As the deployment of L3 autonomous driving surges at a breakneck pace, is the intelligent driving industry ushering in a new high-stakes assessment phase?

On October 20, the First Equipment Industry Department of the Ministry of Industry and Information Technology launched a public consultation on five mandatory national standards including the "Automotive Data Security Requirements".

The most widely noticed update is the addition of provisions related to ADS marker lamps, which explicitly define their dedicated light color as blue-green, finally providing a standard basis for autonomous driving vehicles to "express their status to the outside world".

According to the requirements, vehicles equipped with Level 3 and higher autonomous driving systems shall be optionally fitted with front, rear and side ADS marker lamps. When Level 3 and higher functions are activated, the marker lamps will automatically turn on and cannot be manually switched on or off; they will automatically turn off after the system exits.

At the same time, if a vehicle is driving on public roads without ADS enabled, no undefined lamps shall be lit.

In other words, the widely familiar "smart driving small blue light" will in the future be exclusively bound to Level 3 and higher autonomous driving systems.

However, compared with the small blue light, one issue deserves more attention: Level 3 autonomous driving models were already launched last year, and carmakers have long been promoting Level 2.999 products, so why has large-scale commercialization not been realized after a whole year? How far is it from actual end-consumer deployment?

And once commercialization accelerates, what kind of impact will it bring to the intelligent driving industry?

Level 3: When will real commercial deployment be achieved?

As is known to all, the real difficulty of Level 3 autonomous driving is not only meeting technical standards, but also the distribution of accident liability: the driver is responsible under normal circumstances, and the liability is transferred to the carmaker only after entering specific roads or meeting activation conditions.

For this reason, the main line of the industry over the past year has not been "immediate volume growth", but focused on supplementing regulations and systems.

In July, GA/T 2388-2026 came into effect, providing a nationally unified traffic rule for Level 3 and higher vehicles;

In August, China's first mandatory national standard for Level 3/L4 autonomous driving systems was approved, scheduled for full implementation in July 2027;

Recently, the "smart driving small blue light" regulation has also entered the consultation phase.

As the saying goes, provisions and supplies go before troops are mobilized.

In other words, Level 3 autonomous driving is not stagnant, but continuously advancing at the regulatory and standard level; the only remaining gap before large-scale commercialization for ordinary consumers is the final stage of "full implementation of mandatory standards, clear liability rules, and carmakers completing compliance deployment".

In addition, carmakers are also pushing forward related work simultaneously.

Since the beginning of this year, some models have obtained local Level 3 road test licenses, allowing road tests or pilots on designated road sections, including Zhijie RX, Xiangjie G9, Wenjie M9, Avatr 11, XPeng G7, etc.

From this perspective, all parties are making active efforts now — everything is ready, only the final push is awaited.

So what is the exact timeline for Level 3 to be available to end consumers?

Combined with the timeline of the mandatory national standard to be implemented in July 2027, the earliest window for Level 3 to be opened on a large scale to ordinary consumers will most likely be the second half of 2027.

This is also close to some industry judgments: Jin Yuzhi, CEO of Huawei HiAuto, stated in July that Level 3 large-scale commercial use is expected to be realized in 2027; Yu Chengdong previously mentioned that "Level 3 will be available as early as one year later", which also basically fits this timeline.

However, at the 2026 World Intelligent Connected Vehicles Conference that opened on October 21, the industry's judgments on the deployment timeline of Level 3/L4 showed more obvious divergence.

He Xiaopeng put forward three "anti-consensus" views: Level 3 may be skipped, real Level 4 companies should start exploration from Robotaxi, and Level 5 may also be redefined in the future;

He believes that the deployment of autonomous driving technology will be around the second half of 2028, and policy deployment may fall between 2028 and 2032, varying by country, region and scenario.

Li Bin emphasized that technology and compliance should be viewed separately: from Level 2 to Level 5, the technologies share the same origin, but when Level 3 will be actually deployed depends on when the country allows it.

Yang Dongsheng, Senior Vice President of BYD Group, is relatively cautious, believing that Level 3 conditional autonomous driving is expected to be deployed in 2028.

Cao Xudong, CEO of Momenta, believes that 2027 to 2028 may be an important inflection point for Level 3/L4;

Overall, the more optimistic roadmap regards the second half of 2027 as the starting point of large-scale commercial use, while more prudent judgments push the window to around 2028; He Xiaopeng's view of "skipping Level 3" also reminds the industry that Level 3 is not necessarily a mandatory path for all enterprises, and the Level 4 route may advance in parallel.

What is the impact of Level 3 on the industry?

If Level 3 achieves large-scale popularization next year, its impact on the intelligent driving industry will be far more profound than simply "adding a new hands-free driving function".

To understand this, one must grasp a rule: When technology develops to a certain stage, what is truly rewritten is often the underlying logic of business competition.

