To back up city NOA, BYD fights to turn the tide in intelligent driving.

```

Author | Zhou Zhiyu

Thanks to its early advantages in electrification, BYD has been shining in the automotive industry over the past four years. As the industry enters the second half of the intelligent transformation, how to maintain its own technological dividend is the core issue for it to sustain its industry position.

On the evening of May 28, BYD made another move. At the intelligent strategy launch event that night, BYD Chairman Wang Chuanfu announced that all models can be optionally equipped with the "Divine Eye B", with an option price of 12,000 RMB. For new users with Divine Eye B or above and old users of Divine Eye 5.0, BYD will provide unlimited full compensation for traffic accidents caused by assisted driving during urban navigation—not through insurance, not affecting next year's premiums, with both new and old car owners enjoying one-year coverage.

The urban navigation coverage was one of the most widely spread news that night, but BYD's more important move is that it doubled down in the intelligence sector, launching the self-developed 4nm intelligent driving chip Xuanji A3, the Xuanji Architecture 2.0 integrating cockpit, intelligence, and drive, a physical AI large model, and an autonomous driving platform for L3/L4.

A company known for batteries and manufacturing is trying to prove it can also enter the technical field of intelligent driving. Against the backdrop of Tesla’s FSD accelerating localization and China’s L3 regulations expected to land in 2027, this move’s timing is intriguing.

BYD’s chosen path is also noteworthy. It is still competing on the battlefield it knows best—using cost control, manufacturing scale, and vertical integration to redefine the rules of intelligent driving competition. After last year’s underwhelming intelligent driving strategy, this time BYD is taking a completely different approach.

Coverage

Wang Chuanfu revealed that the team initially recommended a three-month coverage, but he insisted on one year.

Three months is a safer choice—the risk of compensation is controllable, and it’s enough to make a marketing event. But Wang Chuanfu clearly wanted more than a marketing event. The decision is rooted in last year’s lesson.

In February 2025, BYD, along with Changan, Geely, Chery, and others, became "partners in intelligent driving," ambitiously aiming to launch the Year of Universal Intelligent Driving.

BYD executed most thoroughly, with 21 models all equipped with the Divine Eye. However, after a certain automaker had an intelligent driving accident last year, the "universal intelligent driving" campaign faced setbacks. In April 2025, relevant departments banned the use of "intelligent driving," and “universal intelligent driving” encountered policy headwinds. By March this year, BYD’s assisted driving models accounted for 43.2% of total sales.

This failure taught BYD one thing: the obstacle to the adoption of intelligent driving isn’t technology or price, but who bears the consequences. Wang Chuanfu’s data illustrates this: less than 2% of vehicles have urban navigation, and among those, the usage rate is only around 30%. For highway navigation, usage rate among the same group exceeds 90%. The same technology, but people dare to use it on highways, not in cities—the difference is the urban scene is more complex, and responsibility is more ambiguous after accidents. Providing coverage eliminates this ambiguity.

For a long time, BYD's rapid sales growth relied on competing for users with traditional fuel cars. However, by 2025, traditional automakers like Volkswagen, together with suppliers like Bosch and Zhuoyu, gradually addressed their weaknesses in intelligence. Meanwhile, startups like Leapmotor and XPeng were also catching up, launching models with urban navigation in the 100,000–150,000 RMB price range.

The technical dividend wall BYD tried to build early last year quickly collapsed. In early 2025, BYD Seagull Intelligent Driving Edition’s highway navigation was unique in the industry, but by early 2026, according to incomplete statistics from Wall Street Insights, more than 70 models had highway navigation capability, and some even had urban navigation within the 100,000 RMB price range.

For BYD, many buyers are “oil-to-electric” transition consumers. Feedback from BYD sales staff shows they mainly try to convert these customers. A salesperson told Wall Street Insights that recently BYD raised the Divine Eye B option price, but about half of consumers are still willing to select it; they care more about fuel savings and price, and intelligent driving hasn't yet become a "must-have".

