One of the things that Google continues to brag about is how many miles its cars have been driven around under its software’s control.
The company also announced that, since 2009, its fleet of 55 autonomous cars have driven a total of over 1.4 million miles on actual roads around the Bay Area and Austin, Texas.
But how does that likely compare to the miles Teslas have racked up under Autopilot in the months since it was introduced? Well, Tesla is currently coy about those numbers, so let’s make some assumptions and project a likely range for the head-to-head drive-off between the Google fleet of bubbles, Priuses, and others, and the Tesla Model S. We’ll ignore the Model X, as the numbers of delivered vehicles are still small, and also to be more conservative in our estimates.
Let’s start with the number of cars sold with Autopilot. The underlying hardware was introduced in September 2014. In the 5 quarters since then, Tesla has sold 60,414 cars — just going on the quarterly figures. These are the cars which could have used Autopilot.
Autopilot was introduced October 14, 2015. Let’s consider three months of driving for 60,000 cars to account for the discrepancy in cars purchased after that date in Q4 2015 and for the few weeks since January 14, 2016, once again being conservative as it’s currently mid-March. After all, we are comparing Google’s fleet over 7 years vs Tesla’s over 3 months, so we have to give Google a chance.
Let’s consider two scenarios, a low end and a high end.
- Teslas driven less than average because they are second or third cars of well-off people. Assume 10,000 miles annually on average. This gives 2,500 miles for the 3 months per car.
- 20% of owners are regular Autopilot users.
- They use it for 25% of miles driven.
- Teslas are driven more than average because they are awesome, new, and free to take across the country, so 14,000 miles annually on average. This gives 3,500 for the 3 months per car.
- 40% of owners are regular Autopilot users.
- They use it for 40% of miles driven.
A little math gives us a range of 7.5 million to 33.6 million miles driven under Autopilot, with all of that data streamed up to Tesla’s cloud machine learning centre. That’s 5.4 to 24 times as many miles in 3 months as the Google cars drove in 7 years, which is kind of what I expected. It’s around 200 to 850 times as many miles every month of real-world experience with autonomous driving. It’s the power of multipliers. 60,000 Teslas versus a handful of Google cars can’t help but rack up the miles.
For Tesla to have driven as few miles as Google’s fleet under autonomous control, drivers would have to be averaging only 2,000 miles per year or only 4% of them would be using Autopilot or only using it on 5% of their driven miles, or other variants of those numbers, which seems fairly unrealistic. Tesla owners skew to people who drive a lot, multi-car owners are making it clear in forum after forum that their other cars are collecting dust once the Tesla arrives, and Tesla drivers skew to early-adopting technophiles who love the idea of Autopilot.
Of course, Google knows this, so it is now touting its virtual miles driven model, which obviously Tesla has as well. What does Google claim its virtual mileage is? About 3 million miles a day, which is an awful lot. But given that Google cars can only operate on roads scanned and mapped with enormous precision, it’s unclear what their simulator is simulating. If it’s the few square miles they’ve already driven 1.4 million miles on, the area where they recently collided with a bus, they are not experiencing much in terms of new terrain, highways, etc. Perhaps they’ve added SimCity-like modules in, and their car now occasionally runs into Godzilla.
Any way you run the numbers, it seems apparent that Tesla Autopilot continues to rack up impressive gains on the Google approach, especially as they are shipping over 10,000 Autopilot equipped Model Ss and Model Xs quarterly right now.