Last week, Tesla held an event focused on their advances in autopilot and what they call “full self driving”. There, nearly three hours into the event, they made the announcement: not only will they have fully autonomous vehicles ready years ahead of the industry’s best estimates, Tesla expects to have a fleet of one million robotaxis on the road in 2020.
This time, there wasn’t the usual whooping and hollering from the Tesla loyalists. Instead, this was an event for investors. It was a more steadied and almost scholarly affair, with dizzyingly deep dives into the technology powering Tesla’s self-driving ambitions.
And the announcements—carrying world-changing ramifications, if the numbers are right—were delivered aloud into otherwise pindrop silence. It seemed the collective reaction of those in the room was a furrowed brow. Could Elon be believed? Still, onward he pressed, proclaiming a grand vision as though he’d only just returned from the future to share what he had seen. Maybe the claims seemed too far-fetched, maybe the event was too long, or perhaps it’s because the investors at the event were, for the most part, non-technical.
As someone in the tech industry myself, I had a hard time keeping up with the swirling Acronym Soup shared by presenters like Pete Bannon, esteemed former-Apple chip designer. But the overall message resonated: Tesla is ahead, they argued, because Tesla has the data. Whereas nearly every competitor is relying on Lidar, Tesla has placed their self-driving bets almost entirely on computer vision. Choosing to rely on cameras and cheap radar + ultrasonics has allowed them to deploy these sensors on every car they’ve sold for the past several years. Having sensors on every vehicle means they’ve been able to collect data from every mile driven from every Tesla produced in the past several years. That’s a huge number, and it’s increasing rapidly.
Lidar, meanwhile, is power hungry and expensive, adding anywhere from around $7K to $70K to the cost of the vehicle. The upshot is that the major Lidar-based competitors have several hundred cars on the road each, while Tesla has nearly half a million. And machine learning, which is needed for object recognition in any self-driving system, depends on access to mountains of data. In fact, it thrives on it — there’s a direct correlation between how much data you throw at a neural network and the quality of the results. Because they make their own cars, and because they’ve bet on cheaper sensors, Tesla is now sitting on an unmatched (and possibly unmatchable) pile of data, and that pile grows with each mile driven, with the rate of growth multiplying with each new vehicle sold.
In that light, Elon emphatically assures us their self-driving capabilities are improving “exponentially”, which would make the advent of full autonomy arrive much sooner than expected (and, frustratingly, will also make estimating its exact arrival date even more difficult). To demonstrate these advances, they gave investors fully-autonomous rides in standard off-the-shelf Teslas with remarkable capability improvements over the previous generation of software. In other words, given their advances in chip hardware and their substantial lead in real-world data, the final piece of their self-driving puzzle is software. Software which, once ready, can be deployed at the push of a button.
With the facts and figures out of the way, they delivered the real shocker of the event: robotaxis. As early as next year, your Tesla will be able to drive you home as you read a book, they say, as well as go off to make money for you as an autonomous robotaxi whenever you like. You can make money while you sleep, or earn a second paycheck while at your day job. The car can pay for itself, and then some.
Now, even taking that with the massive tablet of salt called “regulatory approval” combined with a healthy schedule adjustment to pad for Elon Time™, that’s still a staggeringly audacious proposal. Especially when you consider their back-of-the-napkin math:
- The base self-driving Tesla costs about $38K.
- As a robotaxi, the car will be able to earn around $30K per year. This assumes rides at half the cost of a Lyft or Uber, with half of the miles travelled being empty “dead legs”.
- Tesla cars will be rated for one million miles, including the battery.
- In the lifespan of the car, it can earn approximately $200K of income for the owner.
A $38K car bringing in $200K of income on its own? That’s insane. That’s impossible. And that’s being spearheaded by a team famous for achieving the impossibly insane.
