AI
Saturday, March 28, 2026 5:17:41 PM
The Tesla Robotaxi Story Is A Myth: Why I'm Maintaining My Strong Sell Into Q1 Earnings
https://seekingalpha.com/article/4886338-the-tesla-robotaxi-story-is-a-myth-why-im-keeping-my-strong-sell-into-q1-earnings
.png)
Summary:
Tesla, Inc. remains fundamentally unproven in autonomous mobility, lagging competitors in safety, transparency, and operational scale.
I maintain a Strong Sell rating on TSLA due to its shrinking core business, compressed margins, and overreliance on sentiment rather than fundamentals.
TSLA's robotaxi efforts are hampered by inferior data quality, lack of regulatory transparency, and a crash rate nine times worse than human drivers.
With Q1 deliveries expected to decline and R&D reliant on automotive cash flow, TSLA's business model faces mounting risks.
.png)
I was recently asked which was the best robotaxi stock play for 2026, here on Seeking Alpha, and I broke down a list of companies with a working operation and/or a clear path to monetization. I noticed a few of you in the comments asking why Tesla didn't even make the list, almost as if I could forget the biggest "bull case" in the industry.
I can assure you, I didn’t forget.
The core of my skepticism comes from the technological divide I discussed in my recent Mobileye article. Mobileye is a company that has been a leader in autonomy for cars, and was originally supposed to provide Tesla Autopilot, but the partnership came to an abrupt end due to a safety dispute in 2016. Mobileye management works on an advanced system, which will be able to determine the best possible behavior for a driverless car in any critical situation.
I bring it up in this article for an argument that Mobileye's management made. They say that thousands of hours of video will never “teach” the neural network how to respond in the infinite number of scenarios, which can happen on the road, some of which may never repeat again. Mobileye, for example, uses a "fast and slow" architecture that mimics human reflexes and deep cognitive processing, and they say that miles collected are an incredibly slow way to teach a car.
Mobileye began collecting data way before Tesla, by the way. It had 200 PB of data in 2022, but since then, the company has stopped disclosing it, the same as the miles collected, because the market realized that data doesn’t get them closer to safe driverless technology. Other competitors like Alphabet also stated repeatedly that relying on a single sensing modality will not bring a sufficient level of safety.
Meanwhile, Tesla continues to insist that cameras are enough, and of course, there are always big headlines about the miles Tesla has collected, like the latest 8.5 billion-mile milestone, even though they aren’t even autonomous miles, but rather supervised miles collected from Tesla owners all over the world. Even if the data is enough, the quality of Tesla’s data falls way behind the data from Waymo, which was collected directly by the driverless cars.
We don't have to guess who is right because we have the data from Austin. Austin was the first city where Tesla’s Level 2++ was deployed, meeting Waymo’s Level 4 robotaxis. The results show everything you need to know about Tesla. The company is best at what is unimportant. For example, it wins "most miles driven," but only Waymo was able to scale to 200 fully autonomous taxis, while Tesla is stuck with road tests in small, geofenced areas. Out of 43 vehicles Tesla has deployed in Austin, only 8 are unsupervised, and even these 8 are believed to be supervised at a distance.
I have to say, I am really concerned about how little data those companies are required to share. It’s absolutely outrageous that a vehicle without the Level 4 clearance is allowed to be “unsupervised,” even in a geofenced area.
And thankfully, it seems I am not the only one concerned over this lack of transparency, and "cutting of corners" when it comes to rules. This month, the NHTSA started an engineering analysis into Tesla’s Full Self-Driving (FSD) system, covering over 3 million vehicles. The probe is specifically looking at the system's "degradation detection," or rather, its failure to detect it. Regulators are concerned that Tesla’s vision-only approach cannot handle poor roadway conditions like sun glare, fog, or dust (as a matter of fact, there are many videos online proving it). The NHTSA even noted cases where FSD completely lost track of or never even detected a vehicle in its path due to visibility issues:
NHTSA cited several crashes, including one fatal, in which the system failed to adequately warn drivers in low-visibility conditions such as sun glare, fog, or dust. The investigation, which began in 2024, is now moving to an “engineering analysis,” a deeper review that could lead to a recall or other enforcement action.
