"I'm goin mobile," The Who sang on one of the greatest albums of all time. Although it is not what they were talking about then, today mobile is the biggest market in the world, especially if you look at the bigger picture and include most of the Internet. Far more people access the Internet from mobile devices than traditional PCs, especially anyone who owns a “phablet”, or large-screen smartphone. Companies like Uber and Lyft are mobile companies, moving into mobile things that used to be done some other way. I don’t believe you can even book an Uber on a PC. Partially because, why would you? And partially because very few (none?) PCs contain GPS capability. The same goes for mobile banking, mapping, managing collections of photos, and so on. These are all things that used to be done offline, then perhaps on PCs, but now increasingly on mobile devices. With 6B mobile devices expected by 2020, people can afford to ignore the other 10% of the market. The big market winners in these shipments are ARM, Google (Android) and Apple. The losers are Intel and Microsoft, although they are having huge success in building out the datacenters that all these markets require as back ends. The markets addressable by mobile sum up to about $7T. Some markets, most obviously telecom, but also others like cameras, have almost completely gone mobile except for a few high-end professional devices. The market for point-and-shoot cameras which were standard a decade ago has gone and many of the participants have withdrawn from the market. In fact, I lost my own point-and-shoot digital camera (I mainly used it for taking pictures of screens and panel sessions at events) somewhere around my apartment and I haven’t bothered to put a whole lot of effort into finding it. My iPhone is not as good, especially with optical zoom, but it is fine. Tomi Ahonen, a mobile consultant based in Hong Kong, who I met there a year or two ago, has a good summary diagram. The key to reading it is that the width of the green arrows represents the size of the market and the distance that they reach into the oval shows how much they have “gone mobile.” His estimate is that mobile today is $1.6T, so the transition is about a quarter complete Not all of these markets require specialized semiconductor technology, but many do. Music needs low-power MP3 players. Mobile payments need NFC. Mapping needs GPS. Image recognition requires neural network implementations. Obviously at Cadence we look at the mobile market, especially smartphones, as the largest semiconductor end market. It has also been one of the fastest growing. Although not every chip in a smartphone runs on the latest process nodes, mobile has been a key driver of process technology and the associated design methodologies. In particular, the application processor and perhaps the modem (in some cases, these are combined into a single chip) have migrated from 28nm to 20nm to 16/14nm, with current designs going on at 10nm and work starting at 7nm. Each of these process nodes has required a wholesale change to the implementation tools in particular, with multiple-patterning and FinFETs just the most obvious. Many of the leading-edge mobile companies design their own IP, but the next wave of users of these advanced processes will require IP to be provided from IP suppliers such as Cadence. Not every design group has the knowledge or the financial muscle to design, say, their own LPDDR controller. Foundries have had to be much more aggressive at working with IP companies to ensure the critical IP is needed when required. When fabs cost perhaps $10B, then they depreciate at $50/second, which is an expensive resource to have idle if IP is not ready when designers need it. In fact there has been a slow change in design methodology, and again mobile is one of the drivers. To truly build optimized systems, everything needs to be designed together, from software down to the silicon. This is what we at Cadence call System Design Enablement. It is not possible to design software and silicon independently since each requires the other as part of their verification (although the software people usually call this testing, which is confusing since in the silicon world that means something else). I see the archetypical case in the fact that the market leaders in smartphones design their own application processors and the semiconductor merchant suppliers produce a complete software stack (with a large part coming from Google in the form of Android, of course).![]()
![]()
↧
