The SD Cars Gold Rush Winner
by Jean-Louis Gassée
After Personal Computers and Smartphones, both richly deserving the “epoch-making” epithet, we now have a new tech Gold Rush: Self-driving (SD) cars. Is Alphabet/Google/Waymo in a position to be the Microsoft of this new era?
The evidence is everywhere. Not a week passes without an SD demonstration on the open-road, or a state granting SD test permits, or the announcement of a new SD technology alliance. Car companies, subsystems, sensor and processor suppliers such as Nvidia… The list of wannabes in the self-driving car arena is too long for this space.
Some of the news bursts are lamentably silly and misleading: Self-driving Ford cars deliver Domino’s pizzas! No, they don’t [MN readers know the routine, edits and emphasis mine]:
“Each car will be driven by a Ford safety engineer, with other researchers onboard, who will zero in on the last 50 feet of the customer experience.”
It’s PR with a whiff of desperation. Ford doesn’t want to be perceived as lagging in the Self Driving Gold Rush. No one does.
The excitement, the haste, the need to appear to be leading the pack is understandable. For millennia, transportation systems have had a profound impact on politics, economics and culture. From the Silk Road to Roman vias, waterways, railways, the “horseless carriage”, and air travel… Self-driving cars will be no exception.
As Benedict Evans surveys in his highly-recommended Cars and second order consequences, SD cars will change everything, from car design, urban planning, and insurance (I recall Warren Buffet saying SD cars couldn’t possibly be good for his huge insurance business), to shopping, drinking and dining, energy supply and consumption, and even tobacco sales… I would add mating rituals and work modes to the impossibly long list of activities, people, places, and habits that will be impacted by the SD car revolution.
That’s why Intel bought Mobileye in what the NYT stenographer calls a “$15.3 Billion Bid to Lead Self-Driving Car Market”. Inexplicably, the article fails to mention Mobileye’s not-so-friendly breakup with Tesla, and, more important, it brushes past the fact that this is Intel’s nth attempt to find its Next Big Thing, a successor to its slowing x86 business. For cultural reasons, the Mobileye acquisition is as doomed as Intel’s previous efforts in security software, IoT, and wearables.
Tesla was one of the earliest and most visible players in today’s SD car theatre. Having decided, early on, that SD would be one of the company’s distinguishing features, Tesla began collecting driving data from its electric cars. Last year, Elon Musk boasted of having collected 1.3 billion miles of driving data.
That’s a lot of data collected very quickly — in the words of the Bloomberg article’s breathless subtitle: “Silicon Valley and Detroit can’t keep up with Elon Musk’s trove of real-world metrics” — but what exactly do these “real-word metrics” measure? The data aren’t offered for inspection.
About the same time as the 1.3 billion mile declaration, Musk tweeted this:
Given these impressive if somewhat inscrutable numbers, why did Autopilot’s Director, Sterling Anderson, resign from Tesla last December? According to an August 24th Wall Street Journal article, Musk was touting SD features that the engineering exec felt couldn’t be implemented on the company’s vehicles, even with a hardware upgrade.
Six months later, Anderson’s successor also resigned, preceded and followed by other departures in the engineering ranks. Notably, Chris Lattner, inventor of the Swift programming language at Apple, resigned his position as head of Autopilot software…after just six months (he has since moved on to Google). This doesn’t bode well for the current state of SD development at Tesla.
I will only briefly mention Uber’s tangled “relationship” with Google/Alphabet/Waymo. That hairball slowly makes its way to the courts. And there is a new sheriff in town, Dara Khosrowshahi, who might change the direction of Uber’s doomed SD efforts.
So who will be the big winners in the Self-Driving Gold Rush? To help refine the question, let’s look at the “The Six Stages of Automation” (ibid. WSJ):
Only one company, Google, made what many see as the right decision: Go straight for the Holy Grail stage known as Level 5: Full Automation.
(Why not take baby steps towards full automation? The reasoning, attributed to Larry Page, is simple: The more a car is partially automated, the less vigilant the driver becomes. When an exception arises, when the partial automation needs help, the driver might not come to the machine’s rescue fast and well enough. Since a partially automated car will only “reach out” under extraordinary circumstances, the resulting damage to the car, the driver, and — least of all, but it’s still a consideration — the reputation of the car maker will be concomitantly extraordinary.)
Thus was born Waymo, Google’s self-driving technology spun out as a separate company last December.
Waymo’s eminently serious Carcraft simulator and secretive “Castle” testing ground in California’s Central Valley, as reported by The Atlantic Monthly, are at the far end of the PR spectrum compared to Ford’s pizzapalooza. And pace Elon Musk, no one is close to Waymo’s accumulated experience. Valley dwellers may recall the little pods around Palo Alto, and the station wagons brisling with cameras, antennae, and other sensors. Then there’s Maps data, Google’s immense computing and engineering resources, its leadership in the AI/Machine Learning field — and Alphabet’s deep pockets. Today, no one seems better positioned than the Mountain View conglomerate.
When the time comes to unleash driverless cars, this could put car makers in a difficult position. They can pay Waymo for the right to use their Level 5 technology, or they can use their own or someone else’s SD technology and expose themselves to selling an inferior, unsafe product.
Waymo/Alphabet becomes the Microsoft of the SD car age, licensing technology without having to soil their hands (or their balance sheet) by getting into the low-margin, capital-intensive car-making business. At a putative $1,000 per car and with today’s “installed base” of 1.1B cars mentioned in Benedict Evan’s essay, this starts looking like real money (albeit over a long replacement cycle). Today’s Alphabet’s yearly revenue is about $100B.
____________________________
Next week or the week after, I’ll start with one of my favorite Horace Dediu quotes and explain the many flaws in the Alphabet-as-the-Microsoft-of-SD scenario.