While Elon Musk has promised for almost a decade now a fully autonomous vehicle, the data we have today and intensifying skepticism from analysts and experts, including probes, suggest otherwise. Also, as Tesla stops Model S and X production and wants to make the "Optimus" robot factory instead, it feels like the company is just jumping from one unfulfilled promise to the next.
Tesla Robotaxi Progress
We are now nine months into the Austin launch, and Tesla has a tiny fleet of about 40 vehicles, even though in October 2025, Musk promised "500 or more" robotaxis in the Greater Austin area by the end of the year. Based on crowdsourced data, the availability of these cars is often below 20%, and of course, as I already said, they are mostly not unsupervised. To put that in perspective, Waymo is operating over 2,500 fully driverless (they are unsupervised; it turns out it is possible with the right technology!) taxis across several U.S. cities. Waymo has over 127 million autonomous miles, and its research shows its crash rate is 85% lower than that of human drivers, and way lower than that of Tesla, which previously had a crash rate 4 times higher than that of human drivers. Also, Waymo’s data is official, provided by the company, and includes all of the details, including differences to the benchmark, the severity of damage, etc. Tesla doesn’t provide research like this, most likely, as the results would disappoint the believers.
.png)
If you think it will improve over time, for now, it doesn’t look like this. Based on the 15 crash incidents reported from June 2025 until March 2026, and the estimated mileage of the fleet, Tesla’s "robotaxis" are crashing roughly once every 57,000 miles. For the average American driver, that figure is one crash every 500,000 miles. This means Tesla’s supervised system is currently performing about 9 times worse than a human driver. And remember, these are incidents that happened even with a human supervisor present who is supposed to prevent them. Only 8 out of 43 cars are stated as “unsupervised.”
When I look at these numbers, I realize how right Mobileye is in its statement:
The sample complexity of policy is very, very high, again, because it's a multi-agent compounding effect, right? The host vehicle performs an action in the world and this action affects other road users. Therefore, there's this compounding effect. So when you have something of a very high complexity, you need a lot, a lot of data. Now the data coming from real world is limited. Even if you have millions of cars sending data, it's still limited. You can do -- you can run on a simulator, say, photorealistic simulators and do training over a simulator but then you are compute bound because then compute becomes a bottleneck.
And by the way, I am not Mobileye’s fangirl. I rated the stock as a Hold for now, but I really liked the fact that Mobileye is a tech-first company. The management there has a large academic background. I respect the company, and I trust their judgment, because to me it seems that studying and developing the technology is a goal number one for them, and not the business. It might not be the best fact from a shareholder’s perspective, but from analysts, I think it is very interesting.
Also, I am pretty cautious about the whole idea, adoption, and, most importantly, monetization of robotaxis. In Seeking Alpha Asks, I said:
Speaking of the robotaxi space, I’ve been skeptical about the wide-scale adoption of this technology since the beginning, and I remain skeptical today. For me, as a consumer, the fact that a car is driverless doesn’t provide any specific value if it is not significantly cheaper than a regular taxi service. Autonomous cars, which can free my hands and potentially free my attention while driving, could be a game changer, especially when we are stuck in traffic for hours. But this is not an advantage for robotaxis. For this reason, I would be investing in mobility, and if I wanted some robotaxi exposure, I’d choose names that focus on different types of mobility.
When I add to my concerns the lack of transparency from Tesla, I think it’s clear I can’t be bullish on this business segment of Tesla. While other autonomous companies, like Waymo, Amazon with its Zoox, or even Baidu with Apollo Go, provide full descriptions of what happened during their crashes to the NHTSA, Tesla redacts everything, marking it as "confidential business information," which means the public and researchers have no way of knowing what happened.
