So maybe AMD was right to skip the smartphone stage after all. The x86 firm has been much criticized for its failure to enter the mobile processor space, despite seeing Intel and Nvidia moving from the PC/server markets into the higher growth segment. AMD has gone no further than chips for netbooks and ultraslim notebooks. Despite a change of CEO, an ambitious multi-platform strategy and an ARM licence, the company is still steering clear of the smartphones.
This may prove to have been a good move, though probably the result of lucky blundering rather than skilful forethought. As Intel knows, to make a real impact on smartphones required making a decisive move several years ago (to be fair, Intel did recognize that, but lacked the right platform until it debuted its ultra-low power Atom last year). Later entrants with high performance, PC-style offerings are really seeing growth coming from new form factors like tablets and phablets, not from handsets ‘ witness Nvidia. Increasingly, the smartphone segment will commoditize and center around a battle for low cost, low power reference platforms. This requires scale and deep ties into the mobile ecosystem and will see the mainstream handset battle royal being one between Qualcomm and MediaTek, not Qualcomm and Intel.
It is hard to see why Intel is bothering to play in that low end, reference design game, though it is making the attempt. The margins will be razor-thin and only worth chasing if a firm can capture massive volumes ‘ unlikely for Intel or most of the other mobile CPU contenders. Most of the profit growth will come from emerging ‘post-PC’ devices such as tablets, phablets, ultrabooks, cloudbooks and hybrids (and a few more form factors yet to be dreamed up). With these premium, highly differentiated products comes, for chipmakers, the chance to win long term, hard-to-break alliances with OEMs rather than just fighting on price.
This, then, is where Intel, AMD, Nvidia and others should be trying to steal a march on Qualcomm, not in commodity devices. And since 95% of smartphones will be commodity devices by the end of this year, AMD may, with hindsight, look wise not to have expended vast energies and resources on chasing that space. Having missed the Orient Express in the middle of the last decade, there is little point in boarding the cattle train now.
The question for AMD is whether it is ready to be a major player in the post-PC device space, where it is essential it carves out a market, unless it is going to shift away from devices altogether as PCs decline and focus on servers and embedded segments. We have often analyzed the vital need for Intel to dominate the products which will replace the PC, with ultrabooks and Windows tablets currently seen as its natural home. It will be delighted with the lukewarm early reception for the ARM-based Windows RT tablets, as its partners ready their own alternatives, running x86-based Windows 8 Pro, for next month. But AMD’s future in the device space also depends on these large-screened products, not on phones, and its strategy has been articulated far less aggressively.
However, the firm has been building up its resources, hiring senior engineers from Apple and Qualcomm and reported to be creating a platform geared to the business tablet segment. For instance, at the recent SolidWorks World conference, Fujitsu showed a prototype slate running on AMD FirePro, an accelerated processing unit..
This product hints at an AMD offering which will target very high performance and professional tablets rather than the mainstream consumer and insiders said the firm is keen to lead a redefinition of the tablet ‘ to be a high quality content creation engine, not just a content consumption device as it is usually perceived according to the norms set by the iPad.
AMD’s FirePro A300 Series of chips are based on its Trinity architecture but current commercial models are far too power hungry for tablets and are targeted at workstations by graphics designers or engineers. The Fujitsu prototype indicates that there is a new version in the works for low-power gadgets, probably a modification of the existing ‘Hondo’ chip. This would have superior graphics capabilities ‘ AMD has been a heavy exponent of integrated CPU/GPU architectures ‘ and would run Windows 8 Pro, with the advantage to business users of supporting legacy Windows applications.
Defining and leading a high margin, if somewhat esoteric, sub-segment is a good plan for a chipmaker which has its eyes on other markets beyond classic device wars. Last year saw AMD putting the foundation stones in place for a new approach to the market, one which would bring it out of the shadow of Intel by looking beyond x86. CEO Rory Read, who took over in 2011, admitted during the firm’s recent results call that the PC decline was still weighing heavily on performance, but that AMD was charting a new course to center on non-PC sectors, notably embedded systems, heterogeneous computing and dense servers (no mention of handsets). He said the strategic shift would take several quarters to translate into significant growth but new graphics chips and processors would appear in the first half of 2013. Read told analysts: “We have the right strategy and a new set of products coming to market in 2013.’
