network at a similar scale.

This scale is important as it reduces wait times and prices for customers and also enables Uber to improve cab utilization rates for drivers. For example, in areas where Uber has enabled AVs (autonomous vehicles), the average wait time is 3 minutes, while it is ~15 minutes on Tesla’s app. The moat is having the largest platform (most customers). Due to its scale advantage, Uber kept ~30% of every mobility dollar booked (since the supply is generic) in 2025.

Delivery: the moat here is a wide selection (of restaurants/products) plus the network that fulfills the order. The segment aggregates 1.3 million restaurants, grocers, and retailers. It fills its orders from the same driver pool as Mobility and locks customers in through Uber One, whose 50 million members now drive around half of combined Mobility and Delivery bookings. Customers on both Mobility and Eats order more and churn less, so the 2 segments defend each other. Advertising monetizes the customers at very high margins and is now a >$2B business.

Freight: This segment matches shippers with carriers and takes a cut. It has had roughly flat revenue for 3 years. Bookings grew 25% last quarter (due to US-Mexico trade flows), but it still posted a $24M operating loss (improved by only $2M YoY). At 10% of Uber's revenue and 3% of bookings, it is not material to the thesis. We don’t see Uber having any real moat here.

Uber saw AVs as a threat to its business model very early. In 2015, Uber hired around 40 researchers from Carnegie Mellon University’s robotics lab and built the Advanced Technologies Group. In 2016, they bought the self-driving startup Otto for about ~$680M. They scaled the AV setup with Travis Kalanick calling autonomy ‘basically existential for us’. Unfortunately, in 2018, an Uber test vehicle killed a pedestrian. After this, the program never recovered.

In December 2020, Uber sold ATG to Aurora (whose founder used to run Google’s AV program) and invested $400M in cash for a 26% stake in the combined company. Aurora currently focuses on autonomous trucking and is not competing for the AV cab market.

So, Uber currently does not have its own autonomous driving technology. They tried for nearly 6 years. The current management team decided to exit the AV development business, whereas the founder felt AVs were existential to Uber. This is key to remember because Uber’s management maintains that Uber can thrive even without owning the AV technology.

Let’s dig into Travis’s logic. He said that if someone else’s AV became cheaper than Uber’s drivers, then Uber would be aggregating the wrong supply. This is platform theory. If the supply is not generic (i.e., one is cheaper than the rest), then aggregators need access to the non-generic suppliers. As we highlighted before, Uber did give AV development a good try but did not succeed. Developing AVs is difficult, and many companies have failed. Lyft in 2021, Ford and Volkswagen in 2022, and General Motors in 2024. The high-technology nature of AVs will in all likelihood make AV supply non-generic. Let’s look at those suppliers:

These companies are following different methods to develop AVs, and it is difficult to say who will win. But the direction is now clear. Self-driving cabs will scale massively in the coming 5 years. This affects the Mobility segment’s moat.

Returning to the framework at the start of the article. Aggregation works when suppliers are generic (the same), abundant, and unable to reach customers on their own. Autonomous vehicles flip all 3 of those pillars. The supply is differentiated with very few providers. The key for customers will be safety. Waymo reports ~90% fewer crashes with serious injuries than human drivers. A study published by Waymo with Swiss Re, using Swiss Re's own claims records, found ~92% fewer bodily-injury claims. Note that this is company-produced data, as is Uber's utilization claim below. We weigh it more heavily only because the underlying claims data is a third party's.

We think people will gravitate towards platforms that offer better safety. Tesla also regularly releases crash data for its supervised consumer system, which it says is ~8x safer than human drivers.

How many AV winners there are decides everything. If only 1 or 2 providers are good enough for people to trust, they will be scarce, and a scarce supply does not need an aggregator. In this case, an owner of robotaxis, such as Waymo or Tesla (both already have consumer apps in millions of hands), does not need an aggregator like Uber. This will put Uber’s Mobility take rate at risk. If there are 5, then Uber is in a better position.

We bet that the first outcome is more likely. Waymo and Tesla are way ahead in the US, and everyone else, including Uber’s partners, is 2-3 years behind on scaled safety data.

The key question is when AV supply at Waymo and Tesla scales to the point that it affects Uber. We think this happens in 2028.

Until this year, the car was a bottleneck. Waymo’s Jaguar cost ~$200K fully fitted out, and the fleet grew by under 300 per month. Its new Zeekr-built Ojai car carries a 6th-generation driving technology with 42% fewer sensors and costs $125k. The plant making these cars will soon scale up to 10s of thousands of cars. Waymo will also use Hyundai Ioniq 5s in 2026, with a reported 50k by 2028. Waymo is at ~3,900 vehicles today.

Tesla has a Cybercab capacity above 125k per year and will look to scale in the coming years, after many years of claiming their technology is approaching maturity. This is supposed to be a ~$30k car.

