Semiconductor upturn now in sight
Will finds encouragement in recent news that the chairman of the planet's biggest silicon foundry has recalled laid-off employees.
“TSMC Chairman asks all laid-off employees to return, says the worst is over,” according to the May 20th issue of Taiwan’s DIGITIMES. As the world’s largest silicon foundry, with the greatest number of chip clients, TSMC is the bellwether of the semiconductor industry. It’s time to perk up our near-term forecasts.
Smartbooks: A rose by any other name…
Between smartphones and netbooks lies the largely untapped world of Mobile Internet Devices (MIDs). Qualcomm proposes to call these devices “Smartbooks,” a shorter, catchier moniker.
Some argue that Smartbooks are just mini-netbooks, while others say that the market is well served with smartphones and netbooks. We believe that the “tween” market is real. Smartbooks will come in a variety of form factors, beginning with VGA or better screens that are 4 inches or larger. That’s higher resolution and larger than iPhone’s display. Some will have optional Mobile TV and TV-out capabilities, and all will have what you have come to expect in smartphones, including Wi-Fi and GPS.
One thing distinguishes Smartbooks from most netbooks: They are truly all-day mobile Internet appliances, with 3G cellular connectivity. Sure, a few (very few) netbooks and even luggable notebooks have built-in 3G connectivity, but they are highly unlikely to maintain an Internet connection throughout an 8-hour day while running on batteries. Smartbooks, on the other hand, will be closer to smartphones in that regard. Some will be touch-screen tablets, while others will have touch-type-able QWERTY keyboards, either as slide-outs or fold-outs.
New Smartbooks were among the many new products at Computex this month in Taipei. And more than a dozen companies plan to be shipping such products this fall. They will be largely based on ARM v7 chip solutions (Cortex-A8) from Qualcomm, Texas Instruments, and perhaps others. Nvidia claims that its Tegra MID (oops, Smartbook) chip solution (based on ARM 11) can result in $99 devices (plus 3G modem). Of course, we can expect Intel’s Atom processor family to be in some, too. We believe that Smartbooks could be a 40-million unit annual market by 2013. We’ll keep you posted.
For the homebound, there’s the “Media Phone”
Computex also featured several demonstrations of “media phones.” These devices appear to be a cross between a digital photo frame and a multi-unit cordless phone setup hooked to a broadband Internet connection.
AT&T's "Home Manager" is a media phone, billed as an “information appliance,” which looks like a digital picture frame with touch-screen capabilities. Home Manager provides “weather, news, sports, and more” along with alarm clock, answering machine, and PDA capabilities while linking several cordless phones that can also act as pagers to any other phone. The Samsung-made devices are priced at $349 or $149 after a $200 discount as part of a qualifying package (for AT&T U-verse subscribers with FTTx service). Oh, and $99 for each additional handset with charging cradle. Isn't that about the price of a low-end netbook plus a cordless multi-unit phone system combined?
But AT&T doesn’t have a lock on the concept. OpenPeak, for example, offers its “OpenFrame” based Media IP phone media center (based on Intel’s Atom processor), which offers the picture frame appliance format plus cordless units with LCDs for video display of things like YouTube websites. Also at Computex, ARC International is promoting its RISC IP and “Media Software Framework” as being ideal for media phones. No doubt, others are seeing this as an untapped market that offers new product opportunities.
Desperate housechips: DSP chip houses dump cell phone business
Although cell phones represent the biggest single market for DSP chips (as basebands and application processors), the companies that made them have been dumping their cell phone product lines. First to go was the Agere cell phone chip product line, sold to Infineon in August 2007. That was only seven months after Agere “merged” with what was to become LSI Corp. (I predicted that spinoff at the time of the merger). Next to go was Analog Devices’ cell phone chip product line that went to fast-growing MediaTek in January 2008. Then, Texas Instruments closed down its 2G/2.5G cell phone operations in early 2009 (though retaining its 3G cell phone business…for now). Now, Freescale is off-loading its cell phone chip operations, leaving only second-tier “DSP chip” suppliers like NEC and Renesas still serving the cell phone baseband market. Coincidentally, those two Japanese companies have begun talks about a possible merger.
Why did this happen? Although cell phones are a billion-unit market, the chip pricing pressures (especially for the China and India markets) and competition from companies that focus on baseband chips (but don’t sell “DSP chips”) created a situation of profitless prosperity for the traditional DSP chip houses, which devoted much of their resources to software and support tools that would appeal to a wide customer base rather than narrowly focusing on cell phones.
Freescale affirms cellular division sale, but doesn’t give names
Freescale has affirmed that it has signed a letter of intent to sell its cellular product group to another company. Freescale has not named the other company, but China-based Cosun Group put out its second news release last month (following one earlier in April) indicating that they will be the acquiring company. Cosun is a holding company, whose subsidiaries were shipping over a million cell phone handsets a quarter in early 2008…all to the China market. That pending sale apparently doesn’t include the LTE RF transceiver development group in Tempe, Arizona, which looks to have been sold to Fujitsu.
