Autonomous Driving: The Ultimate Game in a Trillion-Dollar Arena. Who Will Reign Supreme in the Next Decade?

If the past decade was the golden age for the mobile internet to reshape our lives, then in the next decade, autonomous driving is bound to become the core force rewriting the logic of human mobility.

From tech giants to traditional automakers, and even the astute players in the capital market, all understand that this is not just a technological revolution but also a fierce competition for a multi - trillion - dollar market.

Tesla's FSD, Waymo's Robotaxi, Baidu Apollo's urban NOA... These frequently mentioned industry terms represent the accelerating technological iteration and the expansion of business boundaries.

We are experiencing a paradigm shift in the transportation sector similar to the replacement of feature phones by smartphones. Currently, it is the critical turning point from "assisted driving" to "fully unmanned driving". Whoever can be the first to break through the bottleneck of large - scale commercialization of L4 - level autonomous driving is expected to hold the right to speak in the global technology and mobility industries in the next decade.

01 The Dispute over Technical Routes: The Contest and Evolution of Two Paths

When test - driving mainstream new energy vehicles today, L2+ - level assisted driving is no longer a novelty: functions like automatic car - following, lane - keeping, and high - speed navigation are common, and some models can even change lanes autonomously on urban roads.

Since 2023, the rapid implementation of urban NOA has marked the transition of autonomous driving from "simple highway scenarios" to "complex urban environments", but this is just the prologue of the industry.

According to the standards of the Society of Automotive Engineers (SAE), autonomous driving is divided into six levels from L0 to L5. Currently, most mass - produced vehicles are in the transition stage from L2 to L3. True "driverless" (L4/L5) driving is still limited to specific scenarios, such as Waymo's fully unmanned taxis in San Francisco and autonomous logistics vehicles in closed parks.

Even though leading companies have achieved partial breakthroughs, the large - scale implementation of L4 - level autonomous driving still faces three major challenges: technological reliability, regulatory adaptability, and cost control. The industry has formed two major camps in terms of technical route selection:

1. The Pure Vision School: Tesla's "Data - Driven" Path

The pure vision route centered around Tesla uses cameras to simulate human vision and combines AI algorithms for perception and decision - making. The advantages are obvious: cameras are low - cost and easy to mass - produce. With over tens of millions of vehicles in operation, it can continuously collect a vast amount of road condition data to feed back into the neural network.

Now, FSD V12 has achieved "end - to - end decision - making", which does not rely on preset rules and directly outputs driving instructions, significantly improving the ability to handle complex scenarios. However, its drawbacks are also prominent: in scenarios such as heavy rain, fog, and strong light, the perception accuracy of cameras is easily affected.

2. The Multi - Sensor Fusion School: The "Safe Solution" of Waymo and Huawei

The "multi - sensor fusion route" chosen by Waymo, Huawei, etc., emphasizes "safety redundancy": lidar (for precise distance measurement), millimeter - wave radar (resistant to interference), and cameras (for capturing details) work together to reduce the risk of single - sensor failure.

In the past, the core factor restricting this route was cost. In the early days, the unit price of lidar was hundreds of thousands of yuan, but now it has dropped to the thousand - yuan level, paving the way for mass production. For example, Huawei's ADS 2.0 uses "lidar + camera + millimeter - wave radar" to achieve the urban NOA function without high - precision maps.

There is no final conclusion on which route is better, but both are evolving rapidly: Tesla uses data to compensate for hardware shortcomings, while the multi - sensor fusion school overcomes the mass - production problem through cost reduction.

As emphasized by the Gelonghui Research Institute in 2023, there is no need to oppose different technical routes. Instead, more attention should be paid to the underlying capabilities, especially the breakthroughs of large AI models in decision - making and planning.

Today, the end - to - end logic of Tesla's FSD V12 and the popularization of the "BEV + Transformer" architecture in the multi - sensor fusion school have both confirmed the judgment that "the end of perception hardware is algorithms".

02 Beyond "Building Good Cars": The Battle for the Autonomous Driving Ecosystem

If autonomous driving is compared to an "intelligent robot", vehicle technology is just the "body", and the complete ecosystem is the core of its operation. The Gelonghui Research Institute has broken it down into four major levels, each hiding key opportunities:

1. The Perception Layer: The "Eyes" of Autonomous Driving

Cameras, lidars, and millimeter - wave radars form the "perception matrix". The performance of lidars from domestic companies such as Hesai and RoboSense is approaching international standards. Sunny Optical and OFILM hold important shares in the market for high - pixel automotive - grade lenses.

2. The Decision - Making Layer: The "Brain" of Autonomous Driving

Chips provide computing power, algorithms are responsible for planning and decision - making, and high - precision maps provide road information. NVIDIA's DRIVE Orin (with a computing power of 2000 TOPS) has become the preferred solution for L4 - level autonomous driving. Domestic chips such as Horizon Journey 5 and Black Sesame Intelligence A1000 have filled the gap in domestic computing power.

In terms of algorithms, the "BEV + Transformer" architecture enables vehicles to predict the trajectories of traffic participants, making decision - making closer to human intuition.

