Are we ready for robot cars?
Cars are our culture. We polish them. We name them. We produce entire motion pictures featuring cars who bring us self-understanding, triumph, and love.
Cars are not blenders on wheels. Cars are, for many, expressions of our inner selves.
But thought leader and author Tony Seba and his consultancy RethinkX believe that we are about to give our beloved cars up. All at once. In roughly a decade, probably less.
Does this even remotely make sense?
It seems unfathomable that our culture could change in such a short period of time, but economics has a way of bending culture to its will. If you doubt this, ask how long a beloved mom and pop store lasts when Walmart arrives in town, or how studiously your and your peers look through clothing tags for a Made in America badge.
And so our collective decision about whether to embrace our robot chauffeurs coms to this: Self driving cars will be unfathomably cheap. So cheap that it won’t make sense to drive the vehicle we already own.
This sounds impossible, but the logic is sound. In a previous piece, I laid out why self-driving cars are coming, and how Big Oil is vulnerable to going the way of Kodak and Blockbuster. To follow up, I wanted to share the math: 11 graphs that helped me take the journey from skeptic to believer. Hopefully they will help you understand that yes, our culture can be bought. And that’s not necessarily a bad thing.
Self-driving cars will work at 1/10th the price of a traditional cab. They will be 2-4X cheaper than owning a car and driving it yourself. They will win the battle in the marketplace, and the battle for our culture.
Here is why.
1. Self-driving vehicles will be cheap because the drivers have vanished
This is the dream of every capitalist: No driver means no wages. If a car today costs about 53.5¢ per mile to own and operate (per the IRS), and a New York City taxi charges $2 per mile to drive to your destination, you can guess that humans are responsible for much of the $1.50 difference.
A 2011 study in Ottawa gave an estimate of the cost breakdown for a taxi, and if we consider the cost of humans to be earnings + insurance (since robots should not have accidents, their insurance costs should be negligible), then people account for 56.9% of costs.
A breakdown of the costs of operating a cab. From Ottawa, 2011.
Even in long-haul trucking, with its simple routes and limited traffic, drivers earn about 30% of the gross revenue of the truck. And when you sum in the other costs of being human, including insurance, meals, lodging and the like, people account for just under half the cost of trucking – nearly the same as with a cab.
Annual cost of $103,000 per year for a truck that completes 100,000 miles per year. From the Owner Operator Independent Trucker Association
People are indeed the biggest cost of driving. But simply taking people out of the cost does not, alone, lead to anything transformational.
To understand why self-driving cars will revolutionize transport, you have to look deeper.
2. Batteries last so long, individual car owners will die before the batteries do
A group of Tesla users has crowdsourced insights onto the lifetime of Tesla batteries, and plotted them to understand how the battery degrades over life. You can see a rapid initial drop in capacity over the first 40,000 km, followed by a roughly linear fade at a rate of about 2% every 100,000 km (60,000 miles).
Self-reported Tesla battery capacity reported by drivers in the Dutch-Belgium Tesla Forum. Those in the US should note that the x-axis is in kilometers, not miles.
This failure pattern is normal for lithium ion batteries: the initial, rapid drop is a reaction between the battery’s electrolyte liquid and its electrodes to form a sort of passivation layer, akin to the protective coating on a cast iron pan. Once this layer has formed, degradation trends more or less with the number of miles driven.[1]
Today, the technology community generally considers a battery to be at the end of it’s life when it can only hold 70% or 80% of its original charge (more on this below). When you project that out, you find that a Tesla will last somewhere between 750,000 and 1,200,000 kilometers (500,000 and 750,000 miles). That’s 3-5X as long as an internal combustion engine drivetrain. That’s up to 50 years of car ownership.
Again, those in the US should note that the x-axis is in kilometers, not miles.
On a per mile basis, the capital cost of an electric vehicle may be 1/4 that of a gasoline vehicle. Yet if the car is parked 96% of the time, individual owners can’t take advantage of that savings – most of us simply won’t drive this many miles in a lifetime.
Robots, however, can.
3. Electric vehicles don’t break
Tony Seba has made an astounding claim about electric vehicles – because the electric drivetrain has about 20 parts compared with 2,000 in an internal combustion engine, EV’s will have dramatically lower repair costs. And savings on maintenance will be critical for anyone whole really intends to get 500,000 miles on a car.
