🔓 Unlock the full article with a premium subscription.
"You might own a car like maybe some people own a horse. You know, they might take a ride on the weekends or something."
"I view it as a solved problem. We will be there in a few years."
The first time I hailed a driverless taxi on Reem Island was earlier this week.
I needed oat milk (lactose intolerance, a recent and unwelcome development, long story) and rather than ordering from the dark store visible from my window, I opened Uber, tapped "Autonomous," and a Chinese-built robotaxi with enough LiDAR on the roof to scan for extraterrestrial life pulled up outside my building.
A safety specialist was sitting in the front seat (Abu Dhabi's fully driverless operations launched on Yas Island, Reem still has a human present, for now), the ride was smooth, courteous in the way that software is always courteous, and conspicuously uneventful, which is apparently the highest compliment you can pay a self-driving car.
I got my oat milk. The whole thing cost the same as usual.
That safety specialist, though, won't be there much longer. And when he's gone, so is the job that brought him to this country.
Which leads to a chain of consequences that connects a Guangzhou tech company, a collapsed San Francisco startup, a $30 billion Alphabet experiment, and 70,000 immigrant workers.
So this essay is all because I needed milk. Or oat drink, to be politically correct about it.
What it actually costs to run a robotaxi
Before we can make sense of what's happening in the UAE, it’s important we take a step back to grapple with the global economics, which are, to put it charitably, still very much a work in progress.
Waymo invariably is the benchmark to look to. Alphabet's autonomous driving subsidiary now completes over 450,000 paid rides per week across six US cities, a 45x increase from May 2023. Its fleet of roughly 2,500 robotaxis generated an estimated $350-400 million in revenue last year.
The company is reportedly raising $16 billion (Mubadala waving the UAE flag) at a valuation north of $110 billion if reports last week are to be believed. On paper, this looks like vindication but in practice, each Waymo vehicle (a Jaguar I-PACE carrying five LiDARs, six radars, and 29 cameras) costs around $175,000, and Alphabet's "Other Bets" segment lost $1.43 billion in a single quarter of 2025. Total investment in Waymo now exceeds $30 billion. So yes, whilst this is very much a company that has proved the tech works, it has not it’s fair to say, yet proved the business does.
Riders currently pay an average of $19.69 per trip in San Francisco, a 12.7% premium over Uber's $17.47, though that gap has narrowed from 30-40% just six months earlier. McKinsey estimates the driver accounts for roughly 60% of traditional taxi costs. Eliminate that, and the maths gets interesting, but only if you bring vehicle and operational costs down fast enough. Analyst consensus puts cost parity with human-driven ride-hail at around 2027-2028.
The indomitable Mr Musk's Tesla offers the counterpoint. Its robotaxi service launched in Austin last June with 135 modified Model Ys, a human safety monitor, and a $4.20 flat fare. The approach is radically different: cameras only, no LiDAR, no radar, no HD maps. It's the same instinct that led SpaceX to build rockets from stainless steel instead of aerospace-grade composites, a decision the industry mocked until it worked. Tesla's Cybercab (a purpose-built two-seater with no steering wheel, priced under $30,000) is scheduled for mass production in April 2026.
If Musk's vision materialises, Tesla would produce millions of $30K robotaxis versus Waymo's thousands of $175K ones. That is the core of the economic debate.
A crucial caveat however is that Musk's autonomous driving timelines have a near-perfect record of being wrong: he promised a million robotaxis by 2020, complete autonomy "within two years" in 2015, and cross-country autonomous drives by 2017. None happened.
Ben Thompson, of Stratechery newsletter fame, has perhaps the most useful framing. In his piece 'Elon Dreams and Bitter Lessons,' he invokes computer scientist Rich Sutton's 'Bitter Lesson,' a foundational idea in AI research that boils down to this: brute force wins."
Systems that hoover up enormous amounts of data and learn from it will always, eventually, outperform systems that are painstakingly hand-built by clever engineers. Applied to autonomous vehicles, Waymo is the engineer, meticulously mapping every intersection and coding every edge case. Tesla is the brute, throwing billions of miles of camera footage at a neural network and betting that scale solves what engineering cannot.
Thompson draws a parallel to SpaceX, which a former executive at the European launch monopoly Arianespace dismissed as "selling a dream" in 2013 before being utterly disrupted. But even Thompson hedges.
The graveyard of AV companies (Cruise, $10 billion burned; Motional, $1.5 billion lost) suggests the real question isn't which sensor suite wins but whether the total economics actually work at scale in any configuration.
How Cruise's collapse handed the UAE to Beijing
The most consequential event in UAE autonomous vehicle history happened 11,000 kilometres away. In October 2023, a Cruise robotaxi in San Francisco struck a pedestrian and dragged her 20 feet. Within months, permits revoked, fleet grounded, and by December 2024, GM killed the entire programme after sinking $10 billion.
This mattered enormously for Dubai because Cruise was the RTA's sole autonomous vehicle partner, signed to an exclusive deal in 2021 for 4,000 self-driving taxis by 2030. Sheikh Hamdan had personally announced it. So when Cruise collapsed, Dubai's entire AV taxi strategy collapsed with it.
Into that vacuum rushed three Chinese companies with startling speed…
Stay ahead in MENA’s tech landscape
This is premium content. Subscribe to read the full story for less than $0.27 per day.