3. The Execution Layer: The "Hands and Feet" of Autonomous Driving

By - wire chassis and electronic steering systems convert decisions into actions. In the past, traditional chassis could not meet the requirements, but now companies such as Bosch, Continental, and domestic Desay SV have achieved mass production of by - wire chassis, providing reliable support for driverless driving.

4. The Support Layer: The "Infrastructure" of Autonomous Driving

5G and V2X enable vehicles to be connected to the network in real - time. Cloud computing supports data storage and algorithm training. Simulation platforms reduce the cost of road tests. As of 2024, Waymo's simulation test mileage has exceeded billions of miles, dozens of times that of actual road tests.

The Gelonghui Research Institute has repeatedly emphasized that the competitiveness of China's autonomous driving ecosystem lies in the self - controllability of the underlying supply chain. From the cost reduction of domestic lidars to the breakthrough in computing power of autonomous chips, the domestic industrial chain has achieved a transition from "following" to "running side by side" in multiple aspects. This is also the core logic for our optimism about investment opportunities in related fields.

Of course, technological maturity is just the first step. The popularization of autonomous driving also needs to break through the "policy" and "ethical" barriers:

Policy Level: Countries are still exploring the standards for L3/L4 - level vehicles on the road and the determination of accident liability. Although Germany was the first to allow L3 - level vehicles on the road, it will take time for domestic adaptation. In China, Beijing and Shanghai have opened Robotaxi tests, but full - scale commercialization requires cross - regional regulatory coordination.

Ethical and Safety Level: The algorithm selection in the "trolley problem", data privacy protection, and network security defense (to prevent hacker hijacking) need to be jointly solved by technology, social consensus, and the law.

The promotion of all disruptive technologies is not a "linear acceleration". However, when technology, the ecosystem, and policies work together, the implementation speed of this mobility revolution may far exceed expectations.

03 Investment Opportunities: Is it Too Late to Enter Now?

Although the concept of autonomous driving has been hyped up many times, the industry is still on the verge of explosion. With the leap in the capabilities of large AI models, the decision - making ability of autonomous driving has undergone a qualitative change. In 2024, automakers are accelerating the implementation of urban NOA, and the pilot projects of Robotaxi are expanding. The capital market's valuation of related companies also reflects high expectations. Tesla and NVIDIA in the US stock market remain the focus, while in the A - share market, the industrial chains of Baidu and Huawei, as well as concept stocks related to lidars and high - precision maps, are repeatedly active.

From an investment perspective, the opportunities in the autonomous driving industrial chain can be divided into four levels:

1. Vehicle Manufacturers and Solution Providers: Such as Tesla, Waymo, and the autonomous driving departments of traditional automakers. They are at the forefront of the industrial chain, with high potential returns but also the highest risks.

2. Key Technology Suppliers: Including chip manufacturers (NVIDIA, Qualcomm, Horizon) and sensor providers (Hesai, RoboSense, Luminar). They provide core components, and their business models are relatively stable.

3. Infrastructure and Service Providers: This includes high - precision map providers (Baidu Maps, NavInfo), simulation test providers (Tencent, ANSYS), and cloud platform providers (Alibaba Cloud, AWS). Their profit models are clearer.

4. Application and Operation Service Providers: Covering Robotaxi operations, autonomous logistics, and intelligent transportation, they focus on the commercialization of specific scenarios.

It should be noted that this industry has a long cycle, high investment, and high volatility. Policy risks cannot be ignored. Investors should focus on companies with technological barriers and clear commercialization paths rather than chasing hot - topic concept stocks.

In terms of time, opportunities emerge in stages: in the short term, focus on the hardware opportunities brought by the popularization of L2+/L3 - level assisted driving; in the medium term, look for the commercialization breakthroughs of L4 applications in specific scenarios; in the long term, pay attention to the maturity of the fully autonomous driving ecosystem and value distribution.

04 How to Seize the Big Opportunity in Autonomous Driving?

Since the first DARPA Autonomous Vehicle Challenge in 2004, this technological journey has lasted for more than two decades. Now, multiple technological breakthroughs and policy openings are happening simultaneously. 2025 may be a crucial year for the commercialization of autonomous driving.

As an industry veteran said, "Autonomous driving is one of the most difficult but also the most valuable problems in the field of AI." It is not only a technological challenge but also a comprehensive test of social acceptance, laws and regulations, and business models.

If you believe that technology can change the world and that the future belongs to intelligent mobility, then perhaps now is the best time to pay attention and get involved. The story of autonomous driving has just begun, and its ending may far exceed our imagination today.

The Gelonghui Research Institute continuously tracks the development of the autonomous driving industry and has repeatedly grasped the key nodes of technological evolution, policy changes, and investment opportunities in a forward - looking manner. We not only focus on the technological aspect but also provide decision - making references for investors from multiple dimensions such as business models, industrial chain structures, and investment timings.

Note: The companies mentioned in this article are only for case analysis and do not constitute any investment recommendations. The market is risky, and investment should be made with caution. Be sure to conduct independent research before making decisions.

This article is from the WeChat official account “Gelonghui APP” (ID: hkguruclub), written by the editor of Gelonghui and published by 36Kr with permission.