Is the Zoox plan for a custom car the path to robotaxi domination?


Recently, robocar start-up Zoox came a little out of stealth with this Bloomberg piece and you can also watch a video demonstrating their perception system at work.

I've known Zoox since before its start, when in 2013, founder Tim Kentley-Klay and I (along with Alex and Michael, his two earliest team members) had many discussions about his desire to start a new-from-the-ground-up self-driving car company. We agreed about a lot, but had strong differences on how radical a car should be from day 1. Since then, Tim, who has a design rather than robotics background, recruited Jesse Levinson (from Stanford's project) and got money from Steve Jurvetson of DFJ. Since then, they have raised over $800M and built a large and capable team.

There are a few key differences in the Zoox strategy compared to most others.

  1. They want to build a completely new vehicle designed from day one to be a robotaxi, symmetrical in design and more able to maneuver than traditional car designs.
  2. Their only interest is the robotaxi business -- selling rides in their new vehicles. They have little interest in selling cars
  3. Design will be central, even more than technology, in the success of the project.

While disagreeing on how quickly to do it, Tim and I both like to call the robocar, "Not a car, but the thing that comes after the car."

Is this a strategy that can take a startup to victory against the likes of Google/Waymo on one side, and GM/Cruise and other automakers on the other side?

Doing a car from scratch is mega-expensive

While there are many reasons to want to redesign the car of the future (or thing that comes after the car,) doing so is expensive and even worse, time consuming. Tooling up a new car design is a very long and costly process even for those who are experts at it. While cool concept cars and prototypes can be put together quickly by concept teams and scrappy garage startups, making something that can go into mass production is hard. Both from the production standpoint, but also from the standpoint of knowing it is safe and reliable and getting it certified that way (a process known as homologation). The reality of getting approval from lawyers, boards of directors, insurers and governments doesn't move at the pace of innovative design and technology.

This has led to the most common plan of big players: Make the self-driving part (software and hardware) work on top of existing car platforms, then move to slightly customized versions of those platforms, and eventually to a new vehicle design. For car makers, they know how to build both old and new cars, though they are hidebound with their old knowledge to a small degree. For tech companies, they don't know how to make cars, but they don't need to -- there are tons of automakers who will make a new car for you to spec as a contract manufacturer.

If you want to test a car on the road, bolting on to an existing car is by far the quickest plan, and in fact Zoox has done that for their on-road testing, while doing off-road development of their more radical prototypes.

The cost in money is not a big barrier to the carmakers or to companies like Google, Uber or Apple. It's the cost in time that is more troubling. Everybody wants to be first, to dominate the cities they land in. A new design simply can't have years of safety and manufacturing history behind it, and can't land first.

Many of the other players may see the virtues listed below of a custom design, but figure they can move to that approach later, rather than sooner. That might be right -- indeed 2nd movers have a surprising history of victory -- or it may be wrong because most people usually are way too late in deciding "the time to switch" because of their comfort cushion.

What do you get from a custom design?

Symmetrical design

Zoox has long-touted the virtue of a symmetrical design. In fact their early designs called for 4 way symmetry -- making a car from 4 identical (other than mirroring) quarters. The symmetrical design offers redundancy and thus a low failure rate. It can also offer some safety options, like using other motors as brakes if all the brake systems fail -- unlikely when you have 2 or 4. However, there are arguments that mechanical, non-safety reliability is not super important in a robotaxi where another unit can be dispatched in 2 minutes to continue a passenger on their journey -- ie. the redundancy can be in the fleet, not the car.

Zoox promotes a front and back symmetry which means the car never has to turn around, since it has no front or back. This will add a modest advantage picking people up in certain types of spaces and urban areas. I am less excited about this, because even cars that have a forward and reverse direction can drive backwards with full confidence if they are robocars, so they can turn around quickly, and always back into pickup spaces so that they can leave without a turn.

The symmetrical design promotes face to face seating, which is great for social cars. Some people are not comfortable facing backwards, so they will always want the seat in the direction of travel and will even want to shift if a car reverses direction.

Zoox's goal is the symmetrical design be simpler and easier to make -- you only have to design once for the 4 corners and build twice (a quarter in left and right mirror forms.)

