In a surprise announcement last week, Motorola agreed to buy 50% of UIQ Holdings from Sony Ericsson. Pending regulatory approval, Motorola’s co-ownership of UIQ questions the US-based OEM’s vision for mobile Linux handsets. Motorola has shipped more than 9 million Linux-based handsets to date, while in August reaffirmed its commitment to base as much as 60% of its device portfolio on a Linux operating system by 2012. So did Motorola have a sudden change of heart and was that a diversion or a U-Turn from Linux ?
The dent in Motorola’s Linux vision. Motorola has to date made huge investments in building a mobile Linux platform. The investment began in 2001 by Mark Vandenbrink s Beijing-based team tasked to develop an in-market, for-market operating system with reduced costs for the manufacturer. The OS initially known as EZX is based on MontaVista’s Linux-based kernel, and uses Trolltech’s Qt/E (now Qtopia) for graphics and application framework, both of which have been heavily customised by Vandenbrink’s team. Over time, Motorola replaced the 2.4.20 kernel used in EZX with a newer 2.6.10 kernel and renamed the platform to L-J (for Linux-Java) which uses Sun’s KVM virtual machine for supporting third party applications.
During Motorola’s 6-year Linux history, the OEM has launched around 15 models (A1200, A728, A732, A760, A768, A780, A910, E680, E680g, E680i, E895, MING, ROKR E2, ROKR E6) for the Chinese market and recently the RAZR2 V8 for the US and European markets. At LinuxWorld in August 2007, Motorola re-baptised the L-J platform as the more brand-policy-friendly name MOTOMAGX and reaffirmed that “in the next few years, up to 60% of Motorola’s handset portfolio is expected to be based on Linux”. The remainder of Motorola’s portfolio would probably be powered by Windows Mobile for enterprise devices, UIQ for high-end handsets, and TTP Com’s Ajar for low-end handsets.
Yet only two months later, Motorola bought into a 50/50 ownership of UIQ from Sony Ericsson. As a Symbian spin-off, UIQ (along with its Symbian OS base) is practically a competitor to Motorola’s MOTOMAGX.
Is this is a dual supplier strategy for Motorola ? Hardly, as the OEM is under extreme financial pressure and needs to urgently trim its cost base. Motorola s operating profits have suffered a major blow in the last year, dropping from an operating profit of US$ 819 million in 3Q06 to an operating loss of US$ 332 million in 2Q07, according to Fitch Ratings. At the same time its market share dropped from 21.1% in 2006 to 14.6% in 2Q06, according to Gartner. In response, Motorola announced a reduction of 15% in R&D budgets.
Is the UIQ announcement a tactical move? Certainly not. Motorola could easily continue licensing UIQ from Sony Ericsson as it did for its five Symbian OS -based handset models to date, the A1000, M1000, A925, A920 and the recent Z8 handset.
Analysing Motorola’s change of heart Motorola’s financial troubles are only the trigger behind Motorola’s change of heart in its Linux single-platform strategy. I would argue that there are four reasons for Motorola’s rethink of its Linux strategy.
1. Motorola has been quietly trying to develop a single-core version of its MOTOMAGX platform, in order to reach a planned 50-60% of its handset portfolio, as indicated by the announced-but-never-released MotoRizr Z6. The move from a dual core to a single core architecture would mean a major re-architecture, as a single-core OS has to run both the applications and the modem stack. Virtualisation techniques (see WindRiver, Trango and VirtualLogix) are designed to faciliate single-core OS development, but I suspect Motorola would still have to re-architect major parts of its Linux-based OS even if it used virtualisation.
[updated: a reader reports that the MotoRizr Z6 has been released in China and is based on Freescale’s Starcore single-core CPU. According to the same source, the Z6 is far better in terms of performance compared to the ROKR E2. This finding implies that Motorola is 6-9 months ahead of any other OEM in launching a single-core based Linux stack.
