Tsla's moat in self driving
The self-driving space has been a battleground of ideas: LiDAR vs. cameras, rule-based to end-to-end. A lot of dead bodies with new entrants keep coming. As FSD pivots to end-to-end, there are a lot of discussions around its moat - why can’t other OEMs get there? what about $NVDA or $MBLY?
Data stands at the heart of this discussion. While the importance of data volume and quality is widely recognized, nuances such as the type of data and its accessibility are often overlooked or undervalued.
The right type of data is crucial. $Tsla has 9 (high resolution) cameras per car, while most existing fleets on the road have 1-2 of different quality. Although $Nvda sells ~1M units of Orin chips annually and $Mbly's EyeQ chips have 70% penetration - most of the installed vehicles don't produce the right kind of data for training.
Access to data is equally critical. The presence of cameras doesn't guarantee usable data, with much of it not being saved or simply disconnected. Even when data is stored, third parties like $Nvda and $Mbly face hurdles using it in real-time.
And of course, volume (and quality) of the data matters. $Waymo, rumored to be also pivoting to end-to-end, struggles with <500 vehicles and under 10 million miles of data. And this is not to underestimate the internal execution and organizational challenges of "removing 300,000 lines of code (and likely a complete reorg)".
Potential competitors face three hurdles: producing the right type of data, accessing that data, and achieving sufficient volume and quality. This screening process effectively sidelines most existing fleets and creates obstacles for third-party providers. New EVs might meet the first two criteria but their fleet sizes fall short.
$BYD stands out as a potential well-positioned player, albeit without a clear focus on autonomous driving so far. $Nvda vs. $Tsla in some way mirrors that of $Qcom vs. $Aapl, where $Nvda leads in chip design but lacks $Tsla's level of integration and ecosystem control. Players like $LI $XPEV and $NIO all use $Nvda's chips, yet they are developing their self-driving algorithms in-house.
As the industry gravitates towards end-to-end approach, Tesla's moat actually seems to be widening.
Modify on 2024-04-04 13:01
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
