Qualcomm is proving a far harder nut to crack than many had expected. A few years ago, analysts were full of the dangers to the giant ‘ the decline of CDMA and therefore its IPR stranglehold; the rise of new competitors like Nvidia, especially in sectors where Qualcomm had no track record, like large-screen devices. Yet the giant continues to widen its lead in mobile processors and SoCs, and to expand its reach into new devices (internet of things and home Wi-Fi equipment being important ones), and into new products such as the RF360 front end.
Old rivals like ST-Ericsson and Texas Instruments have faded away in its core markets, and even the great challenger, Nvidia, is finding it tough. The company reported a sharp drop in revenue from its Tegra mobile chipset in its second quarter ‘ down 49% sequentially and 71% year-on-year, to just $52.6m (less than the $66m the company received from its patent licence agreement with Intel). As well as indicating how important is Qualcomm’s unique mixture of licensing and products (and despite the decline in CDMA, it is still a major patent holder in 4G too), Nvidia’s statements highlight the other vital aspect of the larger chip vendor’s success, its tight integration of modems, processors, GPUs and, these days, Wi-Fi and RF. The key to Nvidia’s upturn will be, it hopes, success for its first all-in-one chip, the Tegra 4i, which will integrate the LTE software-defined modem technology acquired with Icera.
This will bring it closer to Qualcomm’s stronghold in mobile SoCs, but others are chasing that end too. For once, the most important of those challengers is likely to be Intel, so long a non-starter in the mobile world, but now finally offering a credible low power processor, as well as riding the rise of ‘post-PC’ devices where it is pushing its ultrabook platform and its Windows 8 tablet chips.
Intel, though, has also lagged in integrating its modem technology, acquired with Infineon Wireless, with its Atom processors. An all-in-one Atom platform will be essential to gain top tier handset deals but the world’s largest chipmaker is determined to push its technology well beyond the catch-up stage. That has been seen in a series of announcements to improve Atom’s ability to outdo ARM-based rivals in performance and even Intel’s usual bugbear, power consumption. Cutting edge new processes ‘ by contrast, Intel’s usual strong suit; a string of small acquisitions to fill gaps in the platform; the increasing integration of extra capabilities at chip level, such as the McAfee security technologies or NFC. All these are steadily turning Atom (and Core, which will also be used in some higher performance mobile products) into a fully rounded and distinctive platform.
However, the incorporation of the LTE modem is an essential next step before Intel can be considered a true Qualcomm challenger. Intel’s big acquisition in this respect was the wireless unit of Infineon, which majored on basebands, though with its strengths weighted to the low end. But a new purchase may prove even more significant, though smaller. The US firm has acquired Fujitsu Semiconductor Wireless Products, an Arizona subsidiary of the Japanese giant which was well known to Intel from WiMAX days, and has now developed a highly advanced LTE RF transceiver.
This move is very significant, because it brings top class RF designers to Intel, which will be able to accelerate the pace of creating integrated Atom offerings, and potentially allow the firm to leapfrog key rivals and their RF partners. Will Strauss, president of analyst firm Forward Concepts, believes the Fujitsu transceiver and design team is ‘probably the best standalone RF team in the world’, with a heritage going back to Freescale and, before that, Motorola (Freescale was originally Motorola’s inhouse chip unit).
The key attraction for Intel will have been Fujitsu’s newest RF transceiver, which supports advanced features such as antenna tuning, envelope tracking, and an internal DSP that supports any application processor, including the x86 instruction set. Strauss calls this ‘probably the most advanced multimode LTE RF Transceiver available in the open market’. Many of its features will gain importance with LTE-Advanced, and indeed, Fujitsu is also working on an RF chip specifically for LTE-A, supporting carrier aggregation. It brings the much-needed LTE modem ‘ Infineon/Intel has 2G/3G modems and is shipping a multimode LTE RF transceiver, but a 4G modem was a crucial missing link.
These two products will be a major boost in Intel’s ability to tap into the high end LTE device market and the nascent LTE-A sector, where Infineon was well behind many rivals in R&D terms, and which will be vital to gain slots in high profile, cutting edge devices ‘ the kind which operators and OEMs promote most heavily, as SK Telecom is doing with its first LTE-A handsets. Slots in such flagships may not deliver huge volumes ‘ LTE-A will not be mainstream until 2015, despite the early moves of some cellcos ‘ but they are essential for product credibility and to attract the attention of other customers. And of course, the OEMs are designing their LTE-A gadgets now, to try to attract carriers for the 2014 fall season and beyond.
While Qualcomm leads the LTE-A wave in multimode handset terms, and is busily proclaiming that ‘LTE-Advanced is here’ with its latest Snapdragons, there is a race to support key LTE-A capabilities, particularly CA. Samsung Electronics and Broadcom have shown off their next generation chipsets for all kinds of mobile devices, while there is also rising interest in LTE-only solutions, especially at Verizon, which will bolster the fortunes of specialized 4G modem providers like Sequans. The French 4G veteran offers the Cassiopeia LTE-A product, sporting 150Mbps Category 4 capabilities, carrier aggregation up to 40MHz, eICIC and eMBMS, optimized VoLTE and enhanced MIMO.
So Intel has catching up to do, and the Fujitsu unit could be a powerful weapon to help it do so (one of the smart, surviving WiMAX-turned-LTE specialists could be another, should it look for further purchases, and bearing in mind Broadcom’s acquisition of Beceem).