Freescale said that it will continue making complete iDEN chipsets for Motorola (for Sprint’s Boost service) and for RIM (for its BlackBerry Curve 8350i targeting the Latin American iDEN market). How they juggle all of this will be interesting.
Qualcomm and Broadcom make nice
Not long ago, the industry buzz was all about the settlement of long-running litigation between Qualcomm and Broadcom. Although many details remain secret, the companies agreed to end their litigation and exchange rights to each others' patents. Qualcomm will be paying Broadcom some $891 million over the next four years, and the companies agreed that during that period they will not make claims against each other regarding their patents on integrated circuits and certain other products and services.
The cash payout direction indicates that Broadcom had a plurality of patents in the trade. Clearly, this treasure trove of patents comes from over $10 billion in Broadcom company acquisitions from 1999 through about 2005 (over two dozen companies, by our count).
Qualcomm will still be able to seek patent royalties from cell phone makers that use Broadcom chips. Those royalties from handset makers (said to be 5 percent of the OEM value) are a key contributor to Qualcomm's profits.
A key question yet to be answered is, “Will Broadcom be able to produce CDMA 1xEV-DO chips under this arrangement?” Another is “How about LTE?” If Broadcom could produce EV-DO/HSPA chips or EV-DO/LTE chips, they would have a leg up on all other competitors except Qualcomm. Either avenue could propel Broadcom from its current 2-3 percent of the cell phone chip market to a respectable double-digit share within two years.
DSP silicon is a huge market; discrete DSP chips are just a small part
As noted earlier, DSP chip companies have largely left the cell phone market. Most of the companies now providing DSP silicon to the cell phone market don’t report them as “DSP chips.” Consequently, the DSP chip market is shrinking, even though the overall DSP silicon market continues to grow (except for 2009). To clarify this, we present three 2008 charts:
Details of this TAM noted in Figure 1 are included in our forthcoming study of the DSP silicon market, mentioned below. Just as in MPU parlance, anything that is not an off-the-shelf Pentium (oops, we mean DSP chip) can be considered the “Embedded DSP” market, which is $24 billion worth of silicon.
Does Intel lead the way for TI?
You may recall that earlier this year Intel announced that it has made its Atom processor core available as hard macros through TSMC to those who want to make their own SoC based on that X86 processor. This brilliant move capitalizes on the knowledge that the bulk of silicon products are headed for SoC devices, and Intel can’t design them all. Besides, the biggest unit market for processors is in cell phones, not in PCs, and Intel no longer has a baseband product to field in that market segment. But this approach (likely employing the “Moorestown” next-gen version of the Atom) promises to get Intel into future LG Electronics smartphone chips.
Intel wasn’t the first to take this TSMC platform technology approach. ARM has enabled several of its popular cores to be available as prevalidated hard macro through TSMC. Using the hard macro version of the IP reduces risk, ensures predictable performance, saves resources, shortens time to market…and enables companies to field SoCs based on ARM’s popular processors.
CEVA Inc., whose licensed DSP cores are already in 100s of millions of cell phones and other devices, began this approach a few years ago, too. CEVA DSP cores are available as pre-validated hard macros at TSMC, UMC, and SMIC. Again, CEVA’s approach expands the number of companies that can produce SoCs based on its popular processors.
Texas Instruments, still the No. 1 supplier of discrete DSP chips and the No. 2 supplier of cell phone chips, is finding that it is being designed out of future Nokia basebands, diminishing its overall DSP revenues. TI would be able to expand its overall DSP market footprint by making its family of DSP cores available through the TSMC platform (or taking a similar approach through another silicon foundry). This would allow hundreds of companies already familiar with TI’s extensive software, design, and debugging tools to produce SoCs, expanding TI’s revenue base beyond the off-the-shelf discrete DSP chips. Moreover, TI would be able to expand its current “embedded processing” emphasis employing this approach.
As illustrated in the previous section, the biggest market for DSP silicon is in “embedded” SoCs that are in hundreds of products…generally not programmable by the user and known as “what they are” or “what they do,” as MP3 player chips, GPS receivers, disk drive controllers, set-top box transcoders, MPEG decoders, video game audio chips, IP phone chips, etc.
Of course, the same silicon foundry path is open to Analog Devices, Freescale Semiconductor, NEC, and other discrete DSP chip suppliers.
Latest study encompasses entire DSP landscape
Two new in-depth market studies covers the entire landscape of the integrated circuit market based on DSP technology, not just DSP chips. “DSP Silicon Strategies ‘09” is the most extensive study on the planet of markets based on DSP technology and details the sea change in the DSP chip market that renders traditional measures meaningless. Many components that were once reported as “DSP chips” are no longer. Rather, they are reported as systems on chip (SoC) in categories like ASICs or ASSPs, even by traditional DSP chip vendors like Texas Instruments and Analog Devices. This is the only market study providing worldwide electronic equipment production forecasts (by region and by application market) and both IC and DSP content for each year through 2013.
Details are available at: http://www.fwdconcepts.com/DSP%2709/index.htm
As always, I invite your comments.
President and Principal Analyst