Tesla's big gamble: Full Self-Driving in the wild

Tesla full self driving, Full self driving, Full self driving. Being called a watershed moment for FSD. It is one of the most controversial parts of Elon Musk’s Tesla strategy, especially since FSD as it’s called. It’s not actually fully self driving, but love it. It’s super cool. Super cool. Like this is game changing. Or hate it, full self driving will do something and it just got lucky. Use it. This is cool and I’m not like some FSD enthusiasts or fear it. This is stressful. This is not what you would hope, which is like taking the stress out of driving. It is now in the wild with a free trial for millions of US drivers, and it could reshape the race to full autonomous driving. This, the contact check. Tesla’s huge FSD Gamble, FSD, or Full Self Driving Supervised is what Tesla calls its Driver Assistant product. You can set a destination and the car will attempt to drive most of the way there. Imagine having a car where you just say where you want to go and the car takes you there, navigating city and residential streets, intersections and turns, plus highways. But if the name Full Self Driving has you picturing yourself in the back seat, taking a nap, or even behind the steering wheel texting, think again. Slow your roll. Slow your roll. Oh geez oh geez. A Tesla manual states that FSD is not autonomous and quote, requires a fully attentive driver who is ready to take immediate action at all times. The manual also tells you to keep your hands on the steering wheel with sensors and the wheel and a camera inside the car monitoring how attentive you are to the road. Some places it it gets things right and and it knows what to do but in others many the roads are very complicated and there are special rules in lots of different places and it’s doesn’t know how to follow all those rules. So not full self driving at all. And a branding that regulators have taken issue with as Tesla uses the term to sell cars. The California DMV in 2022 saying that Tesla statements about FSD were untrue or misleading and not based on facts and that is cars with FSD could not at the time of those advertisements and cannot now operate as autonomous vehicles. And the National Highway Traffic Safety Administration. NITSA has also launched investigations into Tesla’s Autopilot system including one that prompted Tesla to recall nearly all the vehicles it had ever sold. 2 million vehicles had been recalled by Tesla in December and and while the term recall is used, it was an over the year software update meant to address what NITSA considered to be inadequacies in terms of how Autopilot works when it comes to interacting with the driver, alerting the driver about potentially dangerous situations. They did the recall in December. The complaints haven’t stopped. In fact, there’s still been a flurry of complaints, and that has prompted Nitsa to now say, OK, you know what, we’ve closed down our investigation that was looking into Autopilot. That led to the recall in December. Now we’re opening up a new one, looking at whether or not the recall that was put in place did the job. Nitza also investigated nearly 1000 crashes in which Tesla Autopilot was thought to have been used, finding that the system may have contributed to at least 467 collisions, with at least 13 resulting in deaths and others in serious injuries. The agency wrote the Tesla’s design of Autopilot quote led to foreseeable misuse and avoidable crashes, and did not adequately ensure that drivers maintained their attention on the driving task even compared to other car makers with similar technology. According to industry standards, Tesla’s technology is only considered partial automation. That’s Level 2, with features like lane centering and adaptive cruise control. But the driver has to constantly supervise, so it wants to make sure my hand is still on the wheel. So it wants a little bit of firm pressure on the wheel every once in a while. And plenty of automakers have cars with this functionality, including General Motors, BMW, and Audi. The next step is a big leap. Level 3 technology means that you can take your hands off the wheel and your eyes off the road in certain conditions. And it’s actually Mercedes that has beaten Tesla to the punch with its drive pilot tech, becoming the first car maker in the US to start selling a hands free self driving car. Once it’s activated, I basically have a lot of freedom. I can kind of do anything I want in the car, provided that I stay in the seat and provided that I can kind of get a read on my face. And then there are the true robo taxis, Google’s Waymo and GM’s Cruise that have reached Level 4, in which cars can operate completely autonomously without human intervention. So given all the companies with more advanced technology and the regulatory hurdles that Tesla is facing, why is Wall Street tech insiders and many Tesla owners so excited about FSD data, data and more data? It is a butt load of data. It is the essential component that any company developing autonomous driving needs to even have a chance at success, and Tesla is approaching it and gathering it in a very different way. Tesla’s FSD technology is primarily based on a vision centric approach that uses a kind of AI program called a neural network that looks at images from cameras on the car and decides how to drive the car, all happening instantly as the car moves. It is a major departure from the system of cameras, radar, and most critically lidars that