Uber Autonomous – FOMO Syndrome pt. II

Looks like Uber’s autonomous unit is worth $0.

• Recent rumblings are confirming my long-held view that Uber’s autonomous driving effort is worthless which will incense the Japanese investors who invested $1bn just 18 months ago.
• Both The Information and Bloomberg have written articles over the last month highlighting severe problems at the unit including disagreement on which direction to take as well as a need to raise more money.
• This is yet another indication that the due diligence process at SoftBank was badly lacking when it went ahead and led a round (with Denso and Toyota) that invested $1bn at an insane valuation of $7.25bn in April 2019.
• I say insane because one small research outfit with a PC and an internet connection could quickly work out that there were issues with this asset and has written as such on numerous occasions (see here and here).
• The signs have been out there for a long time.
  • First, the data: The data that is released by the California DMV on vehicle testing is deeply flawed and full of caveats, but to date is provided a reasonable estimate of where the different offerings are in terms relative performance.
  • For the last few years, this data has indicated that Uber’s offering is the worst of all the contenders and to add insult to injury, has been getting worse with time rather than better.
  • In 2018, Uber’s vehicle suffered 3 disengagements per mile on average and little seems to have changed.
  • The Information is reporting on an email from one of its engineers highlighting the problem (see here).
  • According to this engineer, the car can barely drive half a mile before encountering a problem and struggles with simple routes and manoeuvres.
  • So, it seems that very little has changed in the last 2-3 years meaning that the Japanese investment has been completely squandered.
  • Second, intellectual property: Events seem to indicate that Uber’s intellectual property for its autonomous driving solution was largely based on Waymo for which a workaround has not been found.
  • As part of its 2018 settlement with Waymo, Uber had to submit its IP to an independent software expert who would determine whether Uber had removed the Waymo IP as it claimed to have done.
  • In its Q3 19 10-Q SEC filing, Uber states that the independent software expert has “made adverse findings as to certain functions in our autonomous vehicle software.” and that these findings are final.
  • In 2017, Uber made 1 disengagement per mile (see here) which then degraded to 3 per mile in 2018 (see here).
  • At the time I was of the opinion that this degradation was the result of removing the Waymo IP and hence I was very surprised that the independent expert concluded that Waymo’s IP was still there.
  • This meant that Uber had to take the axe again to its already battered code meaning that performance, in all likelihood, degraded further.
  • Uber stopped testing in California in 2019 and so there is no data to back up this assertion but from the articles I have referenced above, it looks like my conclusion was correct.
• The net result here is that it looks like the best solution for Uber would be the one I advocated over a year ago which was to throw the whole asset away and start again by acquiring one of the other smaller offerings that are currently struggling.
• Purchasing Zoox would have been a far better use of the money that Denso and Toyota put into this unit, but that opportunity has now passed.
• Hence, the list of quality assets is shorter than it was and there is a real risk that Uber ends up acquiring the last turkey in the shop.
• In the long-term, this is a very serious problem for Uber because when drivers are no longer needed, this combined with its platform will the key assets that define the value of its business.
• At this rate, the long-term for Uber looks very bleak indeed.