Tesla seems to agree, and has offered its owners an eight year, “infinite mile” warranty. Fortunately for us, Tesla is a publicly traded company, so we can see if its repair record holds up to its marketing spin.
In its 2016 annual report, Tesla anticipated that warranty repairs would cost it $3,079 per vehicle over the lifetime of the Model S. But that’s not the only cost of driving a car – there is also tire rotation, break fluid replacement, and other maintenance, which over the four years of ownership will set back a Tesla owner another $2,400.
Is that cheap for an $80,000 vehicle? The nearest competitor to Tesla might be Mercedes, which has an average warranty cost of $2264 on its fleet, though limiting its coverage to just 4 years/50,000 miles. And its warranty does not cover everything that may need fixing in the vehicle: Edmunds expects a Mercedes E-Class owner to spend $8,414 out of pocket in those first four years.
Add up those costs, and after four years the Tesla comes out at half the price. This 2X savings is almost certainly a conservative estimate, since the Tesla warranty lasts over twice as long. And Edmunds estimates that Mercedes will cost its owner another $5476 in repairs and maintenance during year 5. I hope they have the money.
Comparison of costs in the first four years of ownership of the Tesla model S vs Mercedes E Class, two cars roughly comparable in price and features. Data on the Mercedes from Edmunds
These kinds of comparisons are imperfect, but it sure looks like Seba, and Tesla, are on to something here.
4. Electric vehicles cost less to power
Internal combustion engines are relatively inefficient: All other things being equal, an electric vehicle uses only 1/3 the energy of a gasoline vehicle. As I write this, electricity in the US costs about $0.12/kWh retail, while regular fuel costs $2.40. A these prices, a Tesla model S, which gets about 3 miles per kWh, will cost about $0.04 per mile to drive. A fairly efficient, 25 MPG gasoline sedan will require $0.10.
Advantage: Batteries.
A comparison of fuel costs of electric (solid lines) vs comparable gas vehicles (dashed lines). The 3 mile/kWh range is the standard for a Model S. The 1 mile/kWh is proposed for next generation electric trucks. The 10 MPG range is proposed for next generation diesel trucks. Data adapted from a similar plot by Idaho National Labs.
And again, this is, long-term, an underestimate. As the price of solar + battery storage comes down, there will no reason to tie a vehicle recharging station to the grid. In 10 years, the price of locally created energy could be half today’s price, or less, increasing the advantage of EV further. It doesn’t matter how cheap oil gets; even if the oil was $10 per barrel you still have to pay for refining it. EVs win this battle.
5. Multiply these factors, and self-driving cars will be mind-bogglingly cheap
Let’s go back to our big rig, and our taxi. When we exclude the costs of humans, drop fuel costs by 3X, capital costs by 4X, and maintenance costs by 2X, something really interesting happens. The cost of trucking drops by almost a factor of 5, from $103,000 per year to $23,000 per year.
The cost of a cab drops even more dramatically, to about 10% of today.
Graph plotted as % costs; Ottawa did not provide $ figures in their taxi report.
The cost of a driverless car will be 15¢-20¢ per mile. And this is the transportation revolution: At 20¢ per mile, hailing a ride in a self-driving car will cost less than driving your own car. Even if your own car is fully paid for.
A comparison of self-driving costs versus personal ownership. Personal ownership figures taken from AAA. The costs of self-driving are less than personal ownership even if the car is paid off, primarily because of the cost of fuel and personal insurance.
6. Actually, it will be cheaper than that
A Tesla Model S comes equipped with an 85 kWh battery, which gives it a range of about 250 miles on a single charge. Today, that battery costs over $10,000 per pack, and weighs 1,200 pounds – nearly as much as an entire Smart car. A Tesla sedan is like a mommy car, whose energy is sapped by lugging around a toddler car wherever it goes.
So why does a Tesla need this large, 85 kWh battery?
Tesla batteries have a range of 250 miles to allay people’s fear of running out of fuel during a trip, something called ‘range anxiety’. But for non-neurotics, is this range really necessary? The average Uber ride is just 6.4 miles. And thousands of owners of the plug-in hybrid Chevy Volt complete nearly every activity in their daily life within its battery range of 50 miles.