It’s also easy for Tesla to report "fewer crashes" in a month, as they did in February, without mentioning that due to the weather conditions, the fleet didn’t operate for several days, and also, February is a short month, so fewer miles equals fewer crashes. It’s not Tesla’s win. Also, without data about mileage, we can’t claim there was an improvement. If Tesla were truly confident that its tech is safe and working as expected, it wouldn't have anything to hide.
Before Q1
To make matters worse for the shareholders, the market braces as Tesla is heading into next week’s Q1 deliveries report. Analysts at UBS and RBC are cutting their forecasts, with some expecting deliveries to drop as low as 345,000 units, which would be a strong decline both sequentially and year-over-year.
By now, I know the main argument of the fans is that deliveries don't matter because Tesla is an "AI and Robotics" company. And you could be right, because except for the AI and robotics, Tesla doesn’t have any hopes for the future, but in reality, the automotive business funds the R&D for things like Optimus and the Cybercab, and everything else. If the core business continues to shrink, there will be no cash to wait for the new era of Tesla innovation to kick in.
As I mentioned in my previous article, Tesla is winding down Model S and X production to make space for robot factories, but all I see in this is that the demand for their premium vehicles is declining, and margins are being crushed by competition. In my previous article, you can also read how "Musk Premium" is being moved away from Tesla and toward his private ventures like SpaceX and xAI. I think Elon recognizes that Tesla didn’t go as planned, so now Tesla shareholders will fund high-risk AI experiments to hopefully stop the company from crashing.
Conclusion
Writing this article feels a bit like describing a parallel reality. I have been breaking down financials and studying different management styles for a few years now, and while it’s not a lot, Tesla always seems to exist in a category of its own for me personally. I recognize that I have strong opinions on this, and it is perfectly fine if we disagree (that is what makes a market), but for me, the disconnect between the narrative and the facts is inconceivably enormous.
While sentiment can help boost my profits, and I often use it, my investment philosophy has always been to prioritize the business. For these reasons, I am maintaining my Strong Sell rating on Tesla.
https://investorshub.advfn.com/boards/read_msg.aspx?message_id=177083451
.png)
https://seekingalpha.com/article/4886338-the-tesla-robotaxi-story-is-a-myth-why-im-keeping-my-strong-sell-into-q1-earnings
.png)
Summary:
Tesla, Inc. remains fundamentally unproven in autonomous mobility, lagging competitors in safety, transparency, and operational scale.
I maintain a Strong Sell rating on TSLA due to its shrinking core business, compressed margins, and overreliance on sentiment rather than fundamentals.
TSLA's robotaxi efforts are hampered by inferior data quality, lack of regulatory transparency, and a crash rate nine times worse than human drivers.
With Q1 deliveries expected to decline and R&D reliant on automotive cash flow, TSLA's business model faces mounting risks.
.png)
I was recently asked which was the best robotaxi stock play for 2026, here on Seeking Alpha, and I broke down a list of companies with a working operation and/or a clear path to monetization. I noticed a few of you in the comments asking why Tesla didn't even make the list, almost as if I could forget the biggest "bull case" in the industry.
I can assure you, I didn’t forget.
The core of my skepticism comes from the technological divide I discussed in my recent Mobileye article. Mobileye is a company that has been a leader in autonomy for cars, and was originally supposed to provide Tesla Autopilot, but the partnership came to an abrupt end due to a safety dispute in 2016. Mobileye management works on an advanced system, which will be able to determine the best possible behavior for a driverless car in any critical situation.
I bring it up in this article for an argument that Mobileye's management made. They say that thousands of hours of video will never “teach” the neural network how to respond in the infinite number of scenarios, which can happen on the road, some of which may never repeat again. Mobileye, for example, uses a "fast and slow" architecture that mimics human reflexes and deep cognitive processing, and they say that miles collected are an incredibly slow way to teach a car.
Mobileye began collecting data way before Tesla, by the way. It had 200 PB of data in 2022, but since then, the company has stopped disclosing it, the same as the miles collected, because the market realized that data doesn’t get them closer to safe driverless technology. Other competitors like Alphabet also stated repeatedly that relying on a single sensing modality will not bring a sufficient level of safety.