To extend its influence and support its new strategies, AMD has been pioneering heterogeneous architectures that combine CPUs and graphics cores on one chip, and it has been opening up its own Fusion platform to third party cores and partners. Last summer it went a step further by creating an industry alliance around the concept, called HSA (Heterogeneous Systems Architecture) Foundation, which could provide a significant challenge to Intel’s influence. The group includes ARM and Texas Instruments, among others, indicating how hybrid approaches are moving from the high end into end user devices.
The main weapon against Intel will be a software layer which will make it easier for developers to program for complex multiprocessing chips, and so could convert them to HSA-enabled platforms. Along with ARM, TI and AMD, the Foundation currently includes MediaTek and Imagination Technologies ‘ the latter interesting in itself, as it offers the main rival, in terms of mobile graphics processor IP, to ARM’s own Mali. The partners will need a few more sign-ups to claim a de facto standard, and stressed that they were reaching out to influential players like Qualcomm and Nvidia.
Including an x86 supplier in an ARM-centered group, as well as Imagination, emphasizes the claims to be aiming for a universal platform, and point to a heterogeneous world where cores with different architectures can be mixed and matched according to their particular strengths. But more tactically, the line-up could also be regarded as ‘anything but Intel’ (though Intel does have a stake in Imagination). AMD will launch HSA-enabled silicon in 2014, first for PCs and later servers, but increasingly looks likely to leave smartphone versions to other partners such as MediaTek.
The involvement of the Taiwanese baseband giant in such initiatives indicates how it is spreading its own wings, moving up the food chain into 3G ‘ and soon LTE ‘ smartphones with integrated processors, modems and even GPUs. This is the real smartphone chip war now ‘ not around specialized processors with dedicated GPUs, but around the platforms, and here the only serious challengers to a Qualcomm/MediaTek duopoly will be up-and-coming Chinese firms.
Qualcomm – with its famous premium pricing at risk and facing somewhat less profit cushion from its IPR in future years ‘ will have big adjustments to make to avoid being lapped by MediaTek in this sector, while also making these activities profitable. It showed off some of its weapons at its QRD (Qualcomm Reference Design) Summit in Shenzhen, China this week, promising quad-core Snapdragon S4 chipsets with fully integrated modems (3G of all flavors but not yet LTE).
The QRD is at the heart of the US firm’s bid to halt MediaTek’s moves to expand its smartphone system-on-chip base at the expense of Qualcomm. The QRD provides full platforms with tools and pre-tested components, to reduce cost and time-to-market for carriers and handset makers. As such, it targets the heartland of MediaTek’s business, as the Taiwanese firm starts to move up from its featurephone base.
The US chip giant says QRD now supports more than 170 devices, plus more than 100 designs in progress, and some of these support the TD-LTE, the key Chinese 4G standard. The Summit will feature over 40 hardware and software suppliers demonstrating their use of the reference designs to simplify high volume smartphone development, said Qualcomm. Among the featured partners will be China Mobile, Tianyu, Little Chilli, Yulong and Tencent. The company plans to introduce a new online hub for QRD manufacturers to enable them to receive a custom device software package tailored to their unique specifications, quickly and easily.
This reference platform world is not a natural one for Intel with its strengths in high performance, but it is pitching Atom SoCs at the space nonetheless, claiming its huge investment in advanced fabrication processes will enable it to lower cost and power to the levels needed in the mass market. Intel says reference phones are important ‘calling cards’ to gain operator and OEM interest, and to the surprise of many, its first smartphone wins for Atom were not for premium superphones, which might require the performance of an Intel chip, but for white label devices from India’s Lava International, adopted by Orange.
