Uber has signed more than 30 robotaxi partners and committed over $10 billion — and Waymo is trying to get out of its contract

Every figure in this article is drawn from Uber's second-quarter 2026 earnings release and prepared remarks, its SEC filings, or from CNBC and TechCrunch as attributed. Sections marked as analysis are identified as such.

Uber has spent two years buying its way into a technology it has decided not to invent. It has signed agreements with more than thirty autonomous-vehicle companies, taken equity stakes in at least four of them, and committed over $10 billion to a robotaxi network in which it will own the demand, the app and the customer — and none of the driving.

The bet has one obvious vulnerability, and in the last two months it has started to show. Waymo, the only partner with a genuinely proven driverless system at scale, ended its Phoenix arrangement with Uber in July and is now seeking to exit a contract that runs to May 2028.

The quarter that funds it

First, the strength this is being paid for out of. Gross bookings reached $58.0 billion, up 24 per cent — the fourth consecutive quarter above 20 per cent growth, and above the top of the company's own guidance. Mobility bookings rose 22 per cent to $29.0 billion, Delivery 26 per cent to $27.5 billion, Freight 25 per cent to $1.6 billion.

208 million people used the platform monthly, up 16 per cent, taking 3.9 billion trips, up 18 per cent. Income from operations rose 30 per cent to $1.9 billion, adjusted EBITDA 33 per cent to $2.8 billion, and free cash flow over the trailing twelve months passed $10 billion for the first time. Uber bought back $518 million of its own stock in the quarter, which is not the behaviour of a company that thinks it is short of capital.

One line looks wrong and is not. Mobility revenue grew 1 per cent while Mobility bookings grew 22 per cent. The gap is an accounting change, not a collapse: driver payment costs in several markets have moved from cost of revenue to contra-revenue, taking roughly 400 basis points off the reported Mobility revenue margin and about eight points off total revenue growth, with no effect on the underlying economics. It persists for the rest of 2026, and it is the reason revenue rose 12 per cent while bookings rose 24.

Guidance says the run continues. Uber has told investors to expect third-quarter gross bookings of $58.25 billion to $60.25 billion, growth of 18 to 22 per cent in constant currency, adjusted EBITDA of $2.86 billion to $2.96 billion and non-GAAP earnings of 84 to 88 cents a share. For a company that spent its first decade as a byword for unprofitable growth, the striking thing about the 2026 guidance is how boring it has become.

Uber's platform advantage continues to compound: record consumers and engagement, and profitable growth across our business. — Dara Khosrowshahi, chief executive

The fleet Uber does not own

The list of partners is now long enough to be difficult to hold in the head. Lucid has an order for as many as 35,000 vehicles after Uber added $200 million to its original $300 million investment in April, with Uber holding more than 11 per cent of the company. Rivian has a deal worth up to $1.25 billion, beginning with $300 million and an order for 10,000 R2 robotaxis destined for San Francisco and Miami in 2028 and 25 cities by 2031. Nuro supplies the self-driving system for the Lucid vehicles under commitments totalling around $500 million.

Outside America the map is wider still. Wayve raised $1.2 billion in February with Uber participating, plus a further $300 million contingent on a London deployment. WeRide took another $100 million for a fifteen-city expansion across Europe and the Middle East, targeting 1,200 robotaxis in Abu Dhabi, Dubai and Riyadh by 2027. Baidu, Momenta, Pony.ai, May Mobility and Motional each cover other territories. Aurora, in which Uber holds a 19.7 per cent stake, does the trucks. Nvidia supplies the underlying compute, with a stated ambition of 100,000 level-four robotaxis across 28 cities by 2028.

Uber currently offers autonomous rides in seven cities and expects to be live in fifteen by the end of the year. Not all of it is going smoothly: Avride, which received $375 million from Uber and Nebius last autumn and launched in Dallas in December, came under a federal safety investigation in May after a series of crashes.

Delivery, meanwhile, has quietly stopped being the loss-making adjunct it was hired to be. Its bookings of $27.5 billion are now within $1.5 billion of Mobility's, and it grew faster — 26 per cent against 22. Whatever happens to who drives the cars, roughly half of Uber's marketplace is people ordering things rather than rides, and none of that half is exposed to the robotaxi question at all.

Analysis: what Waymo's exit actually costs

The partnership model has an elegant logic. Autonomous driving is enormously capital-intensive and no one yet knows which system wins; Uber, which already owns the largest pool of ride demand in the western world, can decline to bet on any of them and instead rent whichever works. Every partner needs riders on day one. Uber has 208 million of them.

That logic holds only while the best systems still need Uber more than Uber needs them. Waymo is the test case, and it is failing the test. It began with Uber in Phoenix in May 2023, expanded to Austin and Atlanta through 2025, and then in July pulled Phoenix back into its own app — the arrangement Uber had used to prove the model works. It is now trying to leave the contract three years early.

Read plainly: the partner closest to solving the problem has decided it can find its own customers. That is the outcome Uber's strategy is least equipped to answer, because Uber cannot respond by shipping a better driver. It can only respond by signing more partners — which is exactly what the last two years of deal-making have been.

There is a defensible counter-argument, and Uber makes it. Waymo operates in a handful of American metros; Uber operates in thousands of cities on five continents, and no autonomous developer will build a consumer marketplace in all of them. Demand aggregation is a real and durable asset. Delivery bookings growing 26 per cent on a base of $27 billion says the platform has value entirely separate from who is behind the wheel.

Analysis: the strongest possible position, held on someone else's terms

The honest summary of Uber in 2026 is that its financial position has never been better and its strategic position has never been more contingent. It converts roughly 4.9 per cent of everything booked on the platform into adjusted EBITDA, it generates more than $10 billion of cash a year, and it has guided to another quarter of 18 to 22 per cent bookings growth. Nothing in the numbers is fragile.

What is fragile is the assumption underneath the $10 billion of autonomous commitments: that a marketplace can rent its way through a technology transition without ever owning the technology. Uber sold its own self-driving unit, Advanced Technologies Group, in 2020 and has spent the six years since arguing that was the right call. On the evidence of the cash flow statement, it was. On the evidence of the Waymo contract, it is still an open question — and Uber does not get to close it.

There is a neat historical symmetry in how Uber got here. It sold Advanced Technologies Group to Aurora at the end of 2020 in exchange for equity, and that equity is the reason Uber today holds 19.7 per cent of the company hauling freight autonomously between Dallas and Houston. The self-driving unit Uber gave away became one of the partners it now depends on — which is either vindication of the rent-don't-build thesis or a demonstration of what it costs, depending on which line of the balance sheet you read.

For now the company is doing the only thing available to it, which is to make itself impossible to route around: thirty-odd partners, four equity stakes, fifteen cities by December, and a demand base large enough that leaving it costs something. That is a real moat. It is just not the same thing as owning the car.