Toyota, Ericsson, Intel, NTT, and other companies have formed a consortium to figure out ways to manage the massive explosion of data that will be generated by cars over the coming years. As cars become more autonomous, they will need to gather enormously more data from cameras, radar, LIDAR, and other sensors and transmit at least a subset of that data over networks to central repositories for processing and analysis. That, in turn, is going to require some big decisions about which data to process locally, what needs to be sent over the air, and how much and which data to store on an ongoing basis in both locations. Since carmakers like Toyota don’t really have much experience with that kind of thing, network infrastructure vendor Ericsson and chip vendor Intel among others are going to work together with them to figure some of this stuff out, and have left the door open for others to join their effort in future. Notably absent from this initiative are other big automotive chip vendors like Nvidia, any cloud service companies beyond Japan’s NTT, or mapping companies like HERE, and given the strong roles they’re playing or likely to play in this area, the consortium does need to add additional members (including ones who compete with the founding members) if it’s to make real headway here.
The Wall Street Journal reports that Uber is planning to shut down its US car leasing business, which was apparently losing $9,000 per car instead of the $500 Uber projected it would lose when creating the program. It sounds like Uber might have around $800 million in cars leased through the program, which Uber apparently holds titles to in trust rather than on its books, and it may have to sell many of them and the associated leases to get out of the business. The program and this outcome are indicative of Uber’s enormously aggressive expansion strategy and the huge sums it’s sometimes incurred in poorly thought out initiatives which have ended up significantly worsening its losses. Though it’s most common to see Uber’s losses attributed to its subsidies in ride sharing itself, a good chunk of its losses are made in these other aspects of its business and could be cut back significantly as it focuses on more rapid progress towards profitability. I suspect cutting the leasing program in particular wouldn’t dent growth much but would certainly go a long way towards improving margins. It’s also likely another example of an area where Uber might well do better to partner with a small number of large, reputable firms rather than taking such a direct role in the operation – in general, Uber seems far less willing to partner than Lyft, which is arguably holding it back in some areas.
It’s been somewhat heartening lately to see Lyft partnering with a number of different players around autonomous driving technology, and even announcing its own “open platform” (mostly an API for integrating with its ride sharing data) for the space. At a time when Uber, Waymo, Tesla, and a raft of big legacy automakers are all competing around self-driving systems, it appeared Lyft was going to avoid diving into the fray and instead try to partner with the best in the industry. However, all that got rather turned on its head today when it announced that it will also be developing its own self-driving technology, with 10% of its engineers already devoted to this area and a big expansion and new offices planned for the team. On paper, that looks like something of a contradiction given the recent announcements about partnerships, or at the very least a serious hedge that will make at least some of those partners think twice. However, in reading both Lyft’s own blog post and several press articles about the news, I’m coming to the conclusion that there’s actually very little there, and in fact Lyft may be building something far more limited in scope than what most others mean when they talk about a self-driving system. All that’s really described in Lyft’s announcement is leveraging its existing ride sharing data and possibly adding sensors to some of its drivers cars to create 3D mapping. The former is already the centerpiece of its open platform, while the latter seems overly ambitious and probably also redundant given the much larger and more advanced high definition mapping efforts underway for the last several years. What I don’t see any references to are developing LIDAR or other hardware necessary for self-driving, or even software to steer self-driving cars – it’s almost as if Lyft expects its partners to fill those roles, though it still talks about Lyft’s own self-driving cars as distinct from those run by partners. I’m hoping we’ll get more clarity as this project moves forward, but suspect it’s less momentous and therefore also less contradictory than it might at first seen based on the headlines.
Tesla On Track to Deliver First Model 3s by End of July, But Q2 Production Falls Short (Jul 3, 2017)
Update: following another release from Tesla on Monday, I’ve amended both the headline and content on this piece significantly from the first version published Monday morning.
Overnight, Tesla CEO Elon Musk had tweeted that Model 3 production would begin shortly, with the first deliveries happening by the end of July, with production ramping up slowly from there. Hitting the launch milestone is something of an achievement for a company that’s often missed its own self-imposed deadlines, but the real test is ramping production enormously above past levels, and that continues to be the achievement I’m far more skeptical of. The new numbers provided today suggest far lower total production than Tesla has promised in the past, at least in the second half of this year and first half of 2018. Later on Monday, the reason for getting that news out overnight became a little clearer, as Tesla released its production and delivery numbers for the June quarter, including a shortfall in both due to battery shortages. That’s bad news for Tesla, and more evidence of its inability to plan and execute on production in predictable ways, and therefore to meet the targets it sets itself. None of this gives me any more confidence in the longer term projections of Model 3 production.
BlackBerry’s QNX division already makes the operating systems that power many cars today, and it’s just announced a new version of its OS for cars titled QNX Hypervisor 2.0. The key selling point of the new version is that it better partitions the safety-critical and non-safety-critical elements of the OS and the services they support in order to both prevent localized glitches from crippling the whole car and also insulate safety-critical functions from hacks that penetrate, say, the infotainment system. The implication of both of those, of course, is that there could be glitches or hacks that would penetrate certain systems, which seems a realistic if not a heartening concession. But as Android and other operating systems make their way into cars, being able to separate functions relating to driving tasks from those that merely deal with infotainment and other elements of the in-car experience is going to be increasingly important, and BlackBerry/QNX is emphasizing that element. I also wonder if it means QNX will be more able to operate as part of a hybrid operating system environment within cars, where infotainment features might be powered by the new version of Android for center consoles while driving features are still powered by QNX.
One of the best recent examples of the fragmentation that still exists within Alphabet and even within Google specifically is the fact that Waze, the navigation app acquired by Google a couple of years ago, has been working on what’s effectively a ride sharing service, and that it’s been doing so entirely independently of any other part of Alphabet or Google that’s working on related services and technology. It grew entirely out of Waze engineers’ desire to do something interesting rather than out of any strategic imperative from Alphabet management, which means that Alphabet’s Waymo has launched a test of a self-driving ride sharing service while Waze is expanding its Carpool service and the two have nothing to do with each other. To focus on Waze for a minute, it had previously launched its Carpool service in the Bay Area, and now is expanding it to the rest of California. But it’s still more of a true ride sharing service than most of the other services that get painted with that label – this is intended purely as a way for people to literally share rides to places one of them is already driving to, and to help split the driver’s gas money. As such, it also hasn’t generated revenue for Waze, which has merely passed along the entire IRS mileage rate to the driver, so it needs to find some other way to make money, and it looks like that might at least in part be showing ads to users of its app. It’s ironic, then, that even though the interesting disruptive transportation technology has no connection to the rest of Google or Alphabet, but its business model might end up borrowing quite a bit from its parent.