Both the cost of cars and regulations, which were major constraints over the last 5 years, are now clearing up (NHTSA granted the first no-steering-wheel exemption in July 2026 and is writing national AV rules). What stays slow is the city-by-city build: Waymo took ~20 months to reach a quarter of San Francisco’s market. In New York, which is over 10% of Uber's US trips, AVs are banned.

Waymo quadrupled its number of rides in 2025 and is guiding to 1 million rides per week by the end of 2026, 2x its current rate. If this expansion rate holds and Tesla scales as well, the AV fleet from these two companies will be sufficient to absorb the growth in rides in Uber’s most profitable US markets. This is the type of change that will happen slowly and then suddenly (due to overwhelming safety data). At RC, we will not bet against Google and Tesla.

Uber’s strategy is to be the demand and capital provider for every startup or company innovating in AVs that is not Waymo or Tesla. It has partnerships with more than 30 autonomous vehicle companies and is live with them in 7 cities. They expect to be live with AV partners in 15 cities by the end of 2026. The launches for the rest of 2026 include partnerships with companies like Nuro and Lucid in San Francisco, Zoox in Las Vegas, Baidu in the Middle East, and Wayve in London.

Uber has committed >$10B to autonomy with ~$7.5B for fleet procurement (buying cars is a major expense for AV companies) and over $2.5B in equity for software partners. The CFO has said that for every dollar partners raise from Uber, they raise $2.5 from outside investors. This was a smart strategy when cabs cost >$200k. But with cabs costing ~$125k and falling, the constraint is easing. Also, note that Uber's total commitment is $10B, roughly a third of Waymo's cumulative funding of ~$27B. In our view, the industry’s market share will not be decided by who has more capital.

Uber is also buying the value chain around AV technology. It has raised its Lucid commitment to at least 35k vehicles and its stake to 11.5% for $500M, with Nuro providing the AV tech. It invested $100M in WeRide, and Nvidia plans to deploy 100k Level 4 robotaxis on the platform starting in 2027. Uber is trying everything, from owning the cars and riders to renting AV technology, by any means necessary. But will a fleet on rented software survive the best AV technology for safety? We have our doubts. Anyways, these Uber partners will go at it alone if their technology becomes industry-leading. Uber is playing defense and has no way to attack its competition.

Uber’s core argument is that it benefits from the favorable economics of a large hybrid fleet on its platform. It can reduce idle time for cabs on its network, whereas AV fleets from vertically integrated players will have lower utilization. This is the claim we examine first below, and it is one we reject.

There are 3 broad strategies adopted by companies: own everything, license the driver, or aggregate demand.

Uber made the case for their strategy in an AV presentation alongside its Q4 2025 results. Here is our take on each of its claims. We think Uber is right about the things that buy it time (profit pool, edge cases) and wrong about the one thing that decides where this ends (utilization).

We think the way Uber tries to explain why it doesn’t need AV technology lays bare the weakness in its business model and why the bear case is a distinct possibility.

One argument would break our thesis: AV technology becoming a commodity. If you can just buy an AV stack, then AV supply becomes generic again, and Uber wins out. Baidu, WeRide, and Nvidia are all building AV technology and would be happy to license it out. We do not think this will happen anytime soon because safety is both an AI and a hardware problem, and it is difficult to solve.

The key moat in the delivery business is having the largest selection on your app. The app with the largest number of restaurants or retail outlets has an advantage in 2 ways. One it has the most variety to offer its users and secondly, they can run their delivery partners at the highest efficiency due to the density of their operations. This causes the delivery business to consolidate towards the market leader in particular areas/cities.

Currently, the leader in the US is DoorDash. DoorDash has roughly two-thirds of the US food delivery market, while Uber Eats holds ~23%. DoorDash has scaled into Europe as well with the acquisitions of Deliveroo (in 2025) and Wolt (in 2022).

Uber's response was to acquire Delivery Hero for $14.8B ($13.7B net of its existing stake), roughly doubling the markets in which it operates, both Mobility and Delivery. But notice where the growth is. Uber is keeping MENA, Asia, and Latin America, and selling 14 overlapping markets to SSW Partners to clear antitrust. This deal buys emerging markets, not European market share.

In the US, Uber is a distant number 2 to DoorDash. In Europe, DoorDash expanded its footprint through Wolt and Deliveroo, but Uber Eats leads in France and has been gaining share in the UK and Germany. It is a three-way fight there with no clear winner.

We model Uber to continue performing well in this segment and don’t model disruption in any scenario. But high competition will limit profitability in this segment. In our disintermediation (Base Case), delivery margins expand to 4.3% of bookings, and in the Moat Holds case, they reach 5.4% of bookings.

Next up, we model our thesis.