Meanwhile, Tesla continues to insist that cameras are enough, and of course, there are always big headlines about the miles Tesla has collected, like the latest 8.5 billion-mile milestone, even though they aren’t even autonomous miles, but rather supervised miles collected from Tesla owners all over the world. Even if the data is enough, the quality of Tesla’s data falls way behind the data from Waymo, which was collected directly by the driverless cars.
We don't have to guess who is right because we have the data from Austin. Austin was the first city where Tesla’s Level 2++ was deployed, meeting Waymo’s Level 4 robotaxis. The results show everything you need to know about Tesla. The company is best at what is unimportant. For example, it wins "most miles driven," but only Waymo was able to scale to 200 fully autonomous taxis, while Tesla is stuck with road tests in small, geofenced areas. Out of 43 vehicles Tesla has deployed in Austin, only 8 are unsupervised, and even these 8 are believed to be supervised at a distance.
I have to say, I am really concerned about how little data those companies are required to share. It’s absolutely outrageous that a vehicle without the Level 4 clearance is allowed to be “unsupervised,” even in a geofenced area.
And thankfully, it seems I am not the only one concerned over this lack of transparency, and "cutting of corners" when it comes to rules. This month, the NHTSA started an engineering analysis into Tesla’s Full Self-Driving (FSD) system, covering over 3 million vehicles. The probe is specifically looking at the system's "degradation detection," or rather, its failure to detect it. Regulators are concerned that Tesla’s vision-only approach cannot handle poor roadway conditions like sun glare, fog, or dust (as a matter of fact, there are many videos online proving it). The NHTSA even noted cases where FSD completely lost track of or never even detected a vehicle in its path due to visibility issues:
NHTSA cited several crashes, including one fatal, in which the system failed to adequately warn drivers in low-visibility conditions such as sun glare, fog, or dust. The investigation, which began in 2024, is now moving to an “engineering analysis,” a deeper review that could lead to a recall or other enforcement action.
While Elon Musk has promised for almost a decade now a fully autonomous vehicle, the data we have today and intensifying skepticism from analysts and experts, including probes, suggest otherwise. Also, as Tesla stops Model S and X production and wants to make the "Optimus" robot factory instead, it feels like the company is just jumping from one unfulfilled promise to the next.
Tesla Robotaxi Progress
We are now nine months into the Austin launch, and Tesla has a tiny fleet of about 40 vehicles, even though in October 2025, Musk promised "500 or more" robotaxis in the Greater Austin area by the end of the year. Based on crowdsourced data, the availability of these cars is often below 20%, and of course, as I already said, they are mostly not unsupervised. To put that in perspective, Waymo is operating over 2,500 fully driverless (they are unsupervised; it turns out it is possible with the right technology!) taxis across several U.S. cities. Waymo has over 127 million autonomous miles, and its research shows its crash rate is 85% lower than that of human drivers, and way lower than that of Tesla, which previously had a crash rate 4 times higher than that of human drivers. Also, Waymo’s data is official, provided by the company, and includes all of the details, including differences to the benchmark, the severity of damage, etc. Tesla doesn’t provide research like this, most likely, as the results would disappoint the believers.
.png)
If you think it will improve over time, for now, it doesn’t look like this. Based on the 15 crash incidents reported from June 2025 until March 2026, and the estimated mileage of the fleet, Tesla’s "robotaxis" are crashing roughly once every 57,000 miles. For the average American driver, that figure is one crash every 500,000 miles. This means Tesla’s supervised system is currently performing about 9 times worse than a human driver. And remember, these are incidents that happened even with a human supervisor present who is supposed to prevent them. Only 8 out of 43 cars are stated as “unsupervised.”