Higher maneuverability

The ability to simply reverse direction is just one of Zoox's plans for making a car that maneuvers better. They are also adding 4 wheel steering, which allows all sorts of moves you can't do in a real car. Full rotation 4 wheel steering (which I don't think they plan, but the Nissan Pivo implemented) allows you to go fully sideways to get out of jams or into special parking and pick-up locations. Even partial 4-wheel steering will improve the ability to do that.

Zoox is betting that these small touches of more seamless pick-up and delivery will be a winner with customers.

Interior meant for riding (or sleeping.)

Almost every car out there is designed for driving, with driver features and comfort far ahead of passengers. The rare exceptions might be the cars with a focus on the limousine market, and perhaps the specialized taxi designs used in London and New York.

A robocar is only for riding, and so you're going to make it a lot nicer if you can forget entirely about the driver and her needs. Drivers don't just need their seat and console, they need the sightlines to see the road. They also need to be able to feel the road as they drive. One of Zoox's concept drawings (see above) showed a car with no front (or back) windshield. While that's not too likely because many people very much like to see forward either for touring or to avoid motion sickness, it shows what's possible if you can forget about the driver.

Face to face seating is also a very nice thing for the times you want it.

Zoox should be able to make a car that's nice to ride in. But how much nicer than, say, the seats of a minivan as Waymo has chosen for their vehicle?

Zoox also is free to support different vehicle purposes, such as a sleeper car with a bed, or office car with a desk. With the robotaxi, not all cars need be the same -- you can have special purpose cars that are sent to those who have that purpose.

Vehicle designed for maintenance cycles and life cycle of a taxi

The typical car lasts under 200,000 miles and 20 years. A New York taxi lasts 5 years and 250,000 miles. A robotaxi could be designed for different lifetimes. A 10 year taxi might last over 500,000 miles, but it would need to replace many components in that 10 years.

  1. The seats and many interior components will wear out and get dirty with that much sitting, you want them for easy replacement of the surfaces.
  2. Exterior panels may face rust or dings and want replacement at certain intervals. Also elements of the underbody.
  3. The computers and sensors will certainly go obsolete in 10 years -- you may want a car where those can be easily replaced during operation.
  4. Batteries will need replacement in electric vehicles.

A new vehicle can be engineered to have all parts last (except those that go obsolete due to changing technology) the life of the vehicle, or to deliberately last a fraction of it, and be monitored and replaced easily.

The vehicle can express a different aesthetic

Some people imagine they want to make robocars look as much like conventional cars as they can. I think customers, especially early customers, will revel in cars that look different, that advertise to those who see them in one that they are participating in the future. While it's easy to get this wrong, and look stupid, done right, a custom design could be the design of choice for many riders. Of course, it is possible to offer both new designs and classic designs to serve all kinds of customers -- in most rides, people don't care too much about what they are seen in.

Possible higher safety

There are some ways you might improve safety in a vehicle designed from the ground up. Safety can be equivalent in in both directions. 4-wheel steering might allow better accident avoidance. You could also design in safety functions for pedestrians and cyclists, such as external airbags. (Waymo put foam padding on the outside of their Firefly prototype.)

You can more easily create vehicles aimed at specific sub-markets

Design can be aimed at sub markets, such as the one-person city car, which can't even go on the highway. Or the social car or the sleeper car. Plus new markets which have yet to be developed.

You can throw out old ideas on how long it takes to create a car

Zoox isn't the only promoter of this. Local Motors is 3D printing its cars (small shuttle vehicles) and putting them together in a very short time at low cost. Many other small startups are doing the same. The positive regulatory attitude towards robocars means that it's become easier to get an exemption, in the USA, from the FMVSS and other regulatory hurdles may get reduced for up and comers as well as big players.

You can throw out other legacy ideas about a car

I've listed a few of the things in cars which may belong in the past. There will, of course, be others. Perhaps Zoox has or will think of them. They'll have an advantage on how quickly they can move to embrace them, while big automakers, and even high-tech companies like Waymo contracting through contract automakers, will be slower to do.


