Soon after Marc Andreesen founded the venture capital firm Andreesen-Horowitz (a16z) ten years ago, he wrote a piece, Why Software Is Eating the World, that the WSJ published in 2011. I happened to hear the latest a16z podcast this morning, Software Has Eaten the World, where Marc explained in three sentences what he really meant:

• Any product or service that can become a software product will become a software product.
• Every company in the world that is in any of these markets where this is happening has to become a software company.
• In the long run, in any of these markets, the best software company will win.

Chuck Gray of Ford

I was reminded of this earlier this year at Automobil Elektronik Kongress in Ludwigsburg, when Chuck Gray, Director of Electrical and Electronics Systems Engineering at Ford presented Automotive OEM to Software Manufacturer—Follow the Computer Industry? Of course, the big challenge for an automotive OEM like Ford is that they are starting from a standing start with very limited software expertise, and even less of the type of culture of a software company. Automotive is about as far away from "move fast and break things" as you can get. The corollary of Marc's three statements is that cars are increasingly becoming all about software, and so every OEM has to become a software company, and the best software company will win (which might be one of the traditional OEMs, or Tesla, or someone new like Waymo).

Chuck opened talking about another really big transition in automotive:

Our founder obsoleted horses…that was a big deal, especially if you lived in a city.

Chuck had some great statistics about the scale of where we are today. In every hour in the US:

• 30 new vehicles are sold
• 9 million miles are traveled
• 125K Uber/Lyfts are on the road
• 60K shared rides are taken

So Ford wants to design smarter vehicles for a smarter world. Currently, Chuck thinks we are going down a bit of a blind alley. As he put it:

I picked the top 15 cars across the world. We all sell very similar models, so there is no real differentiation. How many parts are standard? It turns out that my engineers figured it was just the tire valve. And that's it. But now we have tire-pressure-monitoring-systems, so we don't even have the same tire valves anymore.

Historically, Chuck said, "the best collaboration happens at the tier 1s when the OEMs aren’t involved." They develop products to share among all their customers, building up a sort of Venn diagram of what is required among them.

That's not really going to work for software and autonomous driving, he feels. It is simply unaffordable for everyone to develop everything without standardization. Already, Chuck estimates that the accumulated investment of 15 OEMs "inventing" their own automotive compute center to exceeds $7.5B.

The PC Industry

What can we learn from other industries? We studied the IBM PC.

The lesson that Chuck takes away from the PC is that the standardized architecture that IBM created gave birth to the modern PC industry. This is clearly true. But I think it is a misleading message. The people who most benefited from the standardized architecture were Intel and Microsoft. IBM lost control early on, and exited the market, finally selling its laptop division to Lenovo about 15 years ago. Manufacturers like Toshiba and Dell had tiny margins on PCs. And they were the ones that survived: remember Gateway (think cows), and others whose names I forget, that left the market. Even Sony, with one of the most well-regarded brand names in the world at the time, couldn't make good money with their Vaio PCs since people just wouldn't pay the Sony premium, so they exited the market, too.

The mobile handset manufacturers took one look at this and decided that they were not going to let that happen to them, and avoided standardizing the processors and operating systems, largely developing their own chips and software for the first decade. Of course, the operating system eventually standardized on Android (except Apple) but since it was free that didn't affect their margins negatively. Google might take money from the users of smartphones indirectly through search, but they take none from the smartphone manufacturers. At the low end of the smartphone market, there is standardization to Qualcomm and Mediatek chipsets, but it is notable that the market leaders in the smartphone market all continue to design their own application processors.

I think that car manufacturers would be foolish to standardize on a single solution for autonomous driving computers and software. This would be equivalent to every automotive OEM buying one of a few standard engines from the standard engine manufacturing company. There's a reason that every single OEM builds their own engines. The accumulated investment of the OEMs in "inventing" their own engines must be orders of magnitude more than they have spent on electronic and software development to date, but they all consider it necessary.

Standardization in the Automotive Industry

If car manufacturers standardize on a single solution then I think two things will happen:

• First, cars will be commodified like PCs, with every car having the same driving characteristics, just like every PC of a given generation had pretty much the same characteristics: same processor, same operating system, a lot of the same software, just a different case and logo. As if every car had the same engine. That might not be bad for us drivers, but it would be terrible for most automobile manufacturers.
• Second, most of the profit will go to the winner of the single solution proving the hardware and software. And it probably would be just one company in this case, since hardware and software seem much more intimately related than in the PC era. Once it is standard, car manufacturers either pay the tax to get the standard solution, or they are not in the automotive market. Before the IBM PC, the market for personal computers (in the generic sense) was fragmented with computers like the Apple II, TRS-80, Kaypro, Commodore, and more. They all either exited the market, or had to build PCs with Intel processors and Microsoft operating systems. (Most famously, Osborne Computer made the mistake of announcing to the world that they realized this, and promptly went bankrupt when everyone stopped buying their current products while they waited for Osborne to ship their PC. This is so famous that it has a name: the Osborne Effect.)

To be fair, the message Chuck takes away from the PC is not that everything should be completely standard. He feels there are three areas where the whole industry can cooperate: standards for sensors and actuators, standard reference architecture, non-competitive shared codebase broker. I guess that would be equivalent to something like the PC industry standardizing WiFi chips and disk controllers, but building their own microprocessors.

The audience was a bit skeptical in the Q&A. Someone pointed out that there is some cooperation among German carmakers but "does GM even talk to Ford in the US?" Chuck said that Ford has cooperated with GM, for example in the transmission world for more than ten years. "But now is a very good time to consider those sorts of collaborations and you'll see more going forward...in the US even!"

Another question was how to decide whether to build, to buy, or to partner. Chuck says that is driven by speed to market. "Just generating a proposal and a quote was a key barrier...so we started with infotainment to be engineered in-house. That led to us hunting down administrative challenges to we can get from early concept to running hardware in less than six months."  Hmm.  When your company is working on an agile RFQ process, it is a long way from having an agile software development process.

My Conclusion

I continue to believe that advanced automotive electronics will develop analogously to smartphone electronics. At the high end, companies will need to build their own electronics, which means SoCs. I said that this would happen and I was pleased when I was vindicated when Tesla announced their SDC earlier this year. At the low end, where companies compete more on price than premium capability, I'm sure we will see more standardization. But if Ford (and other OEMs) want to continue to have premium, differentiated vehicles, then there is no way around the fact that they will need to develop a lot of their own software.

So my advice is not to follow the computer industry, follow the mobile industry.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.