The Obvious Lie Tesla Fanboys Cling To

Is Tesla actually cheaper than Waymo?

Photo by Hoseung Han on Unsplash

While Musk’s long-promised Robotaxis get down to the business of committing constant traffic violations, those who have dumped their life savings into Tesla are busy trying to convince themselves and us that the pathetic meme-obsessed emperor is, in fact, not naked. They claim that Waymo’s cars, Tesla’s main competitor, cost ten times as much due to their complex LiDAR and radar systems and that Tesla’s AI is advancing at a breathtaking pace, which will supposedly make them safer than a human driver in the next few years. Naturally, they want you to ignore the trash fire that was Tesla’s Robotaxi rollout, see the hypothetical “big picture”, go long on Tesla and invest it all in Musk. This stance, of course, has nothing to do with the fact that if Tesla’s robotaxis aren’t a wildly profitable runaway success, the company’s value will plummet, taking these guys’ entire unbalanced stock portfolio down with it. This bias alone should be enough for us to ignore them, but sometimes it’s good to break through the propaganda, touch grass, and figure out what is actually the case. So let’s take a moment to debunk this lie.

Yes, Waymo cars are more expensive, but not ten times so, as I have seen multiple Musk fans claim. Tesla’s current Robotaxi seems to be based on a mid-range Model Y, which costs around $54,000 (with FSD purchased), and because Tesla’s profit margin is currently only 2%, that is likely close to how much it costs Tesla to put these Robotaxis on the street. Meanwhile, analysts have found that each Waymo vehicle costs a little north of $120,000, which is just over twice as much.

But have you ever noticed how taxi drivers often drive really nice cars? Why? Wouldn’t it make sense to get the cheapest car and maximise profits? Well, no, because in the taxi game, it’s all about cost per mile, not cost to buy in, which is exactly where Waymo has beaten Tesla.

Waymo recognised this fundamental difference from the start and has optimised its system accordingly. They have modified the cars to last longer. Their system uses LiDAR, radar, computer vision, and high-fidelity local maps to enable them to operate in adverse conditions, unlike FSD, which turns off in the rain, which significantly increases the number of miles Waymo cars can cover in a year and over their lifetime. This variety of sensor types and local map data provide the system with multiple levels of redundancy, allowing it to increase safety while reducing the performance demand on the AI, which in turn reduces the cost of training, maintaining, and using the AI. The multiple levels of redundancy also decrease the amount of oversight the vehicles need, as they operate at a higher safety margin, which significantly reduces the cost.

As a result, Waymo’s current cost per mile is only $0.30.

Tesla has never publicly stated their cost per mile, so let’s calculate an estimate for comparison. A Model Y can last 300,000 miles if properly maintained and will consume roughly $10,000 worth of energy to cover that distance. Let’s be incredibly generous and say the vehicle never needs a service or maintenance during this time. Tesla’s FSD system (Full Self-Driving) is a general-purpose vision-only system. This means it only uses nine cameras to sense the world around it, and the system has no local map data to check against. As such, not only does the system have zero safety redundancy, meaning the AI must be nearly 100% accurate for it to drive safely, but the AI also has a far more challenging task than Waymo’s. This means that building, training and operating this AI will be far more expensive than Waymo’s. This is also why FSD can’t work in the rain when Waymo can. It’s also why Tesla is almost certainly selling FSD at a loss at $8,000 per car. Sadly, we don’t know the per-mile cost breakdown of FSD, so let’s be incredibly generous and assume that for the entire 300,000 miles, FSD costs Tesla only $8,000.

But wait, there is another cost. FSD is so unsafe that Tesla requires a safety driver to be in the passenger seat. Let’s be generous and say they are paid the minimum wage of $7.25, and the average speed for that 300,000 miles is 30 mph, amounting to a lifetime safety driver cost of $72,500.

Adding up the vehicle cost, FSD cost, energy cost and safety driver cost, these Robotaxis will optimistically cost $136,500 over a 300,000-mile lifetime, for a per-mile cost of $0.45.

For the Musk die-hards out there, that is higher than Waymo’s $0.30 per mile.

And this extra cost doesn’t have any safety benefits. Crowdsourced data of current FSD v13 users shows it averages 493 miles between critical disengagements. However, this is not an accurate number, as these users don’t use FSD in more complex situations, so the actual number will be far lower. Meanwhile, Waymo’s average distance between critical disengagements is over 17,000 miles. For comparison, the average human driver goes roughly 164,000 miles between accidents. So, even Waymo has a long way to go before it can claim to be better than us.

Now, some Tesla fans will call foul on my rough estimate, but to be honest, that doesn’t matter, because Waymo’s next-gen robotaxis will be cheaper than a Tesla anyway. Rather than using a modified $70,000 Jaguar, they are switching to a more affordable Geely-based five-seat EV, and thanks to bulk ordering, these are expected to cost around $28,000 per vehicle. Furthermore, LiDAR and radar sensors have become significantly less expensive than they once were. LiDAR sensors, which once cost $5,000, can now be found for just $200. Therefore, the autonomous driving hardware for this new robotaxi will cost $2,000 per vehicle.

