Reverse Disaggregation - How Silicon IP Will Change the Semiconductor Supply Chain
Kevin Walsh, Director of Product Marketing, Synopsys, Inc.
The title might seem a bit like hubristic marketing jargonreverse disaggregation, what sort of nonsense is this? But often, if you create an interesting expression, such as the expression disruptive technology, or radical innovation, the words help conjure up a mental picture of what is actually taking place. The words can take root and help crystallize a concept. So lets begin to create a picture. The frame of this picture is IP, or silicon-based hardware and software IP (SIP). IP providers will play a central role in the next big landscape change in the semiconductor supplier market, reversing a trend that began over forty years agothe disaggregation of the supplier market in semiconductors. Reverse Disaggregation will drive the creation of a new supply chain for the development of a new generation of products, full featured and assembled at zerocost. IP will influence how the new supply chain is created, reversing the disaggregation that has marked the industry up until now.
Industries aggregate and disaggregate in an effort to improve efficiency. The direction depends on whether consolidation or specialization offers the maximum cost savings (or some other compelling market reason, like time-to-market, or some killer feature). The Semiconductor Industry has disaggregated into specializations Design Automation software (EDA), Pure-play Foundry (as IC factories), SIP (silicon intellectual property), Packaging and Distribution all in an effort to maximize value, differentiate, lower cost, or to offer some other market advantage. Until now these strong trends, driven by the efficiencies of specialization, changed the landscape of the semiconductor industry, creating separate sub-industries. But new market trends are changing this, or in a sense reversing this disaggregation.
Specializations have been advantageous and opportunistic. From the egalitarian availability of high quality FABs providing access to billion dollar silicon factories without billon dollar investments, creating a Fabless Semiconductor industry, to the availability of sophisticated design tools to take advantage of the wealth of available gates to create true Systems-on-Chip (SoC) productsmoving from chip-sets to system-chips. Each disaggregation resulted in achieving real industry gains and many came at just the right time to enable the industry to march forward in line with Moores Law. Not insignificant in this is Synopsys own contribution institutionalizing synthesis as part of the design flow. And not insignificant will be our contribution in redefining the industry with our IP.
The Big Quake
The market is changing from offering products that were either differentiated OR low cost to products that are differentiated AND low cost. Market control is moving to fewer and fewer suppliers. Large consumer grabs are replacing the land grabs of the nineteenth, and the industrial grabs of the twentieth centuries. Fewer suppliers control a significant portion of the market in dollars and volumes. The bigger they are, the harder they can fall, so market share control comes at some risk. Many markets are driven by the consumer, either directly like handsets, or indirectly like medical equipment. Consumer markets are fickle. The consumer mantra is Why cant I have this cool new product now?
Why cant I have this extra service now?
Oh, these things should be cheaper. Such a buying environment creates huge cost and market timing pressures on companies trying to deliver new products with large semiconductor content.
What are these Global product companies (Global Giants) looking formarket franchise, market dominance. To get and retain franchise they need product hits, products with characteristics not unlike movie box office smash hits. The ticket to a steady stream of buyers is capturing consumer buying sentiment, that nebulous combination of cool industrial design and killer functionality. One hit can completely transform (or rebound) a company. Take the iPod. New generations of consumers know Apple as the creator of the multi-media player market and not for creating the personal computer market with the Apple II. Who can argue with the RAZR cell phone as the rebound product of the year? Motorola was down and out against the onslaught of cell phone competition but six months after the release of the RAZR they were back in the game.
More and more the complete product package is special because of the soft features added iTunes for the iPod, ring tones or colored skins for the cell phone. The marketing groups of these Global Giants are spending more time trying to master these soft features. Once they have a hit, the strategy then seems to be to flood the market with consumer choices at different price pointsdelivering a kind of consumer choicebased mass customization. So these Global Giants are demanding a new restructuring to align their soft market feature ideas with hardware/software from their semiconductor suppliers to better position them for the large consumer grabs they need for success. Under these new pressures the fault-line of the semiconductor industry slipped, and we are experiencing the effects of ongoing aftershocks.
