Meng Xing on the Boom and Bust Cycles of Autonomous-Driving Investment

In 2024, Zhang Xiaojun and Meng Xing reviewed investment cycles in autonomous driving. At the time of recording, Meng had just left Didi Autonomous Driving and was preparing to return to venture capital.

Meng says that after joining Shunwei Capital in 2016, he focused on AI, autonomous driving, and other frontier technologies, looking for teams whose technical ability was central early on.

He favors areas without broad consensus but acknowledges that companies eventually need product, distribution, and organizational capabilities, not technology alone.

In 2019 he joined Didi’s spun-out autonomous-driving business to participate more directly in building it, dealing with strategy, partnerships, product, and team management.

He recalls managing a larger team and says scaling operations brought challenges he had not encountered in earlier startups and investing.

After leaving Didi, he compares entrepreneurship and investing to a revolving door and hopes different roles can help technology and products reach users sooner.

He believes operating experience adds a dimension to investment judgment: the problems teams may face as they grow and the consequences of early choices.

Why return to investing when sentiment is weak? Meng says periods without consensus can offer room to rethink approaches. This is his career choice, not market-timing advice.

He hopes to participate earlier, even helping define a direction and form a company. Deep incubation, he says, involves responsibility for outcomes rather than advice alone.

He recalls changing languages and schools between China and the US as a child and thinks repeated adaptation may have shaped his willingness to change roles later.

On accelerators, he says different models serve different teams and stages. Early funding, guidance, and alumni networks can help, though terms vary by organization and period.

He distinguishes his desired incubation approach from short-term acceleration: deeper involvement means organizing talent, setting product direction, and working with the company, not merely investing.

Looking back at Shunwei’s investment in XPeng, he says he supported the early team while crediting other investors and founders. This is his recollection of the process.

He valued the team’s understanding of local users and product positioning and its ability to turn ideas into experiments, rather than simply copying an existing success story.

Discussing Momenta, he recalls meeting its founding team and says it emphasized the relationship between data collection and technical iteration early on.

Meng argues that very high early valuations can pressure teams to continually prove themselves and cites his own limited observations. This is not a universal rule or a prediction of any company’s outcome.

He worries that financing pressure can push teams toward appearances, rapid expansion, and hard-to-verify promises while products and technology still need time to mature.

He acknowledges that founders, investors, and markets jointly influence these choices, so outcomes cannot be attributed simply to one party.

Reviewing the autonomous-driving startup boom around 2016, he cites Cruise’s acquisition, advances in deep learning for perception, and expectations for driverless mobility.

Many technical teams then thought commercialization would follow directly from solving the core technology. In hindsight, he says, vehicles, regulation, and operations made the chain longer than expected.

He recalls rising early-stage valuations and says investors must assess teams, technical direction, product understanding, and execution when evidence is often incomplete.

He describes late 2018 and early 2019 as a funding trough: some companies had higher valuations without the revenue or business metrics later-stage investors typically expected.

Capital later returned and valuations and expectations rose in 2021. Meng recalls valuations for some public companies relying on distant revenue forecasts and therefore on sensitive assumptions.

Discussing conditions in 2024, he says funding and valuations had fallen from their highs. Company status and market prices continue to change; the episode records a particular moment.

He describes a changing talent mix as people from automotive engineering, automation, and computer vision entered at different technical stages.

He acknowledges that early expectations about automaker acquisitions and supplier value did not necessarily materialize. Turning technical ability into customers, revenue, and bargaining power requires real-world proof.

From a demo to everyday service, progress often shows up in reliability and handling more situations rather than appearance. One successful demonstration does not prove a system always works.

He distinguishes automakers, sensor and chip suppliers, software suppliers, driverless-service operators, and simulation and data providers. Their customers and revenue models differ.

After years of competition, he sees concentration and pivots in some segments, while his list of company positions reflects the industry as observed in 2024.

Comparing the autonomous-driving cycle with large models, he sees similarities in funding, talent competition, and changing expectations, while noting more large companies participate directly in models.

He suggests founders consider mismatches between funding and business cycles, work toward a sustainable business loop before financing cools, and manage staffing and cash. This is his experience-based view.

He also discusses personal resilience: industry excitement and outside assessments fluctuate, and founders must manage expectations and accept that some questions have no immediate answer.

His proposed way to handle cycles is to control spending and stay focused during good times, leaving room for conditions to worsen.

He thinks large-model products can reach users faster, while driverless vehicles must clear vehicle, safety, and operational hurdles. Both have uncertain ultimate commercial outcomes.

Looking ten years ahead, he imagines driverless services spreading to more cities and logistics, as well as other mobile devices. These are 2024 scenarios, not guarantees.

On public listings, he says financing windows, shareholder needs, and commercialization progress can all affect decisions; a listing is a milestone, not an endpoint.

He closes by asking which frontier-technology companies can survive booms and downturns and whether their technical, funding, and product paths offer lessons.