When I look at these numbers, I realize how right Mobileye is in its statement:
The sample complexity of policy is very, very high, again, because it's a multi-agent compounding effect, right? The host vehicle performs an action in the world and this action affects other road users. Therefore, there's this compounding effect. So when you have something of a very high complexity, you need a lot, a lot of data. Now the data coming from real world is limited. Even if you have millions of cars sending data, it's still limited. You can do -- you can run on a simulator, say, photorealistic simulators and do training over a simulator but then you are compute bound because then compute becomes a bottleneck.
And by the way, I am not Mobileye’s fangirl. I rated the stock as a Hold for now, but I really liked the fact that Mobileye is a tech-first company. The management there has a large academic background. I respect the company, and I trust their judgment, because to me it seems that studying and developing the technology is a goal number one for them, and not the business. It might not be the best fact from a shareholder’s perspective, but from analysts, I think it is very interesting.
Also, I am pretty cautious about the whole idea, adoption, and, most importantly, monetization of robotaxis. In Seeking Alpha Asks, I said:
Speaking of the robotaxi space, I’ve been skeptical about the wide-scale adoption of this technology since the beginning, and I remain skeptical today. For me, as a consumer, the fact that a car is driverless doesn’t provide any specific value if it is not significantly cheaper than a regular taxi service. Autonomous cars, which can free my hands and potentially free my attention while driving, could be a game changer, especially when we are stuck in traffic for hours. But this is not an advantage for robotaxis. For this reason, I would be investing in mobility, and if I wanted some robotaxi exposure, I’d choose names that focus on different types of mobility.
When I add to my concerns the lack of transparency from Tesla, I think it’s clear I can’t be bullish on this business segment of Tesla. While other autonomous companies, like Waymo, Amazon with its Zoox, or even Baidu with Apollo Go, provide full descriptions of what happened during their crashes to the NHTSA, Tesla redacts everything, marking it as "confidential business information," which means the public and researchers have no way of knowing what happened.
It’s also easy for Tesla to report "fewer crashes" in a month, as they did in February, without mentioning that due to the weather conditions, the fleet didn’t operate for several days, and also, February is a short month, so fewer miles equals fewer crashes. It’s not Tesla’s win. Also, without data about mileage, we can’t claim there was an improvement. If Tesla were truly confident that its tech is safe and working as expected, it wouldn't have anything to hide.
Before Q1
To make matters worse for the shareholders, the market braces as Tesla is heading into next week’s Q1 deliveries report. Analysts at UBS and RBC are cutting their forecasts, with some expecting deliveries to drop as low as 345,000 units, which would be a strong decline both sequentially and year-over-year.
By now, I know the main argument of the fans is that deliveries don't matter because Tesla is an "AI and Robotics" company. And you could be right, because except for the AI and robotics, Tesla doesn’t have any hopes for the future, but in reality, the automotive business funds the R&D for things like Optimus and the Cybercab, and everything else. If the core business continues to shrink, there will be no cash to wait for the new era of Tesla innovation to kick in.
As I mentioned in my previous article, Tesla is winding down Model S and X production to make space for robot factories, but all I see in this is that the demand for their premium vehicles is declining, and margins are being crushed by competition. In my previous article, you can also read how "Musk Premium" is being moved away from Tesla and toward his private ventures like SpaceX and xAI. I think Elon recognizes that Tesla didn’t go as planned, so now Tesla shareholders will fund high-risk AI experiments to hopefully stop the company from crashing.
Conclusion
Writing this article feels a bit like describing a parallel reality. I have been breaking down financials and studying different management styles for a few years now, and while it’s not a lot, Tesla always seems to exist in a category of its own for me personally. I recognize that I have strong opinions on this, and it is perfectly fine if we disagree (that is what makes a market), but for me, the disconnect between the narrative and the facts is inconceivably enormous.
While sentiment can help boost my profits, and I often use it, my investment philosophy has always been to prioritize the business. For these reasons, I am maintaining my Strong Sell rating on Tesla.
https://investorshub.advfn.com/boards/read_msg.aspx?message_id=177083451
.png)
Bearish
Fortes Fortuna Juvat
