Summary of Key Points
Tesla's Robotaxi business has been significantly outperformed by Waymo, mainly due to its reliance on a pure vision-based approach (using only cameras) that has not been recognized by regulatory authorities. The introduction of FSD (Full Self-Driving) in China is likely to be limited to L2-level assisted driving, which does not meet the standards for advanced autonomous driving. Tesla's valuation is supported by Wall Street's faith and expectations of a merger with SpaceX. In the future, Tesla may have to abandon its pure vision approach, but doing so would come at a huge cost—requiring the retraining of all previously collected data. It will be difficult for Tesla to catch up with Chinese automakers.
1. Why Has Tesla's Robotaxi Been Outperformed by Waymo?
Waymo, a Google-owned autonomous taxi company, currently has 4,000 vehicles in commercial operation across 11 cities, holding more than 25% of the market share in San Francisco, and can accept orders 24/7. In contrast, as of June 2026, Tesla only had 40 test vehicles, of which only 20 were fully autonomous (without safety drivers), all located in Austin, Texas. The difference in scale is enormous!
The key reason for this disparity lies in their technical approaches: Waymo uses a multi-sensor fusion system that combines lidar, cameras, and millimeter-wave radar (similar to how humans use eyes, ears, and touch to perceive the world), along with high-precision maps, which significantly enhances safety. Tesla, on the other hand, insists on a pure vision approach using only eight cameras, without radar or high-precision maps, which is cost-effective but less safe.
Regulatory bodies are not convinced by Tesla's approach: The California DMV requires thousands of hours of testing to prove safety, and pure vision-based systems struggle in conditions like night driving with high beams or rainy weather. The CPUC has received numerous complaints about Tesla test vehicles causing traffic disruptions and sudden, unexplained braking events. Waymo has already obtained the necessary permits, while Tesla has not even qualified for autonomous testing, limited to small-scale trials in more lenient regions like Texas.
2. The Pure Vision Approach Is a Dead End: Global Regulations Disallow It
Countries around the world are introducing regulations that prohibit the use of pure vision for advanced autonomous driving:
- United Nations regulations require multi-sensor redundancy (more than one sensing method) for L3 and above levels of autonomy.
- New Jersey, USA, requires millimeter-wave radar for L3 and lidar for L4.
- China's new regulations in June 2026 stipulate that L3 vehicles must be able to detect obstacles 130 meters ahead and 9 meters to the sides in rainy, snowy, or foggy conditions at speeds of up to 120 kilometers per hour. Pure vision-based systems cannot meet these standards.
Tesla chose the cheaper pure vision approach to save costs, but current regulations make advanced autonomous driving impossible.
3. Has Tesla's Valuation Not Crumbled? It's Supported by Faith and SpaceX Expectations
At least one-third of Tesla's market value (about $500 billion) is tied to its Robotaxi ambitions. Despite the poor performance of this business, why hasn't the stock price fallen?
- Wall Street remains optimistic, continuing to fuel speculation until concrete evidence of failure appears.
- There are expectations of a merger with SpaceX, which some analysts believe could lead to a significant increase in Tesla's valuation. Many investment institutions consider this merger a key reason to buy Tesla stocks.
4. Is the Pure Vision Approach Doomed? Musk May Have to Change Course, but at a Heavy Cost
Tesla initially chose pure vision because lidar was expensive (around $75,000 for 64-line systems in 2015), and multi-sensor fusion technology was not yet mature. However, the situation has changed:
- Lidar prices have dropped significantly; Chinese manufacturers like Speedstar and Hesai are producing them for several thousand dollars.
- Huawei and比亚迪 have developed advanced multi-sensor fusion algorithms that have resolved issues with sudden braking and decision-making delays.
- Tesla's own AI5 chip (to be mass-produced in 2027) supports multi-sensor fusion. Abandoning pure vision would mean retraining all data collected using lidar-equipped vehicles, a process that could take two to three years. Chinese automakers like比亚迪 train their vehicles for 200 million kilometers daily; by then, they might have reached L4 levels, while Tesla might never catch up.
5. Is the Introduction of FSD in China a Major Step Forward? It's Probably Just L2-Level Assistance
The article suggests that Tesla's FSD is entering a "sprint phase" in China, but according to Chinese regulations, it still does not meet the requirements for L3 or higher levels of autonomy. Therefore, the FSD being introduced will likely be limited to L2-level assisted driving, with humans still in control. Those hoping for fully autonomous driving in China will be disappointed.
In Conclusion: Tesla's pure vision approach was initially a cost-saving strategy, but it has faced significant barriers from regulators. Its valuation is inflated by market expectations and the potential merger with SpaceX. The introduction of FSD in China is more of a L2-level assisted driving system, not true full autonomy. Investors should be cautious, as there is still a large bubble in Tesla's valuation.