“End of life” in lithium batteries does not describe any limitation of the chemistry. It describes the limitation of our application. A battery with a capacity of 70% can’t take us on long trips, but it still drives shorter distances fine. And as Volt owners know, the vast majority of our trips are simply not that long.
If that Tesla battery kept degrading linearly over time[2], at around 2.25 million miles it would have the capacity of a Chevy Volt – still serviceable for 6 or 7 average rides before it needed recharging. That battery will last 15X longer than an internal combustion engine. That battery is, for this purpose, effectively free.
Of course, rather than start with a huge battery and drive it down to 20% (carrying nearly 1,000 pounds of dead weight in the process), it makes sense to simply build a vehicle around a smaller battery.
How small? Over 90% of trips are less than 20 miles. If fact, half of all trips are less than 5 miles.
Histogram of automobile trip length. Data from SolarJourneyUSA.com
Robots don’t get irritated by having to stop every few trips to refuel. Robots don’t get range anxiety. If the economics work out, why not make lots of cheap, tiny vehicles, instead of just a few big ones? Most of today’s trips could be made just fine with a self-driving golf cart. Prices for a street-legal cart are less than $10,000.
It’s not chemistry that limits vehicle batteries. It’s us.
7. Self-driving cars will ask us to think like capitalists
Most of us perceive that companies love robots because they remove the cost of labor from production – every wage not paid is money in the pockets of investors. Yet often the more important savings come from something entirely different: Reconfiguring the entire system to perform more efficiently than a human ever could.
Removing labor from the cost of driving only nets a 2X improvement. Yet when the person is gone, we can built a fleet. When we build a fleet, we can match the size of the battery to the trip that is needed. When we optimize size to rides, we realize that without a human driver, we deploy a one-person vehicle for most needs, at a fraction of the cost of a full sized sedan.
Taking out the people from the equation nets us not just the 2X savings from their labor, but allows us to reconsider our assumptions about what an automobile is. Without drivers, cars will cost 10X less. At least.
This is a stunning conclusion for anyone who drives for a living. The problem isn’t the wage. The problem is the person.
Yet because most of us buy our own vehicles today, we will each don the hat of the investor when we decide whether to switch to transportation on demand. And I predict that we will each unearth our inner capitalists, because the economics of the transformation to self-driving cars are stunning.
To illustrate, let’s do some middle school math: If a self driving car costs $0.18 per mile, and a personal vehicle $0.58 per mile, how much money will a typical driver (15,000 miles per year) save by switching?
The answer is $6,000. Per car. Per year.
Think about that number for a moment.
The median personal income for a full time worker in the US is $41,500. The person who chooses to abandon their existing vehicle in favor of a self-driving car will get a 15% raise. After taxes. And that savings holds for rich and poor alike.
This economics is what makes the case for self-driving cars so compelling. The strongest case against self-driving cars is this: Individually owned cars do more than just get us from point A to B. Cars provide us with self-expression, spiritual release, and a place to make out. And while self-driving cars excel at transport, they are likely to fail at these other tasks.
But if the opportunity cost is $6000 per year, we can find alternatives. Cars do many non-transportation tasks very well. But they are not unique.
And this is why car ownership will die for everyone but the most dedicated enthusiasts. This time, it’s individuals and families who are auditioning for the role of capitalists. And in that role, they will look at the idea of hiring a human – even if it’s themselves – and ask: Is this really necessary? Isn’t there a better way to invest my money?
When given the incentives of a business, we will respond like one. And the world will change in the blink of an eye.
[1] Battery lifetime is complicated stuff – the electrolyte can decompose, the cathode can fracture or dissolve, etc. But generally speaking, when you charge and discharge slowly enough, the degradation rate is roughly proportional to the number of electrons passed, which is in turn proportional to miles. It’s a simplistic assumption, but it turns out roughly correct for an application like this.
[2] Older battery chemistries such as lead acid did fail at 80% – if a lead acid battery at 80% capacity was subjected to a shock, such as a bump in the road, it could fail catastrophically. But modern lithium batteries don’t do this; they die slowly.