I will point out that I have seen a large number of other, smaller companies hoping to build a completely novel robocar. In fact, almost everybody who was experimenting with new designs (almost all in electric cars) recently declared they were planning for self-driving, since that's a likely future for any new vehicle. Many still have cars oriented around a driver, but a few don't. I wrote earlier about Local Motors which puts a focus on 3-D printing for fast design. They don't plan to do their own autonomy stack, though. Zoox is unique in that they are doing the whole stack, including a new car design, and they have raised enough money to get a serious start at that.

Is this a winning strategy?

These advantages are all good. But do they overcome the big disadvantage in cost and time? The 3D printers believe there are not disadvantages in speed and time, but otherwise it's a tough call. Only real tests with users will reveal of Zoox's and Local Motors' special features will be such a hit with customers that they strongly want to call such a vehicle instead of a Waymo, Uber or GM. Companies like GM, Daimler, BMW, Apple, Uber and Google have some of the world's most famous brands and reputations, and huge marketing resources as well. On the other hand, each unusual looking car advertises itself no matter where it goes, and word of mouth plays a powerful role.

Some analysts believe there will be no competing on safety. All cars will need to be "fully safe" (whatever that is) before going on the road, so no one service can claim they are much safer than the others. This leaves other competitive factors, outlined in this article and the factors which may decide who wins and who doesn't.

Some of these factors can result in a cheaper car, and thus cheaper service, in the long run. A company that does all service with standard 5 passenger sedans will lose on price to a company that offers cheap single person cars. (80% of all rides in a city are solo.) This can be true even in the early, more expensive days of the custom solo car.

In the above article I outline many different ways services will compete. It's a somewhat different question to ask how they will compete in the first few years. My general thought has been that they won't compete much at all in those years -- each will attempt to grab different territory, except perhaps in a few special places (like the SF Bay Area where many are based.) After all, when deciding where to invest, will you want to do it in a place you must compete, or an equally good place which is virgin territory? Nobody has the money to deploy everywhere in year one, and it would be foolish to do so.

So in those first few years, the only competitive edge will be simply existing. Each company's goal will be to convince the locals to ride in their robotaxis instead of Uber, taxis or their own private cars. A tall order to begin with. Zoox's customized car may make that a bit easier. The famous brands of the big players may also make it easier.

In the few places where multiple companies go head to head -- which will be the places so important that nobody can afford to cede it to others -- what will be the main competitive factor? Reputation? Price? Wait times? Service area? Or ride comfort and features? We don't know enough yet to say.


"One of Zoox's concept drawings showed a car with no front (or back) windshield. While that's not too likely because many people very much like to see forward either for touring or to avoid motion sickness, it shows what's possible if you can forget about the driver."

If you can see out the sides, wouldn't that be the same as riding on a train?

Not quite, because trains do very little turning and usually less bumping. Motion sickness is not usually a big problem on trains compared to cars. Of course, as I point out, an all new car might have a better suspension or banking ability to help with that.

But yes, many people can tolerate that just fine. But I think most people, given the ability to see forward, would prefer it. Certainly anybody interested in the surroundings (like a tourist) though less so people who plan to just stare at a phone for the whole trip.

If you can see out the sides, wouldn't that be the same as riding on a train?
I've ridden backwards on trains and in Hackney carriages and they both make me a little queasy.

(Sorry for the late comment; just catching up.) What I find puzzling about the Zoox concept is that they're reasonably investigating the implications of a radical change to driving (the driver not being essential) but yet their designs I've seen still seem focused on 4 passenger designs. If people don't need to own their own cars which must be suitable for all occasions then they won't need to constantly lug around the potential capacity to load up with 4 other occupants. What I liked most about Zoox was they realize that a car that doesn't park 95% of the time (i.e. always driving) will need to be a radically different animal than what we see on the streets now purely from a longevity standpoint. I feel this rightly supports their business model against established car companies since those big automakers really only know how to make cars that remain parked basically all the time. I'm a little surprised that, based on this, bus companies aren't more prominent in the autonomous space (static routes, constant operation, fleet managed, etc.). But I feel a real niche that needs to be carved out is 24/7 operation of 1 person vehicles. My guess is someone in India or China will do that first just from crazy traffic pressures.

Add new comment