We sat down with Christopher Taylor, director, RF & wireless components at Strategy Analytics for the discussion about the market.
Kristin Lewotsky: Wireless capability has gone far beyond phones to everything from DVD players and printers to cars and houses. Can you give us a market size to put it in context? Christopher Taylor: The market for radio components, including semiconductors and RF filters, was about $50 billion last year. We figure almost half of that is components going into cellular terminals, and that's not just handsets anymore. With the average selling price of basic chips coming down, we are starting to see more applications for cellular outside of the traditional handset so that is a growing segment of the market. Some of it we characterize as machine to machine, it's also PC data cards, cellular the USB dongle's, cellular embedded in notebooks and tablets and the list goes on. We estimate now an additional 25 to 30% of the market goes into these nontraditional types of wireless devices.
KL: How has the introduction of the smart phone changed market dynamics? CT: The chipsets are not the whole thing. We now have 32 approved bands around the world. A lot of this is for LTE but nevertheless there are lots of bands and you have to add the RF front end to the equation. That may not be significant for some of these non-handset applications but it's very significant in handsets, especially smartphones where you've got lots and lots of bands, so you have to add dedicated power amplifiers (PAs) and duplexes and RF switches, and all kinds of additional parts and board real estate to cover everything you need to cover. You might have a universal chipset but then you've got to implement it as a global device with lots of bands or be region specific.
KL: Who are the main players in the market and is likely to change soon? CT: Qualcomm, MediaTek, and Intel are the leaders in the chipset space.
KL: Is it a rational market? CT: There have been tremendous changes in past two to three years. Qualcomm keeps getting bigger. MediaTek came out of nowhere by serving the low cost market. They’re having trouble getting a lot bigger with that strategy, so they’re starting to focus on the top OEMs, as well. Texas Instruments slowly getting out, and ST Ericsson is struggling. Intel basically bought Infineon’s wireless business unit, and then you have Spreadtrum Communications coming up. Broadcom has been growing nicely as well. They have pretty low share now but they're probably going to be one of the survivors.
KL: What are the biggest challenges right now? CT: You have to optimize the whole radio as a system. You have the RF parts and the digital parts—you have to partition it properly and optimize it for performance. Qualcomm is in a pretty strong position. They’ve integrated in the baseband the CPU, which is one or more ARM cores; the digital signal processing; the graphics processing; the power management; the baseband processing; and the digital processing for GPS, Wi-fi, and Bluetooth. They have a lot of pieces and that's just the digital side. They've also got good RF technology—they bought Atheros so they’re strong in Wi-Fi now. Even if they don't have digital piece integrated, they can bundle it with the chipset. They're very well-positioned.
[The problem is that] RF doesn't follow Moore's law. A radio chipset for a cell phone consists of two pieces: the baseband processor and the RF transceiver, which has some mixed-signal pieces—typically D/A converters and A/D converters. It also has pure RF pieces. So there's some scaling going on in the transceiver because there are some digital bits but not to the same extent as the baseband processor which is generally purely digital.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.