Various announcements and rumors suggest the major German automakers, including VW/Audi, Daimler and BMW might be planning a real alliance on robocars.

When I say a real alliance, it's cause almost all the announcements of "alliances" in this (and most) businesses are just PR noise -- there to get attention and to give the appearance of progress when there is not much. This suggests something else, in particular because it is getting advanced leaks.

Is this a good or bad move for the German auto industry. In the early days, the Germans took the lead, among automakers, in robocar development, but have gotten bogged down with teams that are too large and too ADAS oriented. Several companies announced teams of thousands of engineers working on the problem. Put that in contrast with Google, which had around 17 engineers on the project when they reached 100,000 miles of real world driving in 2010. Sure, the Google team was top people, but Silicon Valley people aren't that much smarter than Germans. It's the approach that matters.

BMW, Audi and Daimler are rivals to be sure, but in the new global market, the threats are Waymo, Tesla (on powertrain and car design more than on automation) and the Chinese.

At the same time, as I point out above, all of these companies are more than big enough to do the development on their own. Smaller teams can be better than large teams for the core engineering.

That's less true on the testing and QA. Large teams can help there, driving more miles, building more simulation scenarios, finding more bugs. It's also efficient to combine effort on hardware development, which is expensive. In this case the Tier One suppliers, particularly German ones like Conti or Bosch, play a role.

A true joint effort would actually mean some layoffs or reassignments for the core engineers on these teams -- which might cause startups to hire them.

These companies already teamed up to buy "Here" (Navteq) as a jointly owned supplier of maps. Mapping is another thing that's "big" in that you need to map large areas and keep them up to date. One of the big challenges the small mapping startups face comes from the fact that mapping is mission critical. While every car must still stay safe if their map goes out of date, you still want the most accurate and up to date map you can to avoid customer problems and slowdowns. Many large OEMs won't ever bet on a startup to provide anything critical. It's too risky. That's why they like dealing with companies like Here and TomTom which have established relationships and long history, even if the startups might have superior technology.

The rumours suggest MobilEye (Intel) may play a role in this plan. While no longer a startup (and very based in the ADAS world) MobilEye is one of the most nimble automotive suppliers and their management gets the problem.

The downside of working together is that competition can be good, and 3 or more teams working on things will come up with more innovative solutions than one team of the best. A joint team, with too many masters, may lack the independence needed to pursue a vision, especially one at odds with the OEM's traditional views.

I wonder if a better plan might be to encourage and fund lots of startups with the promise that the consortium will buy the best one or just give it the contracts for the combined group.

Does Apple slow down?

We don't get much news out of Apple, but the news on their "Project Titan" has been rocky, with many departures indicating they have had a few changes of plan. A recent report saying that the group laid off 200 people led some to wonder if Apple was killing the project. Not too likely since other reports have put the size of the team at closer to 5,000. This probably just means that they have shut down one avenue after deciding it's not the right one.

The best guess about Apple's plans is that they just want to make the computer systems, not design an "Apple Car." Leave the car making to the companies who have that down for now. That's the right plan for the early days.

Does Waymo speed up?

It was recently announced that Waymo and Magna will open a factory in Michigan where they will product their robocars. They won't build them there, but rather they will take the partially finished cars they get from contract manufacturers (like Chrysler Pacifica minivans and Jaguar i-Paces) and then they insert all the self-driving parts and other modifications and finish the car off.

For many years people have wondered "How can these non car companies make cars?" The answer is, they don't have to. Almost any car manufacturer is open to "contract manufacturing" where you come to them with a large order for some variation of one of their standard cars. The more you order, the cheaper it is to customize. So Chrysler makes minivans and inserts mounting brackets for sensors and other special features, boxes under the trunk to mount computers, empty wiring conduits and so on. It delivers them, unpainted to the customization plant, which adds in all those things and finishes the car.

You can pay other contractors to do the finishing, and in this case, Waymo is working with Magna to make a plant to do that. They may well get the chance to pick up one of the factories GM is shutting down for a song.

My long term scary prediction -- scary for big car OEMs -- is that the contract manufacturing of the main car bodies will take place in China. The Chinese are learning how to do car production, but they don't have brands or marketing outside China. This is a perfect answer for them.

What this all means, though, is that Waymo is serious about the 80,000 cars they have ordered and has a real calendar for deploying cars at a large enough volume to saturate a city with Robotaxi service. They will be to first to learn what happens when you do that.

In the future

When you forecast the future as I try to do, it's always nice when it pans out. Of course, they never all pan out but you want to be a bit better than average.

In 2008, I described how we might build valet parking lots using self-driving tow robots. Now, somebody is building that.

My first idea was that by this point, cars would be mostly drive-by-wire (and many are) and we could just do this with a software interface -- let the car give over control to a small robotic unit. But then I realized it could happen faster if the robot did the moving.