The problem
According to TNW and TechCrunch, both citing Bloomberg, Waymo co-CEO Tekedra Mawakana told Bloomberg’s The Circuit on 4 November 2026 why Waymo will launch its own app in Austin and Atlanta in January 2028. Waymo’s own site says those are the two cities where Waymo is available only through Uber. Her explanation, as quoted by TNW from the Bloomberg interview, was blunt: “In the early days, we would partner because we didn’t have the scale, and in that partnering, it was a little bit of a crapshoot. Will I get a Waymo? Will I not get a Waymo?” These outlets are relaying one Bloomberg report, not confirming it independently.
Most coverage reads this as a split. The record is more specific. When Uber first disclosed the plan in July, its spokesperson told CNBC that Waymo would launch its app “alongside their existing deployment with Uber,” and that hundreds of Waymo robotaxis would stay on Uber through at least May 2028. TechCrunch later corrected its own story to say Mawakana’s comments were about ending exclusivity. Uber says it plans to keep hosting Waymo for as long as both companies agree.
So the decision worth analysing is not “leave Uber.” It is the booking flow, and Waymo’s own help page describes it exactly. A rider in Austin or Atlanta turns on a Waymo preference in Uber’s settings, requests an ordinary UberX, Uber Electric (Austin only), Uber Comfort or Uber Comfort Electric ride “for the chance to get matched,” and gets a pop-up if a nearby Waymo is assigned. A match is possible only “if you’re in the service area and there’s a vehicle nearby.” The fleet “prioritizes ride requests that are close by to an available Waymo vehicle,” and the page says plainly: “we aren’t able to guarantee Waymo rides on the Uber app every time.”
That is a lottery, not a booking. According to TNW and TechCrunch, both citing Bloomberg, Uber said it had asked Waymo to add a dedicated robotaxi option in the two cities, and Waymo said no. TNW, also citing Bloomberg, reports that Waymo declined to comment on the talks. TechCrunch says Waymo did not immediately respond to its own request for comment.
What I concluded
The disagreement, as I read it, is about who owns the rider’s request for a robotaxi.
A dedicated “robotaxi” button would let a rider explicitly ask for a Waymo. It would not by itself guarantee a match, a vehicle, an ETA or a completed trip; the service area and the number of nearby cars would still decide that. What the button changes is expectation: the rider has asked for something specific, and the app that took the request owns the answer. Uber’s own prepared remarks for the same quarter treat it the same way. In Las Vegas, where Zoox and then Motional are due on Uber, the company plans “to introduce a dedicated autonomous option in our app as availability and reliability improve.” Uber treats the button as something a fleet earns once availability and reliability are good enough. I think that is the right standard.
Seen that way, Waymo turning the button down on Uber is consistent, not stubborn. If the explicit request “I want a Waymo” lives behind an Uber button, Uber owns the rider’s expectation, the ETA shown and the support flow, and Waymo becomes the supply behind someone else’s front door. In discussing its own app, Waymo emphasized the same point: “It’s important for us to have that relationship with the rider,” Mawakana said. She also said price is “not the first priority,” and that Waymo is not trying to have “this race to the bottom.” Beyond those quotes, the sources I checked don’t give either company’s business reasons, so I won’t supply them.
Here is my speculation, and it is only that: the lottery may also suit Uber’s economics. Uber says its U.S. AV partners get about 30% more trips per vehicle per day on its network than in first-party deployments nationwide. On the call, the CFO clarified that this compares Uber’s hybrid U.S. deployments with mature first-party deployments that include San Francisco, which he called “a uniquely favorable market.” He added that for Waymo and Tesla, “trips per vehicle per day in the new markets that they have launched this year is in the low teens versus high 20s that we are seeing on our network in these hybrid markets.” Those comparisons span different cities, fleets and launch ages, so they do not isolate the effect of the missing button. Separately, Uber’s remarks give an illustrative example, not a measured result: holding fixed costs constant, a 30% rise in paid trips per vehicle would cut fixed cost per trip by about 23% (1 ÷ 1.3 ≈ 0.77). My guess is that matching an AV to whichever nearby eligible request comes along helps keep a car busy. Nothing I found shows that this is why the button is missing.
So my conclusion: Waymo is right to put the explicit request in its own app. Both entry points are expected to coexist from January 2028 until at least May 2028. The sources I checked don’t say whether the same vehicles will serve both apps or whether each gets a separate allocation. If supply is shared, the dispatch rule between the two doors becomes the real product spec.
Constraints
The first constraint is the contract calendar. The Waymo app arrives in January 2028. Uber says the existing deployment runs through at least May 2028, and TechCrunch, citing the FT, reports the deal had been set to expire then. So for at least four months, two entry points are expected to coexist. If they draw on one pool of cars, the same vehicles would receive requests from riders who chose Waymo and from riders who only asked for an UberX. If the cars are split between the apps, that conflict doesn’t arise. The sources I checked don’t say which it is.
The second constraint applies only if supply is shared: a lottery works best with idle cars. The Waymo help page says matches favour requests close to an available car. If Waymo’s own app had a queue and the fleet were shared, every car sent to a nearby Uber request would be a car not sent to a rider who explicitly asked for Waymo. Under that assumption, the trade-off gets sharper as demand grows, not easier.
