Significance: The mobile chip market has endured five years of intense change but through that, Qualcomm has only strengthened its position. However, its inability to secure sufficient supplies of its newest chips from foundry TSMC have highlighted that it has vulnerabilities now, and Intel, in particular, is rushing to exploit them. More dangerous new challenges may come from Asia, however, especially as the mobile silicon battle shifts from phones to multiscreen platforms spanning handsets, TVs and PCs. This brings MediaTek, which is merging with Mstar to form a dominant TV chip supplier, more powerfully into the Qualcomm arena, and the US firm may consider drastic measures ‘ even abandoning its vaunted fabless approach and taking a stake in a foundry.
The mobile chip market has been in serious flux since the shockwaves when Nokia upturned its supply chain in 2007 (and then went into nosedive), sparking dramatic change at Texas Instruments and others. TI and Freescale have retreated from basebands; ST-Ericsson has faded; Apple and Samsung are increasingly self-sufficient; Broadcom, Marvell, Intel and Nvidia are all trying to knit together their various processor, modem and connectivity components to create the kind of integrated platform at which Qualcomm excels. These are not the only challengers to Qualcomm’s crown though. On the modem front in particular, there are many players gaining strength in Asia, while for its next generation offerings, the CDMA giant has to react to a change in the competitive landscape, and may even invest in its own foundry.
Japan is stepping up its challenge to US and Chinese modem makers. Renesas Mobile, despite the trials of its parent, is advanced in the push for multimode LTE chips. And a joint venture between NTT Docomo and three chip suppliers has been revived, albeit in a less ambitious form than when it was mooted, and cancelled, earlier this year. Docomo, Fujitsu, Fujitsu Semiconductor and NEC have formed a JV to focus on the development and sale of chips with built-in modems. Called Access Network Technology, it is majority owned by Fujitsu with Docomo and NEC each taking about a 20% stake.
Although these partners will achieve some economies of scale in R&D and manufacturing, and provide Docomo with access to advanced technology, it lacks the presence of Renesas Mobile. And two other companies which were involved in the original plan, Panasonic and Samsung, are no longer involved. This suggests scaled-back ambitions ‘ when Samsung was a participant, the JV was potentially a way for the Korean giant to extend its semiconductor offerings as it seeks to create a total platform to rival Qualcomm’s. It recently acquired most of the connectivity assets of the UK’s CSR and is clearly keen to expand its offering, but its decision not to involve the new partnership in its strategy may be a relief to Qualcomm.
Another lurking challenge from the east is seen in the planned merger of MediaTek and Mstar. These Taiwanese giants focus on low end consumer electronics and the Chinese market, and dominate the global TV system-on-chip space, driving out other players like Zoran, Intel and Broadcom. However, as EETimes points out, this deal is not just about owning the TV segment once and for all or scoring efficiencies. It is really about targeting the mobile and multiscreen devices ‘ taking on Qualcomm and Marvell in SoCs which can span phones, tablets, TVs and other content gadgets.
During an interview with EETimes at this year’s Consumer Electronics Show, MediaTek president CJ Hsieh said he wanted to offer a range of solutions with which ‘consumers can easily transport content across platforms’, adding ‘we also find enough synergy in technologies we offer both in smartphone and smart TV platforms.’ As well as the TV presence, MediaTek is the world’s second largest provider of handset modems, though it has struggled so far to move into higher value 3G smartphones. But the enlarged firm could dominate Asian ODMs and then build from there.
The reverse issue is whether Qualcomm, Nvidia and Marvell can really deliver on their own ambitions to support multiple content screens from handsets to TVs, given the dominance of MediaTek/Mstar in the latter. In an interview, Qualcomm’s VP of marketing for CDMA Technologies, Tim McDonough, commented: ‘OEMs are looking for a way to leverage the same operating system to build different products ‘ for one’s pocket, into a briefcase, or onto a wall ‘ without actually making them four times.’ The company has defined its cross-platform roadmap by outlining separate families within its Snapdragon processor range, each targeting a different type of device. Within the S4 series, for instance, the Snapdragon S4 Prime is for smart TVs and set-top boxes; the S4 Pro for PCs, tablets and high end smartphones; the S4 Plus for midrange smartphones and tablets; and the S4 Play for mass market smartphones.
As well as being brave enough to take on MediaTek in TVs and Intel in PCs, the chipmakers who want to cover all the digital device bases also have to learn from MediaTek’s low cost offerings. Qualcomm and others have been pushing down their prices for featurephones and ultra-low cost tablets, and even Intel, usually associated with the premium end of the market, is entering the game.
Most of the coverage of Intel’s mobile efforts focus on processors for high end devices, but the company is not confining itself to the premium end any more. The first commercial handsets sporting Atom are decidedly midmarket (the Lava Xolo and the Orange San Diego), and now the firm is touting RF solutions for sub-$100 3G handsets.
