Written by Lisa Ann Cairns, Ph.D.
Mobility's boom is having a significant influence on the semiconductor supply chain, well beyond the financial. This demand-market event is not only changing the direction of all types of design innovation (through research and development [R&D] investments), it is also pushing the commoditization of chips and subassemblies.
The Device Market’s Ripple Effect on Components
The impact of high demand for mobile devices goes well beyond unit sales and reverberates through the semiconductor and electronics supply chain. Although tablets and smartphones, leaders in the set of smart wireless devices (SWD), are not yet able to replace the full functionality of a PC, sales of these devices show that this is not a deterrent to consumers. Individuals and businesses who favor these SWDs over (or along with) PCs are influencing not only the CE market, but the entire supply chain supporting that market. Furthermore, SWD purchases are contributing to the delay in the ongoing, and long-overdue, enterprise-hardware refresh cycles.
Market trends are showing that the boom in mobile devices is having a significant influence on the semiconductor supply chain, well beyond just the financial. The high demand-market favoring mobile devices is not only changing the direction of all types of design innovation (through research and development [R&D] investments); it is also pushing the commoditization of chips and subassemblies.
Mobility puts chip architecture on the move
The positive, high demand for SWDs, specifically tablets and smartphones, is driving change along the entire semiconductor supply chain. IDC's recent growth estimates for smartphones and tablets forecast an increase of 61%, 500 million units, from 2012 to 2014, and another 25% increase, or 300 million units, from 2014 to 2016. Confirming these growth trajectories, leading OEMs, notably IBM in its recent fab future discussion, offer insight into the sheer number-power of mobile as a driving force for revenue and, more importantly, into tracking the direction of component and chip-level architecture.
As a result of mobile's gaining central place in the semiconductor and electronics industry, there are at least two critical drivers for the current, accelerated changes in chip architectures:
(1) Compressed Time to Market (TTM) demands for increased chip design flexibility all along the supply chain – from OEMs through to the consumer, shorter cycles are expected; and
(2) Significantly increased integration to enable improved device functionality and features while shrinking size and weight, and while keeping costs low.
Delving into these two market forces, and then considering the current semiconductor supply chain landscape, tells us quite a bit about the near- and long-term trajectories of chip design and fabrication from a supply and a demand perspective.
Let's consider what happened in 2012: there were significant consolidations across the upstream semiconductor supply chain, i.e., the manufacturing segment, from memory to analog and even to fabs and foundries themselves. There was a significant increase in fab-lite and fabless companies with a notable concentration of just a few leading-edge fabs and foundries remaining. Among the reasons for these consolidations were (1) the staggering cost of fabs and equipment for increasing complex chip architectures; and (2) the volatile and still conservative end-demand that still dampens many semi and electronics market sectors, like the long-awaited enterprise PC refresh.
So, where financial investments are made now is directly related to what the demand cycles favor downstream in the supply chain. Inventory holds are seen as risky, so utilization and capacities are kept in tight control. These strategic decisions work to reduce supply and support pricing, but not to the point of spurring demand. One critical variable is the dramatically shortened Time to Market (TTM) necessary to provide the quickest end-market upgrade options to consumers who continuously demand more from their mobile devices. These shortened cycles of consumer expectations for new, innovative features and functionality are directly influencing how and where manufacturers are investing resources and what they are requiring of design engineers. In short, these market forces are driving increased flexibility, compatibility, and condensed, more highly-integrated chip architectures than ever before.
In an important way, it all comes back to what "mobility" means to a user and how that change in use is influencing the very architecture of not only chips, but of the semiconductor industry as well. Mobile devices are where market demand and focus are and are likely to stay at least through the decade. The advances thus far in SWDs have pushed the expectation for seamless and constant connectivity into a growing range of devices. At the same time, these growing user expectations, combined with the changes in device capability, are pushing data transmission levels to extremes and challenging carriers and cloud service providers to meet surging demand.
As device users expect more from their hardware, they continue to take for granted, and so demand, the ability to be always connected, always able to update and upgrade, always able to acquire and manipulate data.. The end result of this demand is a notable shift among chip and component design engineers and manufacturers to move rapidly and decidedly towards more System in Package (SiP) structures, with the end-goal of full System on Chip (SoC) architectures. These design changes are pursued to compete in the tough mobile market and provide the complex, integrated architecture necessary to deliver content- and graphic-rich mobile experiences to a wide variety of users.