The third constraint is Uber’s leverage. Ending exclusivity cuts both ways: CNBC reported in July that it lets Uber put other autonomous vehicles on its platform in Austin and Atlanta. Uber says it will have about 700 fully driverless vehicles on its network by the end of 2026 and more than 2,000 by the end of 2027. Those are network-wide figures; the sources I checked don’t say how many will be in these two cities. My reading is that if a competing fleet there accepted a dedicated button on Uber, the “crapshoot” would start to count against Waymo on Uber’s app.
The fourth constraint is precedent, and it is limited. According to EV, Phoenix, the first Waymo-on-Uber market, wound down this summer, and Uber called it “intentionally limited,” just over a dozen vehicles. EV also reports that in Nashville, Waymo rides went live on Lyft on 9 September 2026, with the same fleet serving Waymo’s app and Lyft’s, managed by Lyft subsidiary Flexdrive. Nashville shows that one shared-fleet setup exists. It does not show that Austin and Atlanta will use the same arrangement, and the sources I checked don’t say how requests are split in Nashville either.
The options, and what I traded away
This is my reconstruction of the option set, not anyone’s internal reasoning.
| Option | What it buys | What it costs | Verdict |
|---|---|---|---|
| Keep the lottery on Uber only | Can tap Uber’s demand pool | Riders cannot explicitly request Waymo; Waymo has no direct rider relationship | Rejected by Waymo |
| Add a dedicated robotaxi button on Uber (Uber’s ask, per Bloomberg) | Riders can explicitly request a Waymo, subject to service area and car availability; the demand stays on Uber | Uber owns the request, the ETA shown and the support | Declined by Waymo, per Uber |
| Own app plus Uber lottery side by side (the plan) | Direct relationship; Uber stays a second channel | If supply is shared, two front doors compete for the same cars; the sources checked don’t state a rule | Right call; allocation not stated |
| Leave Uber entirely in January 2028 | One entry point, one app | Gives up access to Uber’s demand pool | Not what Waymo announced |
| Own app first, Uber as overflow with a published priority rule | The explicit request stays with Waymo; Uber gets only genuinely idle cars | Lower Uber volume; Uber may lean on its other partners | What I would do |
Evidence
The strongest evidence is the two companies’ own words. Waymo’s help page says it cannot guarantee a match. According to the Bloomberg interview as relayed by TNW and TechCrunch, Mawakana calls the result a crapshoot and says riders want one consistent app. Per the same Bloomberg report, as relayed by TNW and TechCrunch, Uber says it asked for a dedicated option and was refused. Waymo declined to comment, according to TNW. In its prepared remarks, Uber says it will add that dedicated option in Las Vegas only “as availability and reliability improve.” Those are the stated positions. The sources I checked don’t show what either company thinks privately.
The utilization numbers are Uber’s, and I treat them as an argument rather than a measurement. On the call, the CFO said the 30% figure compares hybrid U.S. deployments with mature first-party deployments, a mix that includes San Francisco. His “low teens versus high 20s” compares trips per vehicle per day in Waymo’s and Tesla’s newly launched markets with Uber’s hybrid markets, so it isn’t a like-for-like fleet. The 23% cost figure is Uber’s illustrative arithmetic, not a reported decline. Neither number separates out the effect of how Uber matches cars to requests.
Two facts on scale come only via Bloomberg, as relayed by The Next Web and EV, so I keep them as background: about 4,000 Waymo vehicles and more than 500,000 trips a week. These are company-wide. The sources I checked don’t say how many of those trips happen in Austin and Atlanta, so I can’t say how much volume the Uber lottery carries.
Where this stops being true
Allocation: if Waymo assigns separate vehicles to its own app and to Uber, there’s no shared-supply dispatch question at all, and the rest of this analysis applies only to how the cars are split. Supply: if the fleets are shared but large enough that Waymo’s own app rarely queues, the conflict mostly disappears and the Uber lottery becomes pure overflow. My concern applies only while a shared fleet is supply-constrained. Partner: if Uber’s other AV partners are slow to reach those cities, Uber has little leverage and Waymo can keep the lottery as long as it likes. Contract: if the two companies agree, before January 2028, on terms that let Waymo set priority for its own app, the question is answered privately, though riders still won’t see the rule. Evidence: if Uber publishes a like-for-like fleet comparison showing that the same Waymo cars earn materially more on Uber, Waymo’s “experience first” stance would have a measurable cost that Waymo would need to justify. And if Waymo exits Uber in May 2028, any two-door overlap lasts only a few months.
What I would do differently
Publish the allocation and the dispatch rule before launch. A short statement would do: are the Uber cars a separate allocation or the same pool? If it’s the same pool, when a Waymo-app rider and an Uber request are both waiting, which one gets the car? Is Uber served only by vehicles with no Waymo-app demand nearby? Riders who switch apps in January deserve to know what they are switching to.
Publish one availability number for the Waymo app in Austin and Atlanta, such as the share of requests inside the service area that are served within a stated ETA, and report it monthly. Asking for a Waymo explicitly is what Waymo has just taken back from Uber. It still depends on the service area and on available cars, so riders should get a number they can check.
Tell Uber riders what they are entering. Until the Uber deployment ends, the Uber-side pop-up could say plainly that this is a chance-based match. It could point riders to the Waymo app to request a Waymo directly, within the service area and subject to availability. That turns the crapshoot into a funnel.
Take the Las Vegas test seriously. Uber’s own rule, a dedicated option once availability and reliability improve, is a fair standard. If Waymo ever meets it on Uber’s network, the button is worth reconsidering on Waymo’s terms, with the ETA and support owned by Waymo.