Virtually every chip provider in the mobile world has spent considerable sums trying to get into the modem game, but it has proved really tough to chip away at Qualcomm and MediaTek. These two ‘ one at the high end, one at the low, but increasingly converging on one another ‘ look like taking the duopoly role that Texas Instruments and Freescale enjoyed in pre-smartphone handsets.
Broadcom is the latest silicon provider looking to shed its baseband business, saying it is assessing its ‘strategic alternatives’ for the unit ‘ only nine months after it acquired Renesas Mobile, another large chip firm’s failed modem business. It has appointed JP Morgan to advise it, and expects to cut its annual R&D and administrative costs by about $700m once the unit is sold or closed. CEO Scott McGregor said the “economic opportunity is not sufficient to justify the continued investment when compared to other opportunities within our portfolio’ especially as, at the high end, there are so few potential customers.
Broadcom did gain a high profile win for its M320 dual-core LTE SoC, which shipped in a Samsung Galaxy Tab model this year, supporting Cat 4 speeds (up to 150Mbps). But that came too little too late, even though it was an early fruit of the Renesas purchase, according to analyst Will Strauss of Forward Concepts. ‘They were at least two years and maybe three years behind in getting to the LTE modem market,” he said, allowing Qualcomm to snap up most of the device wins (it has over 90% of LTE modem market share).
The LTE technology remains the only really valuable element in a baseband unit which otherwise has focused on low end offerings for 2G/3G. That activity had high hopes for a while when Broadcom was chosen as a preferred supplier for Nokia’s midrange handsets, but the Finnish vendor went into decline shortly after anointing the US company into its inner circle.
As for buyers, Qualcomm does not need the former Renesas technology, and it is unlikely that Intel or MediaTek do either ‘ after all, they did not buy it when it was on sale a year ago, and it is certainly the main jewel in the Broadcom business. MediaTek or Chinese provider Spreadtrum, both clawing their way up from low end chipsets into 4G, might be interested, and Marvell is still clinging on in cellular too. However, previous sellers of modem businesses, such as ST-Ericsson, have struggled to find purchasers, in a space which has been somewhat awash with modem makers recently ‘ and where Nvidia could perhaps be putting Icera on the table soon too. A big player in search of real LTE innovation, without too much legacy baggage, could do better to wait for that eventuality, or seek out the LTE-only specialists such as Sequans and Altair.
Analysts at Strategy Analytics estimate that Broadcom has spent over $3bn on baseband R&D over the past seven years, with minimal return ‘ it ended 2013 with about 5% share and expects baseband revenues of $200m to $250m in the first half of this year. It saw only 4% year-on-year growth in 3G modem shipments in 2013, down from 193% the year before. One problem, the analysts said in a research note, was that ‘Broadcom viewed its baseband business as supplementary to its connectivity business. For Qualcomm the baseband modem comes first and everything else is a revenue expansion opportunity.’
Broadcom now plans to merge its connectivity and broadband comms team to form the Broadband and Connectivity Group, to be headed by Daniel Marotta.
Broadcom’s decision follows Nvidia’s recent admission that it would no longer focus on handsets ‘ putting its integrated modem/processor offering, Tegra 4i, and the technology it acquired with Icera, in doubt. TI and Freescale themselves, of course, exited handset modems long ago, ceding the smartphone market to Qualcomm once their tame customers, Nokia and Motorola, were toppled from the premier league. Intel bought Infineon’s baseband unit, among others; while STMicro, Ericsson and NXP combined their activities only to see the resulting ST-Ericsson dissolved last year (though Ericsson has bought back the modems and is still developing them).
This bleak outlook has various important outcomes for the mobile balance of power. It suggests that only a few providers ‘ Qualcomm, Intel (soon), MediaTek and some Chinese newcomers ‘ will have fully integrated, single-vendor SoCs combining the modem and processor, generally considered a sine qua non for the midrange to low end device space because of cost and power considerations. That implies that most processor makers will have to give up on handsets, as Nvidia has, and chase markets with more forgiving cost/power requirements (tablets, set-top boxes and so on), or new spaces where Qualcomm has not yet built its fortress, notably the internet of things (IoT).
The IoT, predictably, is the market which all the also-ran handset chip vendors are proclaiming as their new growth area, but the transition is not an obvious one. While cars, like tablets, will not necessarily need their modems on the same chip as their processors, most connected devices will rely most heavily on the the modem, with limited CPU and GPU performance required. Vendors will need to decide which bit of the vast IoT to target ‘ connected devices which will have considerable graphics and compute capabilities, like Nvidia; gadgets connected by WiFi and Bluetooth rather than cellular; the gateways and hubs which will tie them all together.
Some, including Qualcomm, Intel and MediaTek themselves, will try to address all aspects, while others will invest in new platforms for sensors, microcontrollers or specialized modems ‘ nascent volume segments where there is still room for a new entrant to make its mark. For those which stick with modems, part of the gamble will be deciding which technology will prevail, and when ‘ LTE, an evolved 2G, a new platform? There is great uncertainty over whether LTE will be suited to most IoT applications, at least in the M2M (rather than the connected consumer) segments.
But where it it appropriate, as in the connected car, chip suppliers disappointed in the handset space could do worse than look at the surviving 4G-only modem specialists. Broadcom itself bought one of those, Beceem, but didn’t do much with its technology. Others remain independent, notably Sequans of France, GCT of Korea and Altair of Israel, all survivors of the WiMAX era and now focused on the expected rise of LTE-only devices as the first carriers (such as Verizon) start to switch off 3G. 4G-only will also be relevant in data devices that do not require wide area roaming, such as larger tablets and post-PC products; and in LTE-based M2M.
Sequans, for instance, offers a range of EZLinkLTE modules, which combine baseband, RF and key interfaces in a single package to reduce cost and time to market for vendors. Two were recently certified by Verizonfor its Band 4 and Band 13 spectrum (700MHz and AWS) ‘ the VZ20Q surface mountable module for CE and M2M devices, and the VZ20M M.2 interface module for tablets and PCs (there is a third model, the VZ21M, which includes GNSS for location services).