How to get these users willing to try and pay for intelligent driving is BYD's key agenda for winning this group. BYD's choice is coverage.

Parking has already validated this logic. After launching coverage for intelligent parking last year, usage surged from 21% to 93%. When automakers are willing to bear the consequences, user behavior changes dramatically.

Wang Chuanfu insisted on one year rather than three months—first, because urban scenarios need to experience all seasons, extreme weather, and holidays for trust to truly build up. Second, because of the data. BYD’s intelligent driving vehicles exceed 3.15 million units; Divine Eye generates 200 million kilometers of data daily, but most comes from highway scenes. A year of coverage can generate massive real urban data—the most scarce resource for training intelligent driving models.

On another front, drastic price cuts in the LiDAR industry over the past year have made this strategy feasible in terms of cost. Wall Street Insights learned from suppliers that top manufacturers' mass production costs have fallen to just above 1,000 RMB, with detection distance, precision, and reliability all greatly improved. Manufacturers like Suteng and others have supported BYD’s new intelligent driving popularization plans, both technologically and cost-wise.

Industry chain sources said BYD raised the Divine Eye B option price recently, partly due to higher outsourced LiDAR chip costs and production ramp-up, but with mass production of self-developed chips in the second half of the year, costs are expected to fall further.

The 12,000 RMB option price for Divine Eye B includes LiDAR, Orin X chip, and urban navigation software; Wang Chuanfu said this is already the cost price. According to a Goldman Sachs research report on May 26, investors are most concerned about how low BYD will push urban NOA’s price bracket.

BYD’s answer: as low as possible.

Ticket to Entry

Coverage addresses whether consumers dare to use it. But tonight BYD is also answering another question: Does it have the qualifications to sit at the same technical table as Huawei and XPeng?

BYD started making chips before it started making cars. It has already designed 566 automotive-grade chips, covering electric drive, battery management, body control, and nearly all aspects. But for intelligent driving chips, this is the first. Xuanji A3 is the 567th chip released tonight—a 4nm process, three chips can deliver over 2,100 TOPS per vehicle, supporting L3/L4-level autonomous driving, and has begun mass production.

Looking at parameters alone, Xuanji A3 is not the strongest in its class. NIO’s NX9031 has 546GB/s bandwidth, Li Auto’s M100 single chip has 1280 TOPS, XPeng's Turing chip has 750 TOPS. But Wang Chuanfu emphasized another set of numbers: combined with self-developed algorithms, computational efficiency rises by 100%, and power consumption per unit of computing power is 20% lower than similar chips.

This is a very "BYD" answer. An analyst from a foreign institution familiar with BYD pointed out that the company’s success in the electrification era wasn’t mainly due to having the strongest parameters, but because it could use vertical integration to press costs to the lowest, and scale to spread its advantages widest. Now it’s trying to replicate that logic for intelligent driving chips.

Timing is important. AI computing industry’s demand for high-bandwidth storage chips has tightened supply for automotive-grade chips; HSBC estimates storage chips alone raise single car costs by 1,000-3,000 RMB, and NIO founder William Li said memory price hikes add 3,000-5,000 RMB per high-end EV.

Combined with lithium carbonate jumping from 60,000 to 180,000 RMB, new energy vehicle purchase taxes moving from exemption to halved, and industry sales profit margin in Q1 at only 3.2%. When profit margins are squeezed so tight, controlling intelligent driving hardware costs becomes a matter of survival. BYD’s scale means it can bring chips to cost parity fastest.

The accompanying Xuanji Architecture 2.0 merges cockpit, intelligent driving, and electric drive into a central computing platform, reducing signal scheduling latency to 8 microseconds. In algorithms, BYD introduced a “physical AI large model,” with a cloud world model collaborating with vehicle-end physical AI.

Wang Chuanfu gave an example: a traditional end-to-end model sees a basketball rolling across the road and passes normally, while physical AI can predict “there could be a child behind,” and proactively stops and waits. Cloud training overnight can be equivalent to a million kilometers of real vehicle testing.

