Meng Xing on Waymo, Cruise, and the Challenges of Driverless Operations

In September 2024, Zhang Xiaojun and Meng Xing discussed Waymo and its autonomous-driving competitors. A former COO of Didi Autonomous Driving, Meng shares industry experience and firsthand observations.

He traces Google’s self-driving project from its 2009 start to the later development of Waymo, emphasizing the years needed to accumulate technical and operating experience.

Meng says he once underestimated the length of self-driving research. Early remotely controlled vehicles, robotics research, and later autonomous driving are distinct technologies but provide historical context.

He recalls DARPA’s autonomous-vehicle challenges, which gave universities and unconventional teams a common arena for exploration. He particularly remembers a small team completing the course with limited resources.

He believes the competitions helped researchers gain experience and drew technology companies’ attention. The event details are his recollections.

Turning to leadership changes, Meng says autonomous-driving companies repeatedly balance research, commercialization, and operations over long periods; personnel changes do not necessarily map directly to technical changes.

In his understanding, the industry gradually introduced data-driven methods into perception, prediction, and planning. Changes in individual modules should not simply be equated with an entirely end-to-end system.

He finds end-to-end approaches appealing to engineers but says driverless service must account for rare risks and controllability. His description of Waymo’s internal research and deployment is his understanding, not a company disclosure.

Meng recalls his first Waymo ride in the Phoenix area in 2022: he requested a car as a regular user and waited for a driverless vehicle within its service area.

He returned for repeated rides, looking at shopping-center parking lots, residential areas, and temporary obstacles. Personal experience can provide clues but cannot substitute for systematic safety evaluation.

He describes how a vehicle might handle unexpected conditions through self-checks, remote assistance, and, when necessary, personnel on site. Procedures can differ across companies.

On one trip, a closed entrance prevented the car from completing the route as expected. Meng says remote staff contacted the passenger and offered alternatives, illustrating that the service involves more than onboard software.

He distinguishes staff communicating with passengers from those assisting the vehicle and says information handoffs between teams also require improvement.

He also observed vehicles returning to an operations center from public areas to understand charging, maintenance, and fleet turnover. External observation cannot reveal the full internal operation.

Meng says he wants to understand boundary conditions where systems struggle, because limitations matter as much as smooth demonstrations. He emphasizes safety in testing.

Comparing Waymo with Tesla FSD at the time, he says their goals, hardware, operating areas, and driver responsibilities differed. Personal rides should not be treated as directly comparable comprehensive safety statistics.

He makes a rough estimate of industry spending from public information and experience. The program does not independently verify Waymo’s cumulative cost, so the figures are speculative.

The conversation turns to Cruise. Meng recalls its acquisition by General Motors and how an established automaker’s resources supported vehicle and autonomous-driving development.

He believes Cruise’s early deal shaped other founders’ expectations of financing and exits. After acquisition, the team retained some startup-like ways of working.

GM’s manufacturing capability helped Cruise explore purpose-built vehicles while its team expanded rapidly. Meng sees speed, funding, and parent-company support as influences on its strategy.

On a Cruise safety incident, he says systems must not only avoid collisions but also correctly recognize an incident and stop afterward; communication and regulatory review matter too.

He describes the vehicle’s movement after the incident and questions raised about the company’s disclosure. Responsibility and investigation findings should be assessed from official records; the interview is his retrospective account.

Comparing Cruise and Waymo, he speculates that parent-company backgrounds, financing, and growth pressure may have shaped decisions differently. This is his interpretation of organizational motives.

He says recovery after a major incident involves regulation, public trust, and team stability as well as technology. He was cautious about Cruise’s recovery prospects at the time.

Meng believes Waymo’s long-term research origins may have allowed a longer experimental horizon, while saying an incubation model alone does not determine an outcome.

He distinguishes incubation inside a large company from independent startups later backed by large companies. Both can develop technology and both need sustained funding and operational capability.

On Uber’s autonomous-driving unit, he cites litigation, management changes, and a testing fatality as factors that affected its direction. This is his analysis of a series of events.

He introduces Aurora and Zoox: the former was putting more emphasis on trucks, while the latter developed a purpose-built vehicle with Amazon’s backing. These descriptions reflect the 2024 interview.

On the business model, he says technology, fleets, operations, and customer acquisition all affect unit economics. Higher driver costs in the US may create opportunity, but profitability still requires real-world validation.

He does not expect one instant “ChatGPT moment” for L4 services: technology and service tend to expand gradually, and scaling also depends on costs and operations.

He suggests a large fleet operating continuously as one hypothetical indicator of commercial maturity, while acknowledging that the threshold is his own.

Waymo can operate its own ride-hailing service or partner with an existing platform. Meng says platform building, user acquisition, and vehicle operations all cost money, so the relationship need not be simple replacement.

In 2024 he considered Tesla’s path to Robotaxi technologically and operationally challenging. His views and timing forecasts reflect information at the time, not subsequent outcomes.

He concludes that driver assistance and driverless service in a defined area have different product goals; whether and when broader autonomy can be achieved remains uncertain.