Convergence vs. Fusion
You may have sensed the quake, the convergence of the computer, communications and consumer markets the 3 Cs, the merging of functions from each independent product discipline into a new single product. But semiconductor suppliers are reacting to its aftershocks, a new set of Cs, the fused Cscost, competition and control. Together, these new Cs are at the center of any possibility that the convergence market expansion will continue successfully. A newly constructed supply chain is reversing the previous disaggregation trend and reassembling itself to meet this challenge. The reverse disaggregation is driven by the new demands of Global Giants who are trying to service enormous consumer markets, under intense consumer market pressure. So the semiconductor industry, whose innovations laid the foundation of this consumer gold rush, now must turn to how to master the realities of a new market. Semiconductor suppliers must restructure themselves to better address the needs of these Global Giants, their customers, to meet the needs of their customers customer. Expansion of this new market requires building a new supply chain which can excel in delivering in these three areas:
- Cost (ability to assemble product at near zero-cost)
- Competition (ability to build sustainable differentiation)
- Control (ability to generate profit margin over time)
Figure 1: The new market reality.
So how do you build sustainable differentiation? You may have noticed the semiconductor supplier market is extremely competitive. So our second C, competition, the ability to sustain some real differentiation to compete, needs to be addressed as part of our supply chain re-assembly. The most compelling answer to building sustainable differentiation rests with a concept called vectors of differentiation. Here the idea is that a basic product platform can and should be extensible over time not hinging on a single feature but on the vector-direction of the feature, and how it can be expanded, adapted and delivered to a changing market. A product platform built for vectors of features improvements sets the strategic context for developing an architecture. This eliminates the need for high-level management tinkering in the details over the life of the product family. Once a strategic architecture is set, new features flow naturally into products. IP is an important element in building new vector-oriented platforms. IP designed and tested to plug and play in an orderly way enables an inherently more flexible foundation (put it in or leave it out). Further, the architecture can be built around a common backbone, like a common bus structure in an SoC. This allows for a platform to be extended (as completely new feature requirements come into the market).
The third C is controlthe ability to generate profit on products over time. The issue here is how to create what I call the Gucci effect. This is the tangible differentiation of style with superior quality. Gucci has a certain market cache. It is created mainly by design, but supported with quality as well. Semiconductor companies generate Gucci by providing killer functionality with tangible differentiation while delivering it with a feeling of style (using a cool design or package). TI has a great example in its OMAP platform. For handset and mobile manufacturers, OMAP has a rich feature set for them to choose from and it has a certain package cache that demands copying. In fact, that may be the ultimate measure of success in establishing a Gucci effect, competitors want to copy you. In OMAP you find a combination of highly differentiated proprietary IP blocks and enough supporting IP to provide the platform with legs to carry it into many different end products. The investment in OMAP is substantial. You can see the supplier risk in creating it and the need to have it successful both in its differentiation style and its quality, similar to Gucci style and quality. Can IP be created in a Gucci style?Certainly. Can it feed seamlessly into new OMAP type platforms?Its a must. The IP industry already has examples of high-end products. Processors, DSP blocks, specialized functions, hard PHYs, differentiated IOs, complex protocolsall with some differentiated characteristics. Add to this mix new IP developed with low-power architectures, special packaging options and robust design methods, and IP will further move up in style points toward the creation of real IP brands, which will have a true Gucci type cache.
Front and CenterIP
So IP will emerge as is a key ingredient in the new supply chain, brought on by the requirements of the Global Giants, reversing the disaggregation forces that up until now dictated the construction of the semiconductor market supply chain. The nature of this reversal hinges on IP suppliers. Why? Because fused inside the consumer, computers, communications product convergence are suppliers reorganizing to provide product platforms optimized for cost, competition and control. Whether hardware or software, IP elements will dramatically change the way suppliers will aggregate to provide these Global Giants with what they need to be successful. A disaggregated market does not produce enough flexibility and sustainability to achieve the economies for developing the necessary product set. A business model with fused with cost, competition and control will deliver a new solution capable of meeting the new market realities. IP will be the fulcrum of how a new supply chain is re-constructed. And the elements of that IP will be built:
- for emerging foundry nodes
- to wow the customer
- for new interface protocols
Figure 2: IP leveraged with tools and services.
In order to use a new foundry node you need hard interfaces (I/Os and SERDES). The problem with the march of technology is that these mixed-signal and interface blocks are process dependent. Companies are finding it costly and time-consuming to move these functions to new nodeslike 65nm and 45nm. You need to get these key functions on the process to generate revenue. Mixed-signal and I/O IP built for portability, centered across critical process parameters and built in a robust design style will bring timely availability back into a reconstructed supply chain. Without new hard IP design methodologies, new process nodes will go under utilized. IP providers positioned to deliver robust interface IP quickly will play a key role in providing these functions for foundries. Foundries partnering with these IP providers will go to volume faster with this new approach.
