Hello! I'm your financial analysis assistant. This in-depth report on the three major state-owned enterprises—FAW, Dongfeng, and Changan—competing fiercely in the development of their own autonomous driving technologies contains a wealth of information and touches on the most critical issue in the automotive industry today: the struggle for control of technology and the balance between costs and survival.
To help you easily understand the implications behind this, I will first summarize the key points and then break down the report in simple terms from five different perspectives.
📝 Summary of Key Points
In one sentence:
Faced with rising prices from third-party autonomous driving suppliers (such as Huawei and Xpeng) and soaring chip costs, FAW, Dongfeng, and Changan have decided to stop relying on external solutions and opt for in-house development. However, this is not just a technological battle; it's also an economic and time-sensitive competition.
Key contradictions:
- Must we develop it ourselves, or can we afford to? Regulations require automakers to take responsibility for autonomous driving, but in-house development costs between 5 to 10 billion yuan per year and requires a massive amount of data.
- High sales volume vs. limited data: Together, these three companies sell 3.65 million vehicles, which seems substantial, but due to different models and platforms, they cannot share data, potentially reducing the amount of useful data for AI training.
- Time pressure: The industry predicts a first round of consolidation in the autonomous driving market by 2028, leaving less than two years for these companies to transform their sales volume into algorithmic capabilities.
---
🔍 In-depth Analysis: Five Simple Explinations
1. The Trigger: Why the sudden shift to in-house development?
Previously, automakers found it easier to purchase ready-made autonomous driving systems from companies like Huawei and Xpeng. However, the situation has changed:
- Rising prices from suppliers: Huawei has increased the price of its advanced autonomous driving packages, and other companies like BYD and Xpeng have also raised their prices. This means automakers are paying more per vehicle sold, squeezing their profits.
- Soaring hardware costs: The cost of automotive-grade storage chips has increased by 180% in recent times, significantly raising manufacturing costs.
- Concerns about control: If core autonomous driving technology is in the hands of others, automakers risk becoming mere “shell companies” without control. A supplier's decision to stop supplying or change strategy could leave them at a disadvantage.
- Conclusion: The price increase is just the catalyst; the real reason is that these state-owned enterprises realize they need to control their own technology to remain competitive in the era of automation.
2. The economic cost: How much does in-house development really cost?
Many assume in-house development simply means hiring developers, but it's a much more complex undertaking:
- High labor costs: For example, Changan's total investment of 60 billion yuan includes various aspects, but the labor cost for its autonomous driving team alone (about 1,500 to 2,500 people) amounts to 1.5 to 3.8 billion yuan per year.
- Hidden costs: In addition to labor, there are expenses for computing power, data annotation, testing facilities, and maintaining cloud platforms. Maintaining the autonomous driving team alone costs 5 to 10 billion yuan annually.
- Break-even point: If the cost of in-house development is 5 billion yuan per year and a third-party solution costs 8,000 yuan per vehicle, automakers would need to sell 625,000 vehicles to break even. If the cost is 10 billion yuan per year and a third-party solution costs 5,000 yuan per vehicle, they would need to sell 830,000 vehicles or more.
- Cruel reality: This doesn't even account for the risk of technological updates. Autonomous driving technology needs to be updated every 2 to 3 years, requiring new chips and sensors, making the investment both substantial and unlimited.
3. The data challenge: More sales doesn't equal more data—effective data is crucial
This is a common misunderstanding in the report. Although these companies sold 3.65 million vehicles in the first half of the year, the data from these vehicles may not be valuable for AI training:
- Data silos: FAW, Dongfeng, and Changan do not share a common system, so the data from each company cannot be used collectively.
- Low conversion rate: Just owning a vehicle doesn't mean the autonomous driving feature is used, and even if it is used, the data must be transmitted back to the cloud and cleaned and annotated properly.
- Case comparison: Huawei has 1 billion kilometers of data, and Dongfeng has 1.25 million datasets, but the quality and relevance of the data vary. The key is the proportion of data from high-value, complex scenarios.
- Changan's dilemma: Changan's Qiyuan autonomous driving system is targeted at the 80,000 to 200,000 yuan price range, where users are less likely to opt for it (with a penetration rate of only 11.6%). The more vehicles sold, the less effective data is generated, making it difficult for in-house development to be profitable.
4. Organization and time: The slow pace of state-owned enterprises vs. the fast pace of the industry
The automotive industry is evolving rapidly:
- Tight timeline: Experts predict a first round of consolidation in the autonomous driving market by 2028, leaving less than two years for these companies to turn sales into competitive algorithms.
- Systematic challenges: State-owned enterprises have longer decision-making processes, which may not keep up with the speed of tech companies.
- Talent competition: Top AI talent prefers to work for companies like Huawei, Tesla, or new entrants. Whether state-owned enterprises can offer competitive salaries and retain such talent is a significant challenge.
- Different approaches to in-house development:
- Changan: Goes all in with in-house development, which is costly but provides complete control and data.
- FAW: Takes a cooperative approach to reduce costs through joint development, but intellectual property rights are a concern.
- Dongfeng: Adopts a cautious approach with a dual strategy (in-house + external procurement) to minimize risk, but this may result in less focus on either area.
5. The ultimate question: Is in-house development about saving money or about survival?
This is about more than just economics; it's about strategy:
- Beyond economics: For state-owned enterprises, in-house development also involves supply chain security, technological capability, compliance with export regulations, and responding to national industrial policies.
- The real test: These companies are not short of money or sales volume; what they lack is the ability to convert sales into algorithmic advantages.
- Who can convert real road test data into effective AI models at the lowest cost?
- Who can establish a strong technological barrier by 2028?
- Conclusion: The competition is about who can quickly turn data into useful algorithms. If the experience doesn't match customer expectations, large-scale production won't be profitable; if it does, the high costs of in-house development can be offset by substantial market profits.
💡 Insights for the general public
- Consider the manufacturer behind autonomous driving: When buying a car with autonomous driving, look at who is developing the technology—whether it's in-house or purchased externally. In-house models may offer better updates and better data privacy.
- Autonomous driving in lower-priced cars may be limited: Vehicles in the 100,000 to 200,000 yuan range are less likely to have affordable autonomous driving features due to low adoption rates.
- An industry consolidation is coming: 2028 is a critical year; many current second-tier brands and suppliers may disappear. Choosing leading brands (either automakers or suppliers) is safer for consumers.
I hope this breakdown helps you understand the underlying logic of automotive automation through this complex financial news report!