current leaders like Waymo and Mercedes used to develop its autonomous driving technology. Elon Musk has made his opinions about Lidar clear. Lidar is is a fool’s errand and and anyone relying on Lidar is doomed, arguing they’re too expensive and heavy and ultimately unnecessary. Instead, Tesla is leveraging a vast existing network of vehicles and drivers versus the hundreds of more expensive custom and lighter saddled autonomous vehicles on the road that the likes of Waymo deploys. And when many of those Teslas are parked in driveways at night and connected to a Wi-Fi network, a ton of the data it collects gets uploaded to Tesla to, in turn, further train and improve the neural network algorithms that power its FSD system. I’ve noticed also that it can start to predict things before I can. I didn’t even really think that that person crossing the stairs was going to enter the street, but it’s seen millions and millions of clips. It was able to sort of predict that Tesla recently rolled out a one month free trial of FSD supervised to Tesla owners in North America, feeding more data into that neural net. Well known venture capitalists Brad Gersner and Bill Gurley. They argued on a recent podcast that Tesla’s unique system could deliver not just the quantity of data required, but a better quality of data as well. When you look at these alternative models, right. If this really is about data, and remember, Bill just said an important point, which is it’s not just about quantity of data. Something magic happens around a million cars. Yes, you’ve got to get all that quantity of data, but to get the long tail events right, these are events that occur 10s or just hundreds of times. That’s where you really need millions of cars. Otherwise you don’t have a statistically relevant pool. So that’s the bull case. But the big gamble and the multibillion dollar question is still unanswered. Can Tesla develop safe, reliable autonomous driving technology by using mostly cameras and stripping out a lot of the tech that the rest of the industry sees as critical? Bears also argue that people might not want to pay extra for FSD features like cruise control and blind spot detection that are increasingly folded into the base price of other cars, Bernstein writes. Even auto part technology and lane keeping technology are beginning to be included for free in luxury cars today. So in other words, the space is so competitive. Tesla’s FSD technology might not be as much of A draw once other companies catch up, but Wall Street loves Tesla’s FSD less for its technology, more for its subscription revenue potential. We’ve seen what’s happened with auto margins coming down. They probably will come down some more with a new affordable car, but it doesn’t matter if you can slap on an amazing subscription on it, that gets you a higher margin. So if we could see margins overall in aggregate slowly creep up because of FSD take rate, and we’re hearing already the version 12 free trial was a smashing success, then that translates into greater margins overall subscription revenue. It’s predictable, stable, sustainable and it has lock in effects. In what seems to be an attempt to spur more subscriptions, Tesla recently cut the recurring fee for FSD in half, from $200 a month to just 100 bucks and to 8000 a year from 12,000. It would take someone almost 10 years of paying $100 a month to recoup that cost, a clear sign that Tesla is encouraging buyers to opt for a subscription instead. I still wouldn’t pay 12 grand, two to three at most, and I’ve been saying that for years now, and I still don’t think that it hasn’t proved enough to make me think it’s worth four or five. And with subscriptions, Musk appears to be taking a page out of Apples playbook, finding ways to generate recurring revenue on already sold hardware. Apple Services segment, which includes revenue from iCloud storage, Apple TV plus Apple Music, Apple Pay and many more that hit a record high at the end of 2023. That pace of growth now outpacing its flagship iPhone segment which grew just 6% in that period and is a less reliable revenue stream that is pumping the entire flywheel, its ecosystem of hardware, software and now this consistent and reliable subscription stream. Tesla is hoping that Wall Street sees that potential for its cars as well. But the Street is also accustomed to Elon Musk’s tendency of over promising and under delivering on self driving especially. He’s made bold assertions on the level of autonomy Tesla can achieve, ones that have yet to come true all the way back to 2016. Musk wrote on Twitter that Tesla expects to demonstrate a self driven cross country trip next year that got pushed back to 2019. We expect to be set feature complete in self driving this year and we expect to be confident enough from our standpoint to say that we think people do not need to touch the wheel, look out of the window sometime probably around, I don’t know second quarter of next year then 2020. I’m very confident about full self driving functionality being complete by the end of this year because I’m literally driving it. Same in 2021, 2022 and even last year. I don’t know the the, the, the boy who cried FSD, but I I man, I I think I think we’ll be better than human by the end of this year. This year, it all hinges on August 8th, which Elon Musk has teased Could be Tesla’s big robo taxi unveiling. But Musk has overplayed his hand before, and though FSD might be Elon Musk’s boldest vision yet, we could be waiting a lot longer.