Chasing the mass market smartphone platform is probably a loss leader for Intel, and its fourth quarter results further highlighted the risks and pain of the transition from PCs to the devices that will eventually replace them. The figures showed the impact of the falling growth in PCs, and Intel’s very inadequate transition (so far) to the mobile and embedded markets, which leaves it very dependent on its still-strong server processor base. Even that is under threat from ARM-based challengers, but this is a fortress Intel is well prepared to defend, and it is also seeking new revenue streams and models, notably with a move towards foundry services. According to a Bloomberg report, it has added Cisco to last year’s tentative experiments providing chipmaking to small partners.
Despite the underlying challenges, the results topped Wall Street expectations and analysts have found reasons for optimism in some recent mobile breakthroughs ‘ deals with handset makers such as Lenovo and Google/Motorola, and last week’s launch of the latest Atom SoC roadmap targeting affordable smartphones. There are also hopes for a major role in tablets courtesy of Windows 8 Pro, with models due to start appearing this spring, as well as a slow but steady momentum building behind ultrabooks.
For Q412, Intel reported net income of $2.47bn, or 48 cents a share, down from $3.36bn, or 64 cents a share, a year earlier. Excluding certain items, per-share earnings totaled 51 cents, down from 67 cents. Analysts had expected earnings of 45 cents. Revenue fell 3% year-on-year to $13.48bn, below Intel’s own guidance, made on October, of $13.6bn and also below Wall Street forecasts. Gross margin slid to 58% from 64.5%, slightly better than expected.
However, the analysts may have been moderately pleased, but the results were still down on the year-ago quarter, and Intel’s outlook was cautious. Its projections for the first quarter of fiscal 2013 and for the full year were slightly worse than anticipated. Intel further stoked nervousness by saying it planned to spend $13bn on investing in buildings, plant and equipment, particularly for the latest wafer size, 450mm ‘ $2bn up on last year, though presumably partly to support the move towards foundry business.
Intel said it expects revenue to rise in the low single-digits on a percentage basis in 2013. Gross margin should be around 60%, below its historic norm of 65%, which analysts increasingly think will never be reached again. Analysts expected revenue of $54.3bn this year, which would be up 1.95% from 2012’s $53.3bn. For the first quarter, Intel expects revenue of $12.7bn, plus or minus $500n, while Wall Street was expecting $12.91bn.
“The slightly weak outlook is not a surprise, and gross margin a little higher than people thought is a positive thing,” Bernstein analyst Stacy Rasgon wrote in a client note. “But capex is going to be concerning. It seems awfully high.”
The biggest problem for Intel is that the PC base is declining more quickly than it had hoped ‘ unless the ultrabook can be classed as a PC and gains steam this year ‘ while its mobile business is not growing nearly quickly enough to fill the void. The firm admitted its mobile processors will not ‘move the needle’ for the whole company until next year, when it has high hopes for the emergence of an LTE mass market. By 2014, Intel will have wider range of price points for its low power Atom SoCs, and will have integrated these fully with the modems it acquired from Infineon, to offer an all-in-one platform to counter Qualcomm’s Snapdragon/MSM combinations.
Meanwhile, Intel has reportedly signed a foundry deal with Cisco, to make chips for the firm’s networking products. Last month, outgoing Intel CEO Paul Otellini said Intel was open to being a manufacturing partner for a strategic, but non-competitive partner, and Cisco’s huge requirements would certainly help soften the blow of the stepped-up capex investment in new equipment. Rumors had swirled around an Apple deal, but if that did transpire in future, it would probably be for future generations of products, from 2015 or so.
That is the problem for Intel ‘ its new processes and even some of its SoCs have significant promise but will take another year or more to translate into the kind of growth and profits the firm needs. As it pins its hopes on a strong reception for Windows 8 Pro tablets, and an uptick in ultrabook adoption, the last thing it needs is a rejuvenated AMD biting at its heels in the post-PC world.