Well, it turns out that Motorola’s Z6 is based on a dual core Freescale MXC275-30 SoC (single chip) architecture, and not a single core one. This is according to a Freescale presentation which you can find here. The model name has been changed from MotoRizr Z6 to MotoRokr Z6, according to a LinuxDevices report, while the handset appears to be available in 20 countries according to the same report. Moreover, given that no single core Linux handset has been released by Motorola, this would strengthen the argument on how single core Linux remains a challenge for the Linux OEM champion.]
2. Motorola has invested man-centuries into building MOTOMAGX, based on MontaVista s Mobilinux Linux support package and Qt/E (an old version of Trolltech s Qtopia). Motorola has had to add lots of glue and optimisations on top of Mobilinux and Qt/E, and so a migration away from these components would mean significant re-investment. Yet this is exactly what Motorola would have to do if it is to keep its costs down for equipping more than half of its product portfolio with Linux and at the same time to migrate to a new scalable, end-to-end UI framework architecture as other handset OEMs are doing. MontaVista and Trolltech make money by selling developer seats and Motorola had in 2Q06 ordered 200 developer seat licenses (in addition to the 100 they already had).
3. Motorola s 9 million Linux-based phones up to mid 2007 have shipped in China and Latin America, primarily due to the relaxed device testing/certification and operator customisation requirements in these countries. The RAZR2 V8 which started shipping in July 2007 has been the first Linux-based phone for Western and European markets. It is likely that Moto s Linux platform strategy met with long delays due to increased requirements for handset certification (GCF in Europe and FCC in the US), the stringent network interoperability testing requirements (particularly with US operators) and the need to comply with voluminous operator customisation requirements (in Europe these measure at 4,000 lines of requirements per handset and change twice yearly).
4. Motorola has been the leading force behind the foundation of LiMo, its chairman, its chief architect, according to Nomura’s Richard Windsor. In addition, Motorola has committed to contributing to LiMo multiple software components (package model, application execution model, architecture, registry (with Samsung), security policy, certificate manager, event system+input method (with Panasonic/NEC). However, the LiMo foundation has recently more than doubled in size, while its complex licensing models and unfamiliar processes for new contributions has likely resulted in more delays and higher resource commitments for Motorola. As the leader in LiMo, Motorola may have deemed that incorporating LiMo requirements into its MOTOMAGX platform would prove too costly.
Why Motorola invested in UIQ The number one priority for handset manufacturers is to make money from handsets. Handsets come first, while software and hardware platform strategies come second. At a time of extreme financial pressures and competition from Nokia and Samsung, Motorola had to stick to its product commitments and seek an alternative software platform for launching its handsets in 2008 and beyond.
Ajar, Its other in-house platform which Motorola acquired alongside TTPCom is designed for low-end handsets, not mid-range or high-end ones and was intended to complement the L-J platform. The manufacturer has also launched Windows Mobile based handsets, but these have been targeted to enterprises, not consumers.
Motorola has also launched five handset models based on UIQ, the M1000, A1000, A925, A920 and Z8, so why not continue persuing a typical platform licensing strategy ? Let’s do the math.
Motorola has been looking to scale it s Linux-based handsets from an estimated 2-3% portfolio share in 2007 to a claimed 50-60% in 2012. Assuming a linear growth and a steady 15% mobile device market share this would mean that the OEM would need to build around 225 million handsets based on Linux in the next 5 years, of which around 50 million in the next two years.
Now assuming that a tenth of Motorola’s 2008 and 2009 portfolio of Linux handsets would have to move to UIQ, at $3 royalty, this means a cost of $15 million to Motorola. How much is UIQ worth ? Symbian’s financial statements do not yet account for the UIQ sale, so it’s difficult to tell. Nomura reports that 1.2 million UIQ phones shipped in 2006 (a paltry 2.3% of Symbian-based smartphones), which at around $3 per-unit royalties implies $3.6 million in annual revenues, and at a x10 valuation factor, UIQ is worth around $30 million. Therefore the value of Motorola’s acquisition of 50% of UIQ is the same as the licensing cost for 10% of its 2008/9 planned portfolio of devices, were Motorola to replace Linux with UIQ on these devices.