Once Level 3 is deployed, the first thing to be redefined is "what exactly carmakers are competing for".

This logic is similar to the electrification wave in the past.

After the three core electric vehicle technologies (battery, motor, electric control) were born and matured, the foundation of automotive competition shifted from engines, gearboxes and the petroleum industry chain to batteries, motors, electronic control systems and energy replenishment networks. In essence, industry competition shifted from "oil" to "electricity".

With the rise of intelligent driving, automobiles have further evolved from purely mechanical products to the stage of "software-defined, algorithm-driven decision-making", that is, the competition has shifted from "hardware" to "software".

At the Level 3 stage, the changes are not as earth-shattering as electrification and intelligent transformation, but more in-depth and detailed:

Intelligent driving competition will shift from "focusing on functional experience" to "focusing on safety", from competing for more radical and novel functions to competing for clearer functional boundaries, more sufficient safety redundancy, more rigorous verification systems, and better capabilities to explain, obtain evidence and take responsibility after accidents.

This underlying logic shift may bring three major impacts to the industry:

First: The industry enters a stage of eliminating the false and retaining the true, where only hard power speaks for itself.

Currently, the industry is still in the Level 2 stage, leading to a thought-provoking phenomenon: no matter it is in-house R&D by carmakers or third-party suppliers, almost everyone claims to be in the "first echelon of intelligent driving", claiming to outperform Tesla, and no one admits to being in the second echelon.

The fundamental reason is not necessarily that all players have the same homogeneous capabilities, but that in the Level 2 era, the subject of liability is always the driver — no matter what accident occurs, the car owner bears the final responsibility.

Consumers who are dissatisfied with the intelligent driving experience often have to swallow the losses without complaint.

There is huge room for marketing promotion, while the pressure of liability is limited, so enterprises can easily package "usable" and "safe to use" as "leading".

However, once Level 3 is widely popularized on a large scale, this ambiguous space will be compressed.

When liability starts to transfer from human drivers to systems and carmakers, who really dares to let end users use high-level intelligent driving freely in more scenarios, who has a lower accident rate, and who has clearer takeover boundaries and failure degradation logic, will all be compared openly under the test of real road conditions, real accidents and real supervision — the quality of products will speak for themselves in actual operation.

The second half of autonomous driving is essentially to prove that "cars drive more safely than humans", and Level 3 is exactly the stage where this statement is turned into a hard requirement for the first time.

The pressure of this "mandatory proof" will also be transmitted down the industrial chain: in the Level 3 stage, insurance companies need to assess risks truthfully, and the measurement and implementation of safety indicators will likely follow;

For end consumers, the way to judge the quality of intelligent driving will also transition from listening to "marketing rhetoric" to a "visualized" stage where indicators are visible and comparable.

In other words, Level 3 is more like a magic mirror — it will become clearer who is the "CATL equivalent" in the intelligent driving industry, and who only looks good on paper but performs poorly in practice.

Second: Capable carmakers will further increase investment in in-house developed intelligent driving to hold tight to their "core competitiveness", and the market share of third-party suppliers may shrink.

Since the emergence of intelligent driving, the anxiety of the "soul theory" has always hung over carmakers, and the arrival of Level 3 will undoubtedly aggravate this anxiety.

It can be put this way: Level 2 represents "usable" intelligent driving, Level 3 represents "safe" intelligent driving, and Level 4 represents "user-friendly" intelligent driving. Level 3 is exactly the connecting link between the preceding and the following — some carmakers that carry out in-house R&D will be completely left behind in the Level 4 stage if they fail to catch up in the Level 3 stage.

At the same time, the popularization and market education of intelligent driving in the Level 2 stage have been completed: data from the Ministry of Industry and Information Technology shows that since 2026, the penetration rate of Level 2 new vehicles has reached 70%, and more than 30% of models are equipped with NOA.

With the user base expanded, Level 3 has become a key stage for consumers to form brand perception of intelligent driving, thus Level 3 may become the last window for carmakers to increase investment in in-house R&D.

We can already see this kind of increased investment.

Zhu Jiangming, founder of Leapmotor, said not long ago that intelligent driving has been his top priority since last September;

At the BYD launch event on May 28, 2026, Wang Chuanfu raised the self-developed 4nm automotive-grade intelligent driving chip "Xuanji A3" on site, with a single-chip computing power exceeding 700 TOPS, supporting Level 3/L4, and announced its mass production.

Tan Benhong, Deputy Party Secretary and Director of Changan Automobile Group Co., Ltd., expressed his determination for in-house R&D in July: "Behind intelligent driving is AI, large models, and in-depth understanding of computing power. Adhering to in-house R&D is to explore AI technology and realize Changan Automobile's transformation into a technology company. If we give up in-house R&D, carmakers are essentially just assembly factories that only retain the 'car shell manufacturing' business."

