Welcome to the Halftime Show.

The first time that I took an autonomous car was actually in Phoenix, AZ last year via Waymo. We blasted drum 'n bass at 1 am, windows down, and it felt transformative, euphoric. Now in Shenzhen, my daily routine involves triaging between multiple AV providers for the best deals.

Watching a driverless car navigate rush hour chaos (iykyk) in China never stops being surreal. As a regular customer of three different robotaxi companies, this issue is thrilling.

Let's lock in.

The technology question is answered. At least four Chinese companies and two American ones can drive a car without a human behind the wheel at commercial safety standards.

If you live outside San Francisco or Shenzhen - the two epicenters of robotaxis - this might be news.

Which car ends up on your street comes down to which business model actually travels: economically, geographically, and at scale. 2026 is the year that question gets tested in real time across four continents.

Apollo Go is China's front runner - backed by Baidu, the company that built China's Google, then rebranded as an AI company.

Waymo is America's front runner - backed by Alphabet, the company that built the actual Google.

→ Global Autonomous Vehicles market size expected to grow at a CAGR of 34.84% from 2026 - 2035, driven by advancements in AI and sensors. (Precedence Research)

→ Waymo raises $16B to fuel global robotaxi expansion- valuing the Alphabet subsidiary at $126B (The Guardian)

→ Tesla, Amazon's Zoox, and Uber to challenge Waymo's AV dominance in San Francisco - all launching soon (San Francisco Times)

→ Uber announces partnership with Apollo Go to launch robotaxis in Dubai. Uber controls demand and the customer interface, Baidu supplies the driverless stack. (Uber IR)

→ Apollo Go: 20 million cumulative rides. 200% YoY growth. Weekly rides peaked at 300,000. 26 cities as of February 2026. One airbag deployment per 12 million kilometers. (Benzinga)

→ Baidu's RT6 costs ~$28,000. Waymo's runs ~$75,000- nearly 3x. Waymo operates across five US cities at ~$2.00/mile. Apollo Go operates primarily in Wuhan at ~$0.35/mile. (The Driverless Digest)

→ Uber has signed four Chinese autonomous driving companies: Apollo Go, Pony.ai, WeRide, and Momenta- for commercial deployments across the Middle East and Europe. Every major Uber robotaxi partnership outside the US runs on Chinese AV technology. Pony.ai shares surged 48% on its Uber announcement. (TechCrunch)

The robotaxi race in 2026 has split into two theaters running two completely different experiments.

Waymo and Zoox are playing defense on home turf with premium models.

Waymo is dominating the AV industry, rightfully so, off the large war chest of Alphabet. It holds 90% of US autonomous ride market share, has the deepest safety dataset in the industry, and is expanding to 20+ cities in 2026 including London and Tokyo.

Zoox is building the vertically integrated closed loop model. Own the vehicle, the app, the customer relationship. Amazon backed, premium priced, US focused.

Both are betting depth beats portability. That the US market alone (for now)- premium pricing, high density cities, regulatory familiarity - is large enough to build a significant business on.

US Roster

China's top three robotaxi operators have deployed across 50+ cities by early 2026. The country moved from pilot programs to treating autonomous vehicles as scalable urban infrastructure.

China Roster

Apollo Go, Pony.ai, WeRide, and Momenta built their models in Chinese cities where the margin structure is brutal by design.

Apollo Go proved its unit economics in Wuhan first. Wuhan is a "second tier" Chinese city where fares run 30% below Beijing and a fraction of Dubai, London, or Seoul.

The thesis: if the model works at the floor of the revenue curve, it works almost everywhere else.

The logic: Uber already has the riders. Building a consumer app costs capital and attention that should be pointed at one thing - accumulating miles and proving unit economics.

This is the opposite of the token playbook. Vertical integration and proprietary distribution have been the defaults for every major Chinese tech company that's gone global. These AV firms did the opposite.

Why?

Because the variable that drives profitability in robotaxis is not control of the consumer relationship. It's:

Vehicle cost (fixed) + Operational cost per mile (falls with utilization) = Fare revenue per mile × Rides per day

If that equation turns positive, distribution becomes a commodity. If it doesn't, owning the app just means you're burning capital on two problems instead of one.

Social platforms compound with users. Ecom platforms compound with transactions. Autonomous driving compounds with miles.

Miles can't be scraped, licensed, or synthesized. Every city Apollo Go enters is a data collection operation that makes every other city in the fleet smarter. Dubai trains London. Seoul trains Wuhan.

So the Uber partnership makes more sense than it looks. Miles matter more than brand. Utilization matters more than margin. Speed matters more than control.

Waymo and Zoox: Depth. Superior safety data, dominant US position, premium pricing, methodical expansion. The bet is that the US market alone is large enough to build a significant business on.

Apollo Go, Pony.ai, WeRide, Momenta: Portability. Prove the economics at the floor of the revenue curve first. Partner for distribution. Let the model travel to markets where fares are structurally higher and the cost base is already validated.

In a world where autonomous vehicles could displace ride sharing incumbents faster than current models project, the variable that matters most is which approach scales without requiring the world to change around it.

If you get the chance to take a robotaxi, take it. The gap between reading about autonomous vehicles and sitting in one is the difference between theory and inevitability.