With 1.2 million devices shipped in 2006 and nearly 150 staff, UIQ is definitely a loss-making business. Therefore, Motorola’s move was clearly strategic, with the OEM hoping that the UIQ software will help reduce platform TCO and time to market, at a time of challenged Linux strategy. Motorola would longer-term benefit from reduced licensing costs, a major stake at defining UIQ’s roadmap and hopefully a profitable licensing business once UIQ one-handed interface penetrates higher volume devices. The 50/50 split shows alignment of incentives between Motorola and Sony Ericsson, but also hints at the fragile balance between the two competing manufacturers down the road.
The sale of 50% of UIQ also makes sense for Sony Ericsson, who is dealing with a massive and rapidly growing cost base, with employee numbers more than doubling from 142 staff in February to over 350 in October, and new offices in Budapest and London.
There are two more noteworthy consequences here:
1. To support Linux developers at a low cost, Motorola essentially handed off development of its Linux SDK to Trolltech. The Norwegian software vendor was probably eager to accept, given the poor performance of Qtopia’s dual-licensing strategy and the industry alignment behind the competing GTK graphics library (LiPS, LiMo and GMAE all endorse GTK, not Qtopia). [updated: I spoke to Trolltech who clarified that they don t have direct responsibility for Motorola’s SDK. Motorola licenses Trolltech’s Qtopia SDK and is free to make it available to handset application developers]
2. Motorola s UIQ strategy will likely lead to a reduced presence in LiMo, particularly since the OEM is heavily resource-constrained and because LiMo has specified a GTK-based graphics stack. To keep it s membership in LiMo and save face, Motorola would most easily contract an ODM to release Motorola-branded devices based on other distributions such as Linux stack vendor Celunite, who recently joined LiMo, rather than re-architect its own stack.
Repercussions for Symbian In the short term, Symbian s industry valuation and prospects have been significantly strengthened as a result of Motorola s UIQ co-ownership. In the medium term however, Motorola does not have any ownership in Symbian, and so I doubt the co-ownership of UIQ will impact Nokia s near-majority influence (read: control) over Symbian OS.
Longer term, the OS game for UIQ stakeholders becomes quite interesting. The Symbian stack provides little value above the kernel and drivers (Symbian has essential become akin to a board support package) – read the latest specs of S60 (here) to see that not only the application suites and UI frameworks, but also the vast majority of middleware have been drawn out of Symbian OS. In other words, the value of the Symbian OS software stack is similar to that of a zero-royalty Linux-based stack from the likes of MontaVista and WindRiver. [updated: I realise that the above argument is not backed up with hard data. I hope to delve into some research to detail the components in the Symbian stack vs those in MontaVista/WindRiver distros and what is the *sale* value of each component].
However, while academically it s possible to replace Symbian with a Linux support package (kernel, hardware drivers and base OS functions), it is an expensive undertaking, of the order of $50 million. Given Sony Ericsson and Motorola are facing difficult times competing with the Nokia giant on launching successful handsets, large-scale investments in a software platform is hardly the priority these days.
Diversion or U-turn ? Motorola s acquisition of 50% of UIQ is clearly a strategic initiative, which will likely continue with the launch of several high-end handsets powered by UIQ in the next year. At the same time, Motorola has far too much invested in Linux. It was only late 2006 when Motorola’s Christy Wyatt, said that There isn t a group within Motorola s 70,000 workforce that isn t impacted by Linux and open source in one way, shape or form .
Consequently, UIQ is a diversion, not a U-Turn for Motorola. While R&D investments are being curtailed, the manufacturer will naturally try to re-use its assets and existing Linux-based research, especially since its mobile Linux software know-how is second to none in Western markets. Medium term, Motorola will likely pursue a dual-OS strategy (UIQ and MOTOMAGX), with the more ambitious and demanding handset projects (esp. western markets) powered by the mature UIQ platform.
Longer term, the fate of Motorola s Linux platform strategy will depend on the success of the UIQ investment and how UIQ will help the OEM constrain the total cost of ownership, cost of variant creation and time-to-market for its new handsets. Time will tell.
– Andreas