Waymo’s next-gen robotaxis are likely to cost less than $35,000 to put on the road, which is less than Tesla’s cheapest car, and only marginally more than Musk’s promised $30,000 price for the Cybercab — though Musk has a truly horrific track record of price promises.

In other words, Tesla has no cost advantage at all, either in terms of vehicle purchase or on a per-mile basis.

In fact, Tesla’s no-redundancy and general-purpose AI approach actually gives them a considerable cost and development disadvantage.

This is for three reasons: Tesla’s dependence on AI performance for safety, general AI vs constrained AI, and the diminishing returns of AI training.

Okay, so as I have discussed a million times by now, Tesla’s FSD has no redundancy (read more here). The single sensor type and single driving AI force the entire system’s performance demands onto that AI. In turn, this AI must be almost entirely accurate to make FSD sufficiently safe. Meanwhile, systems with redundancy, such as Waymo’s, which use multiple sensor types and multiple driving AIs, can verify their input and output against numerous data points and disregard any anomalies. This means their AI doesn’t have to be as accurate for the vehicle to drive safely, as other mechanisms ensure safety. This reduces the cost of training, maintaining, and using the AI compared to Tesla’s, as the AI doesn’t need to be insanely accurate.

Waymo’s AI is also more constrained than Tesla’s FSD. It uses high-fidelity 3D maps of its local area to help determine its location and its actions in that location. This external touchstone constrains the AI’s task by reducing potential interpretations and potential outputs, enabling it to be far more accurate. Sadly, this also means Waymo operations are currently restricted to areas that are sufficiently mapped, but as they continue to map more areas, this will no longer be an issue. Tesla doesn’t use local maps; instead, its AI is general-purpose and expected to understand how to drive in any given situation. This wildly unconstrained approach, with far more potential interpretations and outcomes, significantly reduces the accuracy of the AI, requiring its model to be exponentially larger to meet similar safety margins as Waymo’s more constrained model.

Why is that a problem? Well, AI training has reached a point of diminishing returns. I have written about this quite a bit over the years (read more here), but the general gist is that for AI to keep increasing at a linear pace — particularly for general applications like FSD and ChatGPT — the training database and training costs need to increase exponentially. We are now at a point where simply improving these models’ performance by a few percent requires ten times the training data. Because training costs also increase exponentially with the increase in training data (since each data point effectively needs to be assessed against every other data point), the cost to train and operate this larger model is far more than ten times that of the previous model.

OpenAI hit this ceiling a while ago but has gained “improvement” by optimising their models to be better at specific tasks. Unfortunately, this has meant these models are worse than their older models at general tasks. OpenAI can get away with this, but Tesla’s FSD cannot for obvious reasons.

So, Tesla’s AI approach is not only far harder to make accurate due to its less constrained application, but also because of the lack of redundancy and external safety measures, Tesla is totally reliant on creating a fully accurate AI to enable their robotaxis to run safely. And, thanks to the diminishing returns of AI training, developing such an AI will cost exponentially more than Waymo’s.

So, the actual buy-in cost of Tesla’s Robotaxi — that is, actually developing the system, and not the price of the vehicles as Tesla fanboys claim — is guaranteed to be far more expensive than Waymo’s development costs.

And the numbers demonstrate this point beautifully. Waymo has spent a whopping $25.45 billion to date developing its autonomous driving system.

From 2014 to 2023, Tesla reportedly spent over $10 billion on FSD, and in 2024 alone, it was predicted to spend $10 billion on AI infrastructure and training to create FSD v13.

So, the companies have spent roughly the same amount on robotaxi development. Yet Waymo is over 34 times safer than Tesla’s FSD and is already conducting over 200,000 rides per week.

Tesla has no operational cost advantage over Waymo. In fact, the opposite is true. Tesla’s moronic approach to building a self-driving system means it will take even more tens of billions of dollars in AI development for them to make a robotaxi as safe as Waymo’s and eliminate their safety driver, which they need to do in order to have a hope in hell of matching Waymo’s cost per mile.

Oh, and here’s the real kicker: even with Waymo being this far ahead, with a much larger operation, significantly safer cars, and a cheaper per-mile cost, they are still unprofitable.

In other words, there is no way for Tesla to catch up with Waymo practically or financially. So, not only is the emperor naked, but he has skid marks halfway up his back.

---

Thanks for reading! Don’t forget to check out my YouTube channel for more from me, or Subscribe. Oh, and don’t forget to hit the share button below to get the word out!

---

Sources: The Guardian, The Street, Forbes, Ivey Business Review, Teslarati, EW, Inside EVs, Will Lockett, Will Lockett, Energy Pricing.com, Macro Trends