Xpeng's New Robotaxi Has No Lidar. Geely's Has 43 Sensors and No Steering Wheel. China Just Split in Two.

On October 8, Xpeng gave its robotaxi a name: Yoyo. The announcement, posted to Weibo, opened public registration in China and marked the moment the program moved, in Xpeng's own words, from technical validation into "marketization and industrialization." It is a vision-only robotaxi with no lidar and no high-definition maps, built on the bones of an ordinary three-row SUV Xpeng already sells to the public. Five months earlier, at the Beijing Auto Show, China's other automotive giant, Geely, unveiled a robotaxi that makes the opposite bet on almost every axis that matters: a purpose-built pod with no steering wheel, no pedals, and 43 sensors including a high-resolution spinning lidar unit on the roof.

Both vehicles are racing toward the same goal — unsupervised, fully driverless passenger service in China within the next two years. Neither company is Baidu, Pony.ai, or WeRide, the names that have dominated coverage of China's robotaxi race so far. Xpeng and Geely are mainstream car manufacturers that already sell hundreds of thousands of consumer vehicles a year, and their decision to build robotaxis in-house, using two fundamentally incompatible engineering philosophies, says as much about the unresolved state of the technology as it does about either company.

What Yoyo Actually Is

Xpeng Yoyo is built on the Xpeng GX, a three-row flagship SUV that sells in China from roughly 399,800 yuan. Rather than designing a dedicated robotaxi body, Xpeng equipped the GX with four in-house Turing AI chips rated at a combined 3,000 TOPS of compute — one more chip than the consumer version of the same vehicle — and layered on its second-generation Vision-Language-Action model, called VLA 2.0, to handle driving decisions. There is no lidar and no reliance on pre-built high-definition maps, a deliberate choice Xpeng says lets it deploy into new cities faster than sensor- and mapping-heavy rivals, echoing the same vision-only argument Tesla has made for years in the US.

The development timeline moved quickly once it started: robotaxi plans were announced in November 2025, a Guangzhou road-test permit followed in January 2026, a dedicated robotaxi business unit was created that March, and the first mass-produced unit rolled off Xpeng's Guangzhou line in May. Internal testing began in July — CEO He Xiaopeng was the program's first rider — and by Xpeng's August earnings call, the fleet had logged more than 2,000 internal test rides. In August, Guangzhou granted Xpeng a remote-testing qualification that allows road tests on designated routes with no safety operator physically in the driver's seat at all, backed by a cloud-based remote takeover system the company built in parallel.

The cabin experience is clearly aimed at consumer comfort rather than utilitarian ride-hailing: riders can set temperature and choose a cabin fragrance before boarding, and planned configurations include five, six, and seven-seat layouts, with zero-gravity seats and a large entertainment screen in the public-facing version. Xpeng says it will act as a technology and hardware provider rather than a fleet operator, opening a software development kit to outside partners, with Alibaba's Amap mapping platform signed on as the first ecosystem partner — and leaving actual day-to-day ride-hailing operations to partners it hasn't yet named for international markets.

Geely's Opposite Bet: the Eva Cab

Geely's Eva Cab, revealed at Auto China in April 2026, rejects nearly every one of those design choices. It is a purpose-built vehicle with a face-to-face cabin, sliding doors, and — per Geely's own description — "no provision for retrofit human controls" at all: no steering wheel, no pedals, nothing a person could grab if something went wrong. Instead of cameras alone, it carries a 43-sensor perception suite built around a 2,160-line digital lidar unit, powered by dual Nvidia Drive Thor-U chips delivering a combined 1,400 TOPS — less raw compute than Xpeng's Turing setup, but paired with the kind of sensor redundancy that vision-only systems explicitly forgo.

Geely isn't building the Eva Cab to sell as a hardware platform to outside partners the way Xpeng is. It already owns the distribution channel: CaoCao Mobility, Geely's in-house ride-hailing arm, operates in more than 60 Chinese cities today, and the Eva Cab has already been tested inside CaoCao's existing operations in Hangzhou and Suzhou. Geely has set a first commercial fleet deployment for 2027 through CaoCao, and the company is reportedly in advanced talks with partners in the UAE and Saudi Arabia for deployment projects outside China — an international ambition similar in shape to what Pony.ai and WeRide are already executing in Europe, but built around Geely's own vehicle rather than a software layer sold to someone else's fleet.

