The SDE Circle of Life Today

This is the second part of a two-part post about the SDE Circle of Life. The first part was on Friday.

I like to say that the world changed on January 9th, 2007—at the moment when Hang (Hannah) Zhang, a barista at a Starbucks in San Francisco, received a phone call. It was a strange call, since whoever was on the other end of the line ordered 2000 lattes to go, before hanging up on her. It was ground-breaking for the 2000 people watching the other end of the call at nearby Moscone Center. It was the first real call (apart from internal testing) made on an iPhone. Of course, it was Steve Jobs ordering the coffees, calling live from the stage during his famous keynote. "It's just a handset announcement," said the CEO of Nokia. later that day.  But it wasn't. It was the start of the true smartphone era, when phones stopped being mainly about making phone calls and became what they are today. Amazingly, it was only a few weeks over 10 years ago.

Apple Leads the Way

Just as companies like VLSI with their PC chipsets had made it possible for anyone to design a PC, the standard products from semiconductor manufacturers—along with Google's free Android operating system—meant that anyone could design a smartphone. Lots of people did, too; especially the Chinese and the Koreans. Apple's first iPhone contained a chipset from Infineon. Apple couldn't have designed that chip themselves. They just didn't really have an internal SoC design team. But they kicked off the next phase of the SDE circle of life when they built one up. They didn't just put ads on LinkedIn. They acquired a couple of the most respected CPU design companies around, shut down their product lines, and put them to work on what (at the time) was a closely guarded secret. What Apple had realized was that it would be difficult to differentiate from their competitors if they had the same silicon as everyone else, and still give a great user experience (and Apple is all about giving great UX). Eventually, they would announce the Ax series of application processors for iPhone (and later iPad).

This was the current phase of the SDE circle of life. System Design Enablement requires high-end system companies to design everything from software down to physical designs, and in some cases their own standard cell libraries, and in some cases even push the foundries around to get the processes and packaging technology they want. The #1, #2 and #3 smartphone manufacturers are Samsung, Apple, and Huawei. They all design their own application processors, at least for the high end of their offerings (Huawei through its HiSilicon subsidiary). Xiaomi just joined that club. Below that, companies use chips from Qualcomm or Mediatek. They are largely focused on Asia. The #4 and #5 companies, for example, are Oppo and Vivo (I bet you've not got one of those in your pocket), but you can be #4 in the global market only selling in China and India. Remember, China Mobile has more subscribers than the US has people.

Automotive

Unless you've checked out for the last couple of years, you know self-driving cars are coming faster than predicted, and that it is the next Big Thing for semiconductors. Compared to smartphones, it is a small market by volume since "only" about 100M vehicles are built per year. But each will need a lot of silicon. The "system" companies fall into one of three categories: 1. So-called OEMs (such as Ford, Toyota or BWM); 2. Tier-1s (such as Delphi and Bosch), or 3. Upstarts (let's put Tesla, Google, and Uber here, and add a lot of hopefuls that you've not heard of). They all face the SDE circle of life challenge. Do they buy merchant silicon, a lot of merchant software, and try and put a user interface on it that keeps/creates their brand identity?  Or do they build their own silicon so that they can deliver a unique driving experience?

My prediction: the leaders will design their own silicon. They have to have differentiation. As with smartphones, new companies that you've not heard of will spring up. With silicon from companies like NVIDIA, Qualcomm/NXP or Intel, and software from a host of suppliers, and electric drive train technology, which is mostly software, "anyone" can build a car. It will be like earlier when "anyone" could build a smartphone (or a PC—remember the explosion of PC companies using chipsets and Microsoft software?). The semiconductor companies won't be able to sell their chips without a large part of the software load already running, so they also have to get enough system knowledge to do that. One way or another, the companies are becoming alike: automotive companies need semiconductor expertise, and the semiconductor companies need automotive expertise. You can't design the software and the car and the semiconductor completely independently of each other.

The big question is whether GM and Honda are the Motorolas and Ericssons of automotive: masters of the old but increasingly irrelevant technology. Can they transform themselves into the Samsung and Apple of the industry? It is a very different market from smartphones. Cars are not differentiated that much in technology already, but there are very strong brand images (only Apple really has a strong brand image in phones). Will people buy, say, an electric self-driving BMW because they used to build great non-self-driving internal combustion BMWs? Will a big network of dealerships be an asset or a liability? And will they need to build their own silicon to make the ultimate driving machine give a better experience than the non-ultimate driving machines available from a host of "me-too" companies running on standard silicon and software?

The Next Phase of the SDE Circle of Life

Billions of dollars ride on the answer to this question. System companies must adopt the ideas of System Design Enablement to get their own chips into systems. Alternatively, they have to buy them from semiconductor companies who need to adopt the idea since they have to ship a large part of the software stack, not just the silicon. The two worlds are merging in the next stage of the SDE Circle of Life. Systems can only be designed from top to bottom, from software to silicon. The big question for each company is: is it competing on some dimension that requires differentiation in the silicon, or are they using standard silicon and competing on some other dimension (software skin, price, dealer network, brand name, advertising...)?