The dawning of SoC commoditization
Faster, smaller, more power-efficient, multitasking, connectivity, high data transfer capability – these are the hallmarks of what "mobility" means today, of what is demanded from the latest SWDs. These features also happen to describe what is demanded of components. The ever-shrinking device-form has moved us to design complex, highly-integrated SiP solutions. SiPs are providing today's demanded design engineering and manufacturing solutions while we move closer to solving the problems encountered with 3D stacking (cf. Ed Sperling's article in Semiconductor Manufacturing & Design), and work to better and more cost-effectively produce the desired, single-die, SoC solutions.
Today, most CPUs are multi-core and are controlled by symmetric multiprocessing (SMP) architecture. Already, here we have to stop and recognize that the once more-distant domains of chip architectures and operating system (OS) architectures have, necessarily, become more intertwined. The move to interconnect OS and hardware architectures is one of both cost savings and improved processing with the end-goal of not only meeting demand, but leading designs that drive demand.
Along this path of tight interconnection between OS and hardware at the design level, we are seeing increasingly-integrated chips, put first into multi-chip modules (MCMs), now put into chip stacks (bringing to mind an early version of 3D ICs). As we improve designs and continue to architect solutions that combine more functionality into a single package to reduce heat, power consumption, and size, we move toward the single die SoC solution series. However, at this time, what these SoCs gain in feature capabilities, they also lose in cost-reduction because of the manufacturing complexities and the still relatively narrow device markets.
But we have to think about the market situation, as mentioned above – SMPs are really the domain of PCs, or at least the larger category of computing devices that are not bound by power efficiency requirements. Unfortunately, at least for the PC sector, there is a significant slump in demand, and that, of course, is slowing demand for these multi-chip modules (MCM) and the SMPs that drive them. Because mobile solutions require more efficient battery use and greater heat dissipation, they have not followed this SMP, multi-core, or MCM solution path. Rather, the solutions used in mobile have been asymmetric and/or heterogeneous, as was recently discussed in an article on SMP and AMP by Kurt Shuler, V.P. of marketing at Arteris.
Recalling the issue of cost and the 2012 history of consolidations in the semiconductor industry, the business strategy in the design sector of the semiconductor and electronics supply chain has been to maximize design flexibility to increase use of the chip sets to meet multiple product needs. As recently noted by Shuler in System-Level Design:
In 2013, many of the companies that once designed multiple SoCs to address the needs of a variety of different customers will migrate to a single platform strategy because of the rising costs of silicon. This single-platform approach requires greater options so that chips can be reconfigured on the fly to address multiple markets.
We've come to realize that companies need to redesign portions of their SoCs twice as often as they previously did. […] chip designers […] use[d] the interconnect fabric to incorporate elements into the SoC so that the chip was better suited to meet specific markets [sic] demand. […]
A lack of flexibility in the late stages of a design, or the inability to use a single SoC platform to address multiple markets, has left some companies sitting on the sidelines while more nimble players reap the rewards.
There is much to think about in these statements – from a supply chain or industry metamorphosis perspective, and especially from the perspective of understanding what is happening at the chip design level within the semiconductor industry. These two events, industry metamorphosis and chip design change, are correlated and are simultaneously driven by cost and by continuous demand for ever-evolving, M2M mobile solutions, particularly leveraging cloud computing (cf. this recent article in Chip Design Magazine by Jamie Little, ESD at Mentor Graphics).
Where are these design changes and market strategies leading us? At the forefront is a new, commoditizing era for SoCs. Given the design requirements for mobile solutions, and given the decline and shift away from the hardware and firmware solutions for PCs due to the end-device declines, the opportunity and the strategic business value to push SoC designs into more flexible, more commoditized levels are dawning.
Why does SoC commoditization matter? GROWTH.
It is not that SoCs are by any means a golden goose for semi at this point in time. Nor is it the case that the moment flexible SoCs can be successfully commoditized will solve the ongoing concerns across the semiconductor and electronics industry. Yet, given the phenomenal adoption rate of mobility into end-device solution series, those who do not stay abreast of this quiet but revolutionary shift in chip architecture are already finding it difficult to compete.
It is not that SoCs are the breakthrough; they are the expansion of mobility and M2M connectivity. The data transfer at the core of M2M is changing how we interact with and use end-devices and what we expect of them. If we continue to think that improving the PC is the solution, then we've missed what the demand side of the market is telling us: mobility matters to users. Mobility matters even more than full functionality because if users are mobile, they can connect and access what they need via cloud computing from wherever they are. The devices that serve users for such use case scenarios, and these devices and the infrastructure that supports the M2M and data handling both require different architectures than what PCs brought us in the form of MCM and SMP solutions.