By Phil LoPiccolo, Editor-in-Chief, Solid State Technology
While the semiconductor industry continues to promise ever higher rewards — with compound annual growth rates in the high single digits for many years to come — few companies will be able to survive the increasing risks of competing in the market unless they form novel heterogeneous manufacturing alliances, similar to the research partnerships that have proven successful in managing escalating R&D costs. That was the message delivered by The ConFab’s keynote speaker Ajit Manocha, EVP and CMO of NXP, who kicked off the event on Monday and set the tone for three days of top-level executive discussions on a range of issues facing the semiconductor industry.
In his opening address, Manocha painted a picture of a rapidly expanding semiconductor market expected to nearly double in revenues this decade, from $204 billion in 2000 to $400 billion in 2011 (according to data from IC Insights and World Semiconductor Trade Statistics). “Not too many industries can provide that kind of growth,” he said, adding that among factors fueling growth is the fact that the amount of semiconductor content in electronic devices will also have risen dramatically during this period, from 10.7% in 1990 to 24.5% in 2011 (see Fig. 1, above).
Another trend that bodes well for the future is that the industry is becoming increasingly segmented, Manocha noted. There are more than 300 distinct markets, including general-purpose sectors such as memory, MPUs, optoelectronics, discretes, and analog devices, which currently make up about two-thirds of the market. Application-specific sectors such as wireless and wired communications, data processing, consumer, and automotive segments account for the remaining third (see Fig. 2, below).
On one hand, this kind of application fragmentation is a positive growth factor because it means that there’s a great deal of innovation occurring, Manocha said. However, it also presents challenges to manufacturers in that it means smaller product volumes and, in the consumer era, shorter product lifecycles. Moreover, there’s a great deal of fragmentation in terms of the sheer number of competitors, many of whom account for a relatively small percentage of industry revenues. Of more than 150 chipmakers, the top 50 garner 75% of total revenues, while the middle 50 account for 20%, and the bottom