The opposite side of carmakers' in-house R&D is a less friendly environment for third-party suppliers.

Nowadays, as carmakers keep increasing investment in in-house R&D, the market size for third-party suppliers is indeed not as large as imagined.

Data from Gasgoo Automotive Research Institute shows that in the high-level intelligent driving (Level 2++) market from January to May 2026, the combined share of the self-developed solutions of six carmakers including BYD, Li Auto, NIO, Changan, XPeng and Leapmotor has reached 53%, while the total share of third-party players including Huawei, Momenta, Yuanrong Tech, and Zongyu is only 31%.

Third: The overall market cake may shrink, but Level 3 does not only bring impact to third-party players — it may also push the market to a pattern of "the strong get stronger, the weak get weaker".

The arrival of Level 3 adds redundant systems to vehicles, increasing the hardware cost per vehicle by about 30,000 to 50,000 yuan. Carmakers have higher investment and smaller room for trial and error, so they will naturally be extremely cautious when selecting suppliers.

At the same time, according to national standards, under the Level 3 type approval framework, the manufacturer, model, quantity and installation position of the perception system, the model and version of the software, as well as the manufacturer and model of DMS, are all specified one by one in the certification conditions.

This means that replacing any supplier or any software version is equivalent to triggering a brand new type approval, and the Safety Case also needs to be reconstructed.

Once a supplier's product is deployed on vehicles, it becomes much more difficult for OEMs to replace it.

In the past, OEMs were accustomed to "multiple bets", assigning fixed-point orders to multiple suppliers at the same time to make them compete on price, and replacing underperforming suppliers at any time — for OEMs, changing a supplier was no more than a matter of a contract.

In the Level 3 stage, the deeper binding model may bring a turning point to this practice.

Head suppliers with scale, data and mass production experience may gain stronger user stickiness as a result, while small and medium-sized players will find it more difficult to enter the long-term cooperation list of carmakers.

In addition, the third-party intelligent driving market has already been highly concentrated among top players, and the arrival of the Level 3 era will only accelerate this centralization process.

Overall, Level 3 is not just a technical node, but also an industry screening mechanism, or an industry college entrance examination — it pushes intelligent driving competition into a white-hot stage.

Cao Xudong, founder of Momenta, once put forward the "endgame theory" — apart from in-house R&D by OEMs, only two to three intelligent driving solution providers will remain in the domestic market in the end;

He Gang, Vice President of Changan Automobile, also predicted that by around 2028, the industry will complete the first round of drastic reshuffling, and at that time there will probably only be five to six truly competitive solution providers in the Chinese market.

When popularization is completed, safety is forced to be compared publicly, and market share may accelerate to concentrate on top players, the Level 3 stage will very likely be the node where this endgame theory of intelligent driving begins to come true.

The second half of autonomous driving: Beware of repeating the old mistakes of the battery industry?

As mentioned above, Level 3 requires OEMs to carefully select and bind suppliers for a long time on one hand, and on the other hand, it puts most of the liability for intelligent driving accidents back on OEMs.

In the past, car owners were liable for intelligent driving accidents, but now OEMs are liable, and OEMs will then argue with suppliers afterwards.

According to the requirements of national standards, OEMs need to break down safety goals into verifiable technical indicators and pass them on to suppliers through contracts;

Suppliers need to provide evidence chains including simulation, site testing and road operation, which are summarized by OEMs into the Safety Case.

Although intelligent driving suppliers do not directly face regulators, the safety pressure will penetrate through OEMs to the end of the supply chain — whoever's sensor, software or algorithm link causes the Safety Case to fail, the OEM will claim compensation according to the contract.

From this perspective, the relationship between intelligent driving suppliers and OEMs is somewhat following the old path of the battery industry.

Let's first look at two old disputes in the battery industry.

From mid-2021 to the end of 2023, Sunwoda was the cell supplier for the Zeekr 001 WE86 version, and the two parties even jointly funded the establishment of a joint venture at the peak of their close cooperation.

But the honeymoon period did not last: more and more car owners complained about slower charging, capacity degradation and reduced range, and some vehicles were even exposed to fire risks.

In the last three months of 2024, Zeekr quietly provided free battery pack replacement for users in the name of "winter care activity" — ostensibly a customer care measure, but essentially a recall.

This operation helped Zeekr maintain its reputation, but the bill fell on Geely.

Geely naturally asked Sunwoda to share the cost, and Sunwoda's attitude was very clear: replacing the battery was Zeekr's own decision, no similar problems occurred in other batches, and it stated that the issue had nothing to do with it.

According to the Tianyancha App, the result of the dispute between the two parties is that Geely (its subsidiary VREMT) formally sued Sunwoda in December 2025.