As well as expanding its actual product capabilities at an accelerating rate, the giant has also been enhancing its production processes to support its next steps in many key businesses ‘ defending its core server business from ARM by developing more microserver chips; signing up foundry customers; broadening the telecoms infrastructure range (Crystal Forest will get a refresh this fall); and of course driving down Atom’s cost and power consumption still further for the mobile and IoT spaces.
As it finally seems to respond with the required urgency to ARM’s threat, it is supposedly preparing to slash six months off the timescales for next generation Atoms based on 14nm processes, which would improve its competitive position in the 2014/2015 timeframe. Currently, one of Intel’s problems is that there is a gap of a year or more between the release of new Core chips and Atom parts built on the same architecture, a reflection of the old order in which PCs took precedence over mobile. That has to reverse now, and reports suggest that there will be a far shorter gap between 14nm Core offerings, due in volume in the second half of 2014, and 14nm Atoms, which could arrive just six months later. The 14nm generation is widely expected to put Intel in the forefront of the mobile market, in power/performance/cost terms, for the first time ever, unless ARM pulls a major rabbit out of the hat.
One reason for keeping Atom as a clear second string has been the lower revenues and margins that come with the mobile chip family, encouraging Intel to prioritize Core. But with the decline of the PC space and the massive expansion of devices that use low-power silicon, Intel has to find a way to adapt to those new cost/profit assumptions, and sacrifice margin for the mobile market share it badly needs, offsetting the losses with other revenue streams such as foundry services, as well as rigorous cost control. With Chinese chip suppliers getting more capable by the day, neither Intel nor Qualcomm will be able to rely on the premium prices and margins of the past, and there will be painful adjustments to make. As Qualcomm knows, these will involve many cuts and compromises, but none of these should be in developing and acquiring the best technology for the next wave of mobility. Both these giants have impressive R&D and IPR investments, but Intel’s have often been less focused on real market opportunity than its mobile competitor’s ‘ or have been translated less well into actual revenue. That needs to change, and one of the most pressing challenges for new CEO Brian Krzanich will be to align R&D and M&A effectively to short term market opportunity, and ensure excellent technology translates into sales as directly as it does at Qualcomm.
Ericsson in the modem business:
While ST-Ericsson has disappeared from Qualcomm’s competitive landscape with the break-up of the ailing joint venture, theoretically its mobile products remain on the market.
Under the terms of the break-up (since the reluctant parents failed to find a buyer), Ericsson keeps the modem properties and STMicro gets the applications processor. That breaks STE’s flagship all-in-one platform, NovaThor, in two and ends the chance for either owner to play in that space. However, the two firms could deploy their respective assets, either in partnership with more mainstream mobile chip suppliers, or in non-smartphone segments.
Ericsson’s modem business effectively recreates the former EMP (Ericsson Mobile Platforms), which was a leading vendor of 3G baseband/RF integrated platforms before merging with the ST-NXP joint venture, and a key rival to Infineon Wireless. The reconstituted EMP still has 11% share of TD-SCDMA modems, according to Forward Concepts, and 9% of GSM/EDGE, but only 2% in 3G, though the HSPA flavour of the Thor modem was used last year in smartphones from HTC, Samsung, Motorola, Panasonic and Sharp.
However, it does have a credible multimode LTE modem/RF transceiver, and it will be interesting to see whether Ericsson puts any real effort behind establishing this as a rival to Intel’s portfolio and other vendors, as an option for processor-only firms, or a future sale prospect. Its best options may lie in developing TD-LTE modem/transceivers for China Mobile, a route STE was already going with some success.
Meanwhile, STMicro has the ‘Nova’ half of the STE equation, the app processor. It is unlikely to target smartphones, given the crowded nature of the market and ST’s vocal desire to move away from mobile devices and into embedded products. It could deploy Nova in its strong automotive segment or in multimedia platforms, and could even partner with Ericsson again to create an integrated TD-SCDMA/TD-LTE offering for China.
Intel to roll out two LTE modems this year:
As well as the plans for Fujitsu LTE-A products and an integrated SoC, Intel is making more immediate 4G modem moves and will deliver two products from its Infineon unit this year.
Next month, it will ship the XMM 7160, as multimode LTE modem which can work on as many as 15 LTE bands as well as 2G and 3G networks. Thiswill differentiate itself with small size and low power consumption as well as flexibility, according to Intel, and Hermann Eul, general manager of the firm’s mobile and communications group, says the former Infineon unit has grown up a lot under its new owner, moving from being a ‘fast follower’ targeting mass volume 2G and 3G markets, to focusing on early mover advantage.
“For us, it is the clear strategy, the clear plan, to go into a leadership position, and we are currently in the transition to do that,” Eul told an analyst briefing,.
Qualcomm already sells multimode LTE chips for a large number of bands, but Intel is claiming a performance advantage together with 12% smaller size and 20% to 30% lower power consumption than rival products.
The first version will support Category 3 LTE (up to 100Mbps) and an update, to appear by year end, will move to Cat-4 or 150Mbps and will also support VoLTE. It may also add TD-LTE to support carriers like China Mobile and Sprint/Clearwire.
In parallel with the 7160, Intel has been developing the XMM 7260, which is due to ship in the first half of next year. This will support LTE-Advanced features like carrier aggregation, which Intel demonstrated on an evaluation board at the launch event. This chip will definitely add TD-LTE.
While the former Infineon unit currently has its products manufactured by foundry TSMC, Intel plans to bring this activity into its own fabs in order to reduce cost, increase control and ease integration with other product lines ‘ the new Fujitsu additions, and with Atom for combined SoCs.