Cars are going to change a lot in the next few decades. Electricity on one hand and software on the other change what a car is, how it gets made and who might own one. They might also change the key players. As is often the case when an industry is going to be turned upside-down, there are actually a number of separate things happening, which feed into each other and accelerate the pace of change.

Electricity changes what it takes to build cars

First, the shift to electric reduces the mechanical complexity of cars a great deal. No transmission or internal combustion engine means far fewer moving parts. That may also change the sophistication and capital required to design and build cars, which, in turn, may change who can build them and how they get built. Gear boxes and premium sports transmissions turn into software in the same way that electromechanical calculating machines or cameras got turned into software.

The key change in the phone business in the last 15 years is that you used to need deep technical understanding of cellular technology to make phones, and now you don't - you can buy commodity components off-the-shelf. This happened to the PC industry too: the technology changed from a core enabler to a module. There's obviously a lot of variation within this: Apple and Samsung create very sophisticated custom systems whereas cheap Chinese manufacturers just buy a system-on-a-chip from Mediatek or its competitors, but IP is no longer a condition of entry (as Apple itself showed). In parallel, the industry moved almost entirely from in-house factories to outsourced contract manufacturers. These two changes together mean that entering the phone business now needs much less IP and much less capital.

The car business is some of the way there - there are contract manufacturers and there are modular systems and components (especially within groups, of course). But the shift to electricity might take us much further down that road. One of the important dynamics of smartphones today is that there are companies effectively outsourcing scale to Shenzhen and adding their own software, brand, marketing and distribution instead, with much lower capital requirements - Xiaomi most obviously. This may not be directly applicable to the manufacture of a large piece of metal instead of a small piece of glass and plastic. But it might be.

On-demand changes car ownership, and also changes which cars get bought

Second, the rise of on-demand car services changes what it means to own a car and changes who buys them, and that in turn may also change what they look like. These models won't work for everyone everywhere: there will be a point of equilibrium in each urban area where supply, demand and price stabilize at a sustainable level (after the price wars and sign-up bonuses are gone), and that point of equilibrium will look different in different places. The number of people who stop having a car (or using one) or who stop using public transport will vary, and won't be universal, but will be substantial.

But to the extent that this does happen, who owns those cars and what do they look like? It may make more sense for the cars themselves to be owned by someone with a big balance sheet - a GE Capital, if you like - that owns hundreds or thousands of cars with an optimised financial structure, rather than individual drivers getting their own leases. That in turn means that the cars get bought the way Hertz buys cars, or - critically - the way corporate PCs get bought. In this world what matters is ROI and a check-list of features, not flair, design, innovation or fit and finish. The US car-rental companies account for around 15% of the US industry's output, and some models are specifically designed with this market in mind. They're not the cool ones. That poses a challenge for Apple, and indeed Tesla. If the users are not the buyers, the retracting door handles or diamond-cut chamfers don't matter.

Autonomous self-driving cars

Third, autonomous self-driving cars, presuming for the sake of argument that they happen, change vastly more than just traffic accidents. They have the potential greatly to expand the adoption of on-demand, and so to transform who buys cars and why.

Removing the drivers from an on-demand car service cuts the cost, since you don't have to pay them and also since lower accident rates mean cheaper insurance (though this applies to your own car too). But in addition, autonomous cars expand supply for on-demand services, since many more cars are available to be used for on-demand when their owners aren't using them. This will creates all sorts of second-order effects and feedback loops. We saw the same thing with cars themselves: as Carl Sagan said, it was easy to predict mass car-ownership but hard to predict Wal-Mart.

Hence, if your car doesn't need to wait for you where you got out, then city-centre car parks disappear and retail gets remade (such of it as survives the shift to ecommerce, of course). No more worrying about parking. If you don't need to worry about parking yet can be driven there directly and affordably, how much travel shifts from public transport to cars? How many people visit a busy central area they might previously have avoided for that reason (the West End of London, for example)? But then, where does that car go afterwards - does it drop you off for dinner and drive off to a cheap carpark, or does it spend the next few hours driving other people around for a fee? The more autonomous cars there are, the more appealing on-demand becomes. Quite where the second-order effects end up is hard to predict - for example, where does it leave public transport if routes start emptying out, and what does that mean for people on very low incomes? What does it do to cycling?

@BenedictEvans will be interesting to see how it affects cycling, which becomes perfectly safe.
— Aidan (@aidano) August 21, 2015

From a technology point of view, what's really happening is that we move road transport from circuit-switching (with manual switchboard operators) to packet-switching. The cynical counter-argument would be that in a world of self-driving on-demand cars no-one will ever have to worry about parking, but that's just as well because we'll need the space for traffic. But the traffic will work better too, so there may actually be less of it. This video of car management is really showing packetized traffic-management, but arguably this is still TDMA rather than true CDMA - why do you need lanes at all?