Whether or not they accomplish this latest feat, it’s a fascinating thought experiment. Assuming the tech works, what would happen next? Let’s assume this back of the napkin math is right, and that every car is worth more than what you’ll pay for it:
- The existence of “a machine that prints money” will put extraordinary demand on Tesla vehicles. People will buy fleets of entry-level Model 3s to get as many cars on the road as possible and maximize the number of concurrent rides. Most cars will likely be black, as it is the lowest cost and there’s not much incentive for differentiation.
- These cars can run 24/7, automatically charging themselves when needed. They’ll only need to get pulled off the road for cleaning and as-needed maintenance.
- City streets will be flooded with Model 3s. Tesla will make money on every sale and a cut of every ride, sending their revenue (and their stock) skyrocketing. With a pushbutton software deployment, Tesla effectively transforms from a car company into a ridesharing company.
- Riders will have the choice between standard ride sharing apps and Tesla robotaxis, which they estimate will be half the price. If the tech does what it says, users will take the half-priced rides nearly every time. Uber and Lyft are at serious risk, and many (or most) drivers will be out of a job. There will be a backlash from rideshare drivers and the companies themselves.
- The “secret” of the too-good-to-be-true cars will spread quickly, and robotaxi supply will increase until it starts to outstrip profitability. Bear in mind, there is A LOT of demand for ride sharing, which means this number could be very, very high.
- High demand for Model 3s to be used as fleet robotaxis will make it hard to buy them for personal use. Owning a personal Model 3 may become somewhat rare. Most people will interact with Tesla vehicles as robotaxis, alone.
- There will be substantially increased pressure on charging infrastructure, both with charging stations (especially self-charging stations) and the power grid itself. Increased daytime charging may put pressure on peak power use which would drive electricity prices up. Tesla will need to continue increasing battery storage at Superchargers to deliver off-peak energy collected overnight at a much larger scale.
- Until automatic self-charging stations are deployed at scale, Supercharger attendants may be needed to tend to the influx of driverless cars.
- There will be a huge aftermarket years down the road for used Model 3 robotaxis near the end of their million-mile life.
- It may be cheaper to take robotaxis than to drive your own car. Used car prices will then plummet as people shift to robotaxis to save money and as unemployed rideshare drivers seek to unload expensive vehicles that they can no longer afford. As regulatory restrictions will vary by location, the market for used vehicles will likely narrow down region by region as robotaxis come online.
As with the current ubiquity of ridesharing traffic today, the demand for robotaxis may be so high that entry-level Model 3s become the majority of cars on the street. The number of these robotaxis will increase unless and until we encounter one of several possible limiting factors:
- Government regulations stymie the rollout of driverless cars
- Power infrastructure limitations (not enough charging capacity) prevent cars from getting charged quickly enough to remain profitable
- Power costs climb too high
- The supply of robotaxis climbs so high that each car has too much competition to be profitable (their car doesn’t get selected for enough rides anymore)
- Backlash from ridesharing interests prevents rapid expansion, possibly through government action or boycott
- Tesla is unable to keep up with a spike in demand (which would only slow the growth of robotaxis, not limit it)
- None of these other factors turns out to be limiting, and the streets simply run out of room.
Tesla may choose to limit the number of cars a single individual purchases to inhibit the creation of fleets. But as their goal is to “accelerate the world’s transition to sustainable energy”, allowing intrepid entrepreneurs to flood the roads with fleets of cheaper, safer, and greener robotaxis would certainly help this mission.
It doesn’t hurt that they’ll also be stuffing Tesla’s pockets (and their own) while they do.
What’s your take? Will Tesla be the first out the door with full self driving? Will robotaxis be the money printing machines they’re purported to be? Or will this be one promise Elon & Co. fails to deliver on?
Would love to hear your thoughts. And if you found this interesting and haven’t seen it yet, I’d encourage you to watch the announcement for yourself. Or, if you don’t have several hours to spare, check out this 15 minute cut below.
Follow me on Twitter: @verbiate
Disclosure: I own some Tesla stock.