The Same Fight Tesla and Waymo Have Been Having, Now Playing Out Inside China

Readers who've followed the US robotaxi race will recognize this split immediately. It is, structurally, the same disagreement that has defined Tesla's vision-only approach against Waymo's lidar-and-HD-map stack for years: one camp betting that cameras and a large enough AI model can eventually match or exceed sensor-based perception at a fraction of the hardware cost, the other betting that redundant, purpose-built sensing is worth the added expense and slower per-city rollout because it fails more gracefully when something goes wrong. What's notable here is that this exact argument is now happening entirely within China's domestic auto industry, between two of its largest manufacturers, independent of whatever Baidu's Apollo Go, Pony.ai, or WeRide are doing with their own robotaxi fleets.

Xpeng's framing leans hard into the Tesla parallel without naming the comparison directly: faster deployment, lower per-vehicle sensor cost, and a vehicle that started life as a car a regular customer can already buy. Geely's framing leans the other way, closer to Waymo's or Zoox's logic: accept a more expensive, purpose-built vehicle in exchange for a system explicitly engineered so there is no human fallback to design around in the first place.

Two Different Businesses, Not Just Two Different Cars

The hardware split comes with an equally sharp business-model split. Xpeng has positioned itself as what it calls a technology provider and ecosystem enabler: revenue comes from hardware sales, software and data services, and a share of ride revenue, while actual fleet operations are handed to outside mobility partners, both domestic and international, that Xpeng has not yet fully named. Geely, by contrast, is vertically integrated from the vehicle up through the ride-hailing app a passenger actually opens — CaoCao Mobility is Geely's own platform, already running in dozens of cities, which gives the Eva Cab a built-in distribution network Xpeng has to go build through partnerships instead.

Neither approach has an obvious advantage yet. Xpeng's path is faster to scale across many partners and cities if the vision-only technology proves reliable enough without the lidar safety net; Geely's path is slower to replicate outside CaoCao's existing footprint, but gives it direct control over pricing, operations, and the rider experience from day one, closer to how Waymo runs its own service rather than licensing software to a dozen other fleet operators.

Why This Matters Beyond China

Both companies have said, in different ways, that China is the proving ground rather than the destination. Geely is reportedly negotiating Eva Cab deployments in the UAE and Saudi Arabia. Xpeng chose an English name specifically, by its own account, for "easy pronunciation worldwide," describing Yoyo as a first step toward going global, with its SDK model designed to let overseas mobility partners plug in the same hardware and software stack Xpeng is validating domestically. If either approach proves out at scale in Guangzhou or across CaoCao's network, it becomes a template Chinese automakers can export the same way Pony.ai and WeRide have already exported their own robotaxi software to Zagreb and Madrid.

That export dynamic is exactly what should worry Western automakers watching from the sidelines. Mercedes-Benz and BMW are still rolling out Level 3 systems gradually, market by market, inside existing luxury sedans. Xpeng and Geely are each already running a dedicated robotaxi business unit, a production line, and in Geely's case a captive ride-hailing network, less than a year after their respective programs went from announcement to road testing. The question these two vehicles pose isn't just vision-only versus lidar — it's whether mainstream automakers outside China can move at anything close to this pace once they decide to compete.

What to Watch Next

Whether Xpeng's vision-only Yoyo can secure a genuinely driver-free commercial launch in 2027 on the timeline the company has set, which would be the clearest real-world test yet of whether a camera-only system can match lidar-backed rivals at meaningful scale outside Tesla's own fleet.

Whether Geely's UAE and Saudi Arabia talks convert into signed deployment agreements before Pony.ai, WeRide, or Baidu lock up those same Gulf markets first — the same geography already seeing early Chinese robotaxi activity through Baidu's Dubai and Abu Dhabi partnerships.

Whether other Chinese automakers — BYD, Chery, or SAIC among them — announce their own in-house robotaxi programs now that two of the country's largest manufacturers have shown both a fast, asset-light path and a slower, vertically integrated one actually work well enough to reach public roads.

China's robotaxi race has mostly been told, in English-language coverage, as a story about dedicated AV companies — Baidu, Pony.ai, WeRide — expanding outward. Xpeng and Geely's entry changes that story: the country's mainstream car industry is now building robotaxis in-house, and it can't even agree internally on whether the car needs a lidar.

Related: Tesla Robotaxi vs. Waymo: An Honest Head-to-Head Comparison · Europe's First Fully Driverless Robotaxi Didn't Launch in London — It Launched in Zagreb · Is Tesla FSD Worth It in 2026? An Honest Review

Xpeng Yoyo vs. Geely Eva Cab