This video gives a good sense of what that might look like in the real world.

This is how self-driving cars would work at an intersection. pic.twitter.com/5z29Zeibk9
— oliver binkhorst (@obinkhorst) July 23, 2015

From a technology perspective, I think there are three parts to what's happening here:

A single autonomous car driving down the street and not hitting anything - this is what most of the attention is on now.
Optimizing traffic flows when some or all cars on any given road are autonomous (two very different problems) - this is what we really see in the videos above.
Optimizing a fleet of autonomous on-demand cars in a city on a real-time basis. Where are the cars now and where do you want them to be in 17 and 34 minutes to minimize traffic, trip times and wait times?

Of these, the second and third feel much more like Google or Uber challenges than Apple or (perhaps) Tesla challenges. These are questions of algorithms and large scale computing systems, not design, experience or ease of use. They are also why maps have become so important - maps are PageRank for the real world. For car companies maps used to be an accessory no different in strategic terms to the CD player, but now they're a site of existential worry.

A further issue for Apple and (perhaps) Tesla is that, again, this changes what cars look like and who buys them. On one hand, removing steering wheels and other manual controls in an autonomous car is a further reduction in mechanical complexity beyond moving to electric, and also an occasion for re-imagining what a car should be, which is certainly appealing for Apple. (I suspect that the first autonomous cars will still have manuals controls, just as the first steam ships still had masts and sails just in case - the original hybrid technology.) But, again, if all those autonomous cars are also on-demand cars, then they are not bought by you and you don't summon them based on the design. If you're calling a car owned by another ordinary person who's out to dinner then they might have bought an Apple or Tesla for the design, but if you summon one from a corporate fleet then it will be bought in the same way corporate PCs are bought.

Scale

Fourth, scale. As I've written elsewhere, the smartphone has supplanted the PC as the dominant ecosystem, in volume of devices, install base and usage. But phones were always bigger than PCs - it was just a matter of time before they converted to software. What other manufacturered product is susceptible to being taken over by tech in the same way and can provide the same size of opportunity? Cars look like the only candidate.

The problem with this chart of course is that this isn't static. It's not clear quite how on-demand and autonomy will change car ownership in the next few decades, but it can only push demand down. Fewer cars are going to be being sold - some analysts are talking about unit sales halving over time (with growing demand from China and other newer markets offsetting new technology). Meanwhile, moving to electric can reduce the price of a car, or of course (Apple's preferred option) expand margins.

Then within that, what is addressable by Apple? It seems unlikely that Apple will go after the $10,000 market, reserving that price band for watches. Rather, it's the luxury or premium segment that fits Apple best - the bubble on the chart above shows Mercedes-Benz, BMW, Audi and Lexus, which combined sell 5-6m cars a year for $220bn in revenues (and so averaging $40,000 per car). That's where Tesla is aiming now, and where one might expect autonomous cars to arrive first. For comparison, iPhone revenue in the last 12 months was $146.5bn.

Do people buying these cars want one that drives itself, or prefer the 'driving experience'? Probably both. But one thing that doesn't seem likely is that Apple can expand the market for such premium cars in the way it expanded the market for premium phones. You can choose to spend $600 instead of $200 on a phone (especially if that premium is masked by a monthly contract) but you can't choose to spend $40,000 instead of $14,000 on a car no matter how good it is.

To look at that another way, if Apple created a car business as big as BMW and Mercedes combined, that business would generate less profit than the iPhone.

Featurephones

I bought a second-hand car when I moved to San Francisco last year. It's a 2009 model and it reminds me very much of using a Nokia in 2002 - a perfect feature phone before they started adding smart to it, badly. There's a point for many such devices (cameras might be another example) where the UI is perfectly optimised, and a trend after that for new features to grow like ivy, each one making perfect sense by itself, until you are swamped by multifunction controls and can no longer work out what anything does. Then software arrives and sweeps everything away. This was the death of Nokia, and arguably the Japanese consumer electronics industry, and probably many other products, and it's where cars have arrived now.

That is, it's now pretty easy to look at a car and say 'this should be a smartphone' - somehow. It also seems likely that the right way to do that is with software companies, not engineering companies, and that it should be driven by the software-powered device that you replace every two years and not the car that you replace every ten years. To the extent that you add smart to a car, it should really be from the smartphone, with the car dashboard itself being 'dumb glass' just like a connected TV. That's not limited to the navigation or entertainment either - the valuable place to add smart is to the driving controls and displays. That might be Apple's CarPlay or Google's Android Auto, or it might be this universal standard for adding smart to a car.

In the long term, though, Apple's CarPlay or Google's Auto are both bridging technology. The important software in an Uber car is in the driver's iPhone, and the important software in an autonomous car is never seen by the user. They'll sit with their back to the road and reminisce about traffic on Twitter. That's another second-order effect - if no-one drives themselves, what does that do to mobile use of the internet, or radio?