The global robotaxi fleet will grow from today's pilot deployments to 331,000 vehicles by 2031, an average annual growth rate of 78%, according to market researcher Yole Group.

That headline figure underpins a parallel forecast: the combined sensor and processor markets serving robo-X vehicles will reach roughly $4.4 billion for sensors and almost $4 billion for processors by 2035. Together, they redraw the commercial logic of autonomous ride-hailing — and, by extension, the silicon pipeline feeding it.

What changes the business case?

Yole points to a 70% reduction in vehicle bill-of-materials and manufacturing cost as the turning point. Layered on top are additional commercial services, end-to-end algorithms, hybrid modes that mix robotaxi and driver-based ride-hailing, and convergence between L2+ driver-assist and L4 self-driving stacks. The combination, Yole argues, has converted pilot fleets into scalable services rather than demonstrations.

"The robotaxi industry has shifted from proving that the technology works to proving that the business works," said Yole analyst Yu Yang. "Removing up to 70% of the vehicle cost is what turns a demonstration fleet into a service that can scale, and a handful of operators have now crossed that line."

Unit-economics milestones remain rare. Pony.AI has reached them in two Chinese cities by exploiting synergies between robotaxis and robottrucks.

Who leads, and who is catching up?

The competitive ranking is fluid, but Yole's snapshot places Waymo first on operational unsupervised fleet size, with Pony.AI, WeRide, and Apollo Go close behind. Apollo Go runs the largest raw fleet with near-full coverage in some Chinese cities. Waymo is scaling its commercial fleet in the US while running tests in the UK and Japan.

• Tesla is launching its Cybercab.
• XPeng is entering with a foundation model shared across its AV programs and a dedicated vehicle.
• Geely supplies the base car to Waymo and runs the Caocao mobility platform.
• Zoox has released a purpose-built robovan for commercial operation in the US.
• Uber has exited in-house robotaxi development but partners with multiple operators across several countries.
• DiDi has stopped in-house vehicle development and turned to vehicles co-designed with GAC.
• WeRide is expanding globally and merging its L4 robotaxi and L2+ ADAS stacks.

Automotive OEMs are no longer bystanders. Tesla, XPeng, and Geely have all joined the field, and most other carmakers now engage robotaxi players through co-development, contract manufacturing, or direct investment. Although most operators start at home, competition has gone global: the US and China are incubating giants, while Europe, the Middle East, and parts of Southeast Asia are the favored overseas destinations for both camps.

What does this mean for chip suppliers?

For semiconductors, the most consequential finding is that the robotaxi does not stand up a new supply chain. It extends the existing ADAS one.

"The robotaxi does not create a separate semiconductor supply chain," said Yole's Pierre Boulay. "Instead, it stretches the existing ADAS one. The same processor and sensor suppliers serve both markets, and the real differentiator is vertical integration."

The supplier landscape therefore largely overlaps with established ADAS names across processing and the multiple sensing modalities — camera, radar, LiDAR. Three exceptions stand out, and they will shape which silicon gets designed in:

• Waymo: proprietary TPU, in-house LiDAR and radar.
• Tesla: proprietary SoC.
• XPeng: proprietary SoC and power modules.

With global fleet volumes on track for two million vehicles by 2035 — roughly six times the 2031 figure — pricing, sensor mix, and the depth of in-house silicon will increasingly decide which chip vendors ride the cycle and which get pushed into commodity ADAS sockets.