Uber ATG And Aurora Merge To Staggering $10B Valuation


Uber's self-driving unit (ATG) has been merged/sold to Aurora, a high flying startup with a combined valuation of $10B, but a big drop for Uber and climb for Aurora. I outline the odd nature of the deal in this new article as the robocar news just keeps on coming. It's not winter any more.

See a new Forbes site article at Uber ATG And Aurora Merge To Staggering $10B Valuation


I think you'll find while Uber is the Western worlds largest taxi company, it is in fact dwarfed in size on a global scale by China's DiDi under every metric, it's 5 X bigger

"Uber retains the top position in the world in selling rides"

That's simply not correct, DiDi has 550 million passengers and does 10 billion trips a year,
Uber has 110 million passengers and does less than 7 billion trips a year.

Though Didi is essentially in one country and its brand is not known outside, while Uber is the most recognized name in most of the world.

Hard to see how Uber will be alive in 10 years. Their entire business model is based on exploiting human drivers, and human drivers are going away.

Maybe they'll get bought out. An Uber/Tesla merger would probably make sense. Subsidize the cost of Teslas for Uber drivers in exchange for collecting lots of data. Sell the service as Uber Green (a premium service using electric cars only, charged exclusively with clean power).

Problem is, Uber's stock is way way overvalued. If not a merger, maybe they can make a deal now that they're not competitors any more.

I presume some sort of relationship comes from this deal, such as:

  • Aurora will only licence its driver to Uber for robotaxi service
  • Car companies can integrate Aurora driver, but any car made that goes into robotaxi service has to go into Uber, though possibly not exclusively into Uber.

Or maybe Uber got nothing of this sort, which makes it harder for them.

Is there any evidence that Aurora is ever going to matter at all?

If they thought merging with Uber ATG was a good idea, probably not.

If Aurora produces the best self-driving stack that is not tied up in a proprietary company -- then they can matter a lot. The stacks from Zoox and Waymo will not be for sale, nor will Tesla's if it ends up working. The one from Cruise may be fore sale, or may only be for Cruise investors like GM and Honda.

Whoever has the best stack that is for sale will be able to do well. Aurora got this valuation I think because the founding team gave the message, "we built those other stacks you can't get." Waymo is the best, and Chris played a big part in that. Car companies want what Waymo has, but if they can't get it, they will buy it from Chris.

This deal I am sure will allow Aurora to sell their stack to car companies. It may come with a requirement that any such car that goes into a taxi network must go into Uber, or possibly only Uber.

Whoever has the best stack that is for sale will be able to do well.

Maybe if they also have the best patent lawyers. This stuff is going to be so easy to steal. The keys to software 2.0 are concepts and data, not code, and neither one of those is protected by copyright law.

Concepts are not very protected. Training data, where 500 people at Tesla have been busy labeling images to improve their systems, can be protected, though it might be hard to detect if somebody stole it and used it through a leak.

But do you think a major team, one of those seeking to be the supplier for major automakers, would make use of a pilfered training set? It's not obvious in what you release, but you take a big risk of word getting out.

As is clear from recent Tesla hacks, FSD is quite far from "Software 2.0" at present. Yes, it uses lots of neural nets (as does Waymo) but nobody is about to bet their life on neural networks entirely.

For now, while it's not impossible, it's pretty unlikely in the view of most that you build a system you can deploy on the roads with just software 2.0. But some people do think that.

You talked about Tesla, but I was talking about Aurora. I just don't see why they matter. But I'm sure you know more about them than I do.

Of course you don't build a system with only software 2.0 (just like you don't build a system with only software 1.0). The software 2.0 is the (software) part of the system that is valuable, though. And over time, the software 1.0 becomes smaller and smaller.

Moreover, software 2.0 (plus real world experience) is the way you figure out what software 1.0 you need to build.

Add new comment