Autonomous Driving Is the Next Growth Era — for a Very Few Players

The consensus view of the car market is a replacement story. 彦山堂's Weibo post argues it is about to become a creation story: the next five years will bring a new growth era driven by autonomous driving — and it will belong to a very few players. This is part two of our three-part series on that forecast; part one covered the saturation argument the post is arguing against.

Parts: 1 — the smartphone argument · 2 — autonomy as the next increment · 3 — the 2027–2030 timeline

The claim

The post's counter-thesis is compressed into one line: the next five years will see a brand-new growth era driven by autonomous driving, and this round of growth will belong to only a few top players.

It names two companies as the examples: Tesla and XPeng — the ones that could, in the author's framing, achieve Robotaxi at scale and drive the popularisation of consumer L4. The change starts in China and the United States and spreads outward. And autonomy, the post argues, creates entirely new categories of demand rather than replacing existing cars: home travel, ride-hail and taxi services, parking-lot operations, cultural tourism, even property amenities tied to residential developments.

Why "a few players" is the load-bearing part

Strip the enthusiasm out and the argument's real content is a cost structure. To do what the post describes, a company needs, at the same time:

• a driving model good enough that removing the human is not a marketing claim;
• compute — XPeng's own stack has moved to its Turing chips and VLA 2.0, the vision-to-action model it has been rolling out across the G9L and other models;
• data and fleet operations, meaning cars on the road that can be observed, updated and paid for;
• regulatory work, which is slow, jurisdiction-by-jurisdiction, and mostly invisible until it is done;
• capital, because all of the above runs for years before it earns.

XPeng is one of the companies actually accumulating these pieces in public. Its Yoyo robotaxi won a city-wide operating licence in Guangzhou this year — a permit covering a whole city, not a pilot zone — which is the kind of unglamorous milestone the forecast depends on.

The corollary is uncomfortable for everyone else. If the increment belongs to a few players, then for most of the industry the smartphone story — replacement demand, price competition, consolidation — is still the true story. The post is not saying the market grows for everybody; it is saying the growth moves to a different place and a different set of companies.

The new demand the post is pointing at

The list of scenarios is the part that deserves more attention than it usually gets, because each one is a different business:

Scenario	What actually has to change
Home travel	A car that drives itself is a car whose time becomes usable; the sell is hours, not horsepower.
Ride-hail and taxi	The unit economics of a driverless fleet versus a human-driven one — the core Robotaxi bet.
Parking operations	Valet-style autonomous parking turns parking from a fixed cost into a service.
Cultural tourism	Self-driving shuttles and tours in places where a driver is a cost and a constraint.
Property amenities	Buildings that offer autonomous transport as part of the residence — the most speculative item on the list.

Not all of these will land on the same schedule, and the post does not claim they will. The point is that demand of this kind does not appear in today's replacement-cycle accounting at all.

The counterpoints

Two things temper the thesis, and both are worth stating plainly.

The "few players" claim cuts against the timeline. The companies best placed to win are also the ones with the most expensive path, which is why the winners list is short — but it also means the growth arrives later than a replacement-market recovery would.

Autonomy claims have a long history of arriving late. The post's own dates put "beyond-human driving in all scenarios" at 2027 for Tesla and 2028 for XPeng. Whether that is a forecast or an aspiration depends on evidence that does not yet exist in public: safety data at scale, disclosed cost per kilometre, and permits in more than a handful of cities.

Part three takes the post's 2027–2030 timeline apart and lists what would have to be true for each phase.

Series: 2 of 3. Source: 彦山堂 (Weibo) — translated & adapted for XPENGTech.