Chip, architecture, large model—BYD has completed a full set of intelligent driving entry tickets in one night. Specific performance will depend on real car deployment. But the message is clear: BYD doesn’t intend to be just a price disruptor in intelligent driving, it wants to join the technical game. Considering last year’s lessons, how far it can go this time remains to be seen.

Tough Battle

Wang Chuanfu said at the event that China’s L3/L4-level autonomous driving regulations are expected in 2027, and BYD is fully prepared.

The coverage announced on May 28 is, in a sense, BYD preparing ahead for this role transition. The Divine Eye autonomous driving edition showcased that night carries the Xuanji A3 chip, equipped with ultra-high line LiDAR, detection distance of 600 meters, 1,000fps snap camera, and tenfold system redundancy. This hardware isn’t for selling cars tomorrow, but for preparing for 2027.

This is a tough battle. L3 means the responsibility for intelligent driving transfers from users to carmakers. Automakers including GAC, Lantour, and others have also been announcing that their models have L3 hardware pre-installed since last year’s new launches. The goal is to reassure buyers worried that rapid tech development might make their purchase obsolete if they delay.

The Divine Eye C OTA plan also tries to encourage purchase willingness. BYD revealed that Divine Eye C will upgrade to a map-free end-to-end model, planned to push to all users in December. Divine Eye C covers models from 70,000 to 200,000 RMB, and does not use LiDAR. If its end-to-end model can deliver urban memory navigation—even just for commuting—cars under 100,000 RMB can run urban navigation without extra cost.

Currently, models with standard urban NOA start above 150,000 RMB; Leapmotor A10 LiDAR edition optional urban NOA price is 86,800 RMB. If Divine Eye C delivers, the bottom line for urban NOA pricing will drop lower.

Especially with Tesla’s FSD accelerating localization. For the domestic intelligent driving industry, FSD brings not only a new competitor, but a completely different approach: pure vision, no LiDAR, global data training. This will force all domestic car makers to reassess their cost structure and technical routes. BYD’s coverage and self-developed chips are, to some extent, a preemptive move before FSD is fully deployed locally.

However, actual deployment is key. BYD launched its second-generation Blade battery and flash charging tech in March 2026, but Wall Street Insights has learned that due to process adjustments, second-gen Blade battery capacity is still ramping up, and high-end model capacity is expected to ease only in Q3 2026. Other automakers also have similar battery plans, but whether BYD can deliver before reaching full capacity is crucial for maintaining its advantage.

Intelligent driving protection plans from domestic car makers are evolving rapidly. XPeng, Huawei, Avatr all offer intelligent driving protection, but most require insurance claims first or have compensation caps. BYD’s coverage challenges the industry: whether to follow. With industry sales profit margin at only 3.2% in Q1, this decision is especially hard.

BYD sells millions annually, its per-car coverage cost diluted by scale; smaller makers following suit would face much bigger hits to their profit and loss. This mirrors BYD’s logic in launching price wars with scale—the move is bearable for big players, tough for smaller ones.

Moreover, scale advantage will only snowball. If coverage can raise urban navigation usage from about 30% to parking’s experience, the urban scenario data from BYD’s 3.15 million intelligent driving vehicles will far surpass competitors. Data trains models, models improve experience, experience drives usage, usage generates more data. Once this flywheel spins up, smaller makers can't catch up.

Intelligent driving was once the core of startup car makers’ valuation stories. As urban NOA’s threshold is lowered, these stories will need a rewrite.

Whether BYD can truly turn its hardware and manufacturing capabilities into competitive strength in intelligent driving—last year’s setback already proved it’s not easy. It must keep bridging its shortcomings as its tech gap with peers shrinks. Time is running out for BYD.

Risk Disclosure and DisclaimerThe market entails risks and investments require caution. This article does not constitute personal investment advice and does not take into account the individual investment goals, financial situation, or needs of particular users. Readers should consider whether any opinions, views, or conclusions in this article suit their particular circumstances. Invest at your own risk. ```