Baidu, the tech company sometimes called the “Google of China” announced its new 6th generation robotaxi, with plans to deploy 1,000 of them in Wuhan this year. They also forecast they will be profitable in 2025 (on a “unit economy” basis, a type of gross margin.) The new vehicle, which costs them 200,000 RMB to build (around $27,500 USD) is half the cost of their previous generation vehicle.
While the new vehicles retain a wheel and pedals, they will operate without safety drivers in Wuhan, where 70% of Baidu Apollo Go’s robotaxi operations currently are without a safety driver, according to the company.
They are also lowering costs by automating a lot of their depot operations. These vehicles use automatic battery swap to recharge, and Apollo Go states that vehicle activation, dispatch, cleaning and retrieval will all be done without human intervention. On the software side, they announced a new foundation model (AI technology) will now be doing the driving, and is flexible enough to deploy in new cities with about 6 months of effort.
While Chinese car economics are different from those in the west, and Chinese EVs cost less than in the USA, the ability to put together a nice EV along with all the compute and sensor suite for under $28,000 is a big step, and will eventually be duplicated around the world. Once you get the other factors of a robotaxi service (such as those depot operations listed) to work at scale, the largest cost of operating a robotaxi derives from depreciation of the vehicle, so the lower the vehicle cost, the lower the cost per ride. (This differs from regular taxis where the human driver is the largest cost.) In fact, once you start removing features for the driver (like wheels, pedals, adjustable seats and a complex dashboard) a robotaxi can cost less to make than a consumer car.
Indeed this vehicle—though made to Chinese standards and pricing—undercuts the likely forecast price for the “Tesla Model 2” which Tesla recently put on the backburner in favor of a robotaxi. Chinese standards are improving, but there is one snag in bringing those lower costs to the USA—a planned 100% tariff on EV import from China. But even at double the price, this vehicle is cheaper than any planned robocar in the USA. Today, US consumers don’t know and can’t buy Chinese brands, and they may take some time to do so, but those rules don’t apply to robotaxis. Indeed, Waymo has contracted with Zeekr, a unit of Chinese auto giant Geely, to build their next generation robotaxi. Taxi customers don’t really care about what nameplate is on their taxi. They will care about the Waymo/Google brand or Uber brand, but not about what’s underneath. It could easily be a Chinese brand they’ve never heard of.
In fact while an imported robotaxi must not have any significant safety issues, the same standard does not apply on mechanical reliability and quality. When a consumer car breaks down, the owner is greatly inconvenienced and angry. If a robotaxi breaks down or has a quality problem, it’s just taken out of service. If passengers were in it, it pulls over and a new taxi arrives in 60 seconds to continue them on their way—almost no inconvenience at all. As such, Chinese quality standards in low cost vehicles may well already be high enough for the western markets.
Battery swap is not reasonable for consumer cars, but can be a good choice for robotaxi fleets. It will be interesting to see just how they will automate cleaning—I have reached out to Baidu for comment on that.
Robocar and automaker players outside China must take notice. Right now Waymo is dominating the space outside China, but a tiger is waiting to strike. Due to current politics, it is unlikely the USA and other countries would let a Chinese based company command and operate a robotaxi fleet in the USA, but once the Chinese companies seek to expand, they will find non-Chinese partners to deploy their technology, and a tiger may be ready to pounce.