The latest product in the Smarti range, which Intel acquired with Infineon’s wireless arm, integrates an HSPA RF transceiver with power amplifiers on a single 65nm die, and does not require a separate power management component. The Smarti UE2p will reduce the cost of adding 3G to a basic handset, says Intel, and could even be squeezed into modules for machine-to-machine connectivity.
The chip will ‘simplify the product development and supply chain logistics with the reduction in component count and system complexity,” according to Stefan Wolff, VP of the Intel Architecture Group. “This will allow our customers to introduce lower cost 3G handsets and support the transition of the M2M market segment toward 3G-based connected devices.’
Such launches indicate the importance of the Infineon purchase, often overshadowed by the work on mobile Atom. Intel is not yet offering fully integrated SoCs like Qualcomm’s, but enhancement of the Smarti range, as well as ultra-low power Atom iterations, will pave the way to this ‘ and will be essential for mass market phones and even low cost tablets. The Smarti UE2p chip will be available as samples in the fourth quarter.
Beyond its massive R&D, its channel presence and its broad range of products, Intel’s proudest competitive boast is often about its foundries, and the resulting ability to control its own supplies, and to innovate in manufacturing processes for competitive edge.
Qualcomm, by contrast, has been vocal in defending the fabless model, with its flexibility and relatively low upfront cost. But the stance has come into question amid recent shortages of new Snapdragons and the company is reported to be considering taking a stake in foundry UMC. Qualcomm has been searching for new suppliers to supplement its main foundry TSMC and its CEO Paul Jacobs said recently that it was exploring various options to ramp up the flow of its 28nm processors ‘ and longer term, to increase its ability to control and guarantee its sources of supply. Early in July, it was reported to be negotiating foundry deals with UMC, GlobalFoundries and Samsung, to add to that with TSMC, but now sources indicate it will go further and invest in the former firm. That would increase its power to influence UMC’s directions and ensure it takes first place in the queue for capacity.
UMC recently offered a 10% stake to anyone wanting to become a ‘strategic partner’, as it seeks to catch up with the dominant TSMC. At the smaller firm’s current market valuation, 10% would be worth $540m.
If a closer relationship with Qualcomm materializes, it would fit in with an industry trend for chip giants to assert stronger control over their manufacturers. Intel has made much of Qualcomm’s discomfiture, stressing that, because it owns most of its own fabs, it can adapt process innovation precisely to its roadmap, as well as guaranteeing supply. And TSMC recently said it might set up dedicated factories earmarked for single customers, to enhance the level of influence they could promise their most important clients. TSMC has been trying to win significant Apple business away from Samsung, and such an arrangement might well suit Apple’s famous desire for control.
Announcing its plan to secure an investor, UMC’s CEO Shih-Wei Sun told analysts: “This is not for financial purposes, really we are looking for strategic partners.” He added: ‘As you know, the industry landscape is shifting aggressively now. The structure is changing. The radical supply chain partnership is getting more and more important. So we very much welcome the partnership.” He would not, of course, comment on the reports that Qualcomm was in the running, but said the stake could be open either to a customer or a technology partner.
As EETimes points out, Dutch semiconductor tools maker ASML recently engaged in a similar push, offering up to 25% in the company in return for strategic investment in developing its next generation process and R&D. Intel has signed up for 15%, for about $3bn, and Samsung and/or TSMC may take the remaining 10%.
AMD hires Apple’s processor architect:
AMD has been stranded in no-man’s land as far as mobile processors are concerned, but that may be about to change. The company has hired Jim Keller, one of renowned CPU designers of his generation, poaching him from Apple.
Keller has designed several generations of processors for Apple, and will now be corporate VP and chief architect for microprocessor cores at AMD, reporting to CTO Mark Papermaster,. He has also worked at Broadcom and was a pioneer of modern RISC processors when he worked on Digital Equipment’s Alpha architecture in the early 1990s.
AMD said its new recruit (who has spent a previous stint at the firm) will focus on the cores which will underpin the next wave of products, including high performance and low power chips. The latter may hint at AMD finally coming up with a convincing offering for power sensitive products like handsets and embedded gadgets. There have been persistent rumors that it would adopt the ARM architecture for this purpose, but so far has only signed an ARM licence specific to security ‘ although it has also unveiled a processor neutral architecture which can support ARM or its usual x86.
Papermaster also stressed low power credentials in his statement, saying ‘Jim is one of the most widely respected and sought-after innovators in the industry and a very strong addition to our engineering team. We expect that his innovative spirit, low power design expertise, creativity and drive for success will help us shape our future and fuel our growth.”
Keller was most recently a director in the platform architecture group at Apple, designing the chips which are found in the iPad, iPhone and Apple TV. He joined Apple when it acquired PA Semiconductor, where he was VP of design ‘ PA, combined with a Samsung partnership, was the foundation of Apple’s A4/A5/A6 processors and its bid to be self-sufficient in silicon.