How can IP create excitement? Providing unique architectures that take advantage of new low-power tool flows and are architected to allow system control power optimization is a good first step. The skill and creativity in IP design itself will be a differentiation that IP companies will use to deliver wow features to customers. Wow features will be the norm for IP developed in a fused C world. Without these features, IP will be quickly commoditized and will not provide enough profit to pay for sustaining engineering. Products will entropy and become a support burden, or worst, become the one element not working in a SoC. IP optimized to quickly provide multiple ports using a shared clock or constructed with partitioning in mind to allow for performance tradeoffs between hardware and software are two more examples of how IP will change to help provide adaptability with performance, wow features for next generation product platforms.
Access to IP early in the market lifecycle of new protocols will be another part of the new fulcrum. Unless the IP industry begins development early, the economics of the new fused environment will be sub-optimal. Costs will be incurred twiceonce when a lead company develops a new protocol to enter a market quickly and again when they realize they have to buy externally because of exploding internal costs or new revisions that require redesign. The IP industry must begin to invest earlier as a new protocol develops. Further, they must begin building earlier to deliver during the first generation of a new standard. Only then will the cost models work to the benefit of everyone.
The final point in the IP fulcrum is quality. The new supply chain cant tolerate poor IP quality. Horror stories abound about IP providers who dont have the proper design methods to insure a high quality product. IP development is not a garage business anymore. Look at the risks in the context of a printer SoC which is deploying in multiple product families. Defects, returns, rework cannot be limited anymore to a single product or product family, so any issue with any part of the common platform is disastrous. New standards for quality IP are becoming common in the industry and IP built to new quality standards will be the only choice for the new supply chain.
Even as the IP industry addresses these four pivotal elements it must expand its view of deliverables beyond its block centric present to a broader user future view that includes how IP will integrate and interact in a system. Take a look at a frank comment from an IP user posted to a popular feedback site and youll understand why.
Yes, the IP blocks work standalone and conceptually we the buyers believe that we can integrate and verify the whole system and be happy ever after. But Toto, we are not in Kansas anymore. It is a lot more than connecting the ports and passing/overriding parameters on an instance to integrate the IP blocks in our systems. All sorts of issues ranging from initialization and register/signal accessibility to test benches and debug process become severe bottlenecks. Integration of multiple IP blocks, possibly not from the same vendor, is no trivial matter. Sometimes, you have to run three different license daemons just so that you can integrate these models who needs lawyers?
The point to discern from the above quote is that IP suppliers must realize the challenge of integration, assembly and system test in addition to meeting the fulcrum needs of foundry IP availability, wow features for differentiation, delivering on new protocols and developing IP with superior quality.
A New Supply Chain
How will the new supply chain aggregate again to meet the challenges of the new market landscape? What should we expect in the next few years? My view is that IP will link with Tools and Services to form a tight integration that can produce the flexible subassemblies for product platforms and then even complete product platforms for product consumerization. Shortly after this transformation its likely we will see companies who have all these elements (and more), or companies that build alliance partnerships successfully, move further up the supply chain and become Original Design Manufacturers (ODMs). ODMs will produce a complete design or product that will be branded by a System OEM, one of the Global Giants, for selling under their label and into the market franchise they own.
Figure 3: The new supply chain.
How can this possibly succeed? IP providers must morph to provide complete configurable templates or platforms (including hardware and software) that perform some important system function. Tools must be assembled into flows to build flexible platforms. Tools must work in flows with IP to deliver completed products. The flows must address the vectors of differentiation that will build differentiation into the end product. Flows must address unique market needs like lower power systems. The must address the unique needs of advanced semiconductor processes for increasing yields. Services must move from assisting with elements of a project to providing a complete solution in line with ODM buyer requirements. Providing ODM content will require mastering new disciplines in marketing and engineering to better understand consumer requirements and to develop the methods to deliver.
What makes Pixar great and a great buy for Disney? Is it the creative force of producing the movie theme or the superior execution of the animation?In truth, both. In the consumer electronics space it will be similar. Killer functionality and the superior industrial design will rule the markets. A new supply chain (reverse disaggregated from the existing one) will rise to meet the challenge. It will deliver zero-cost platforms, platforms with vectors of features to sustain differentiation in consumer markets and it will deliver it with Gucci style. Only this new supply chain will meet the demands of the consumer markets emerging across the world. IP and IP providers will be squarely in the center of creating the delivery engine to build the platform to deliver the products to generate the profits.
For more information on Synopsys DesignWare IP, visit www.synopsys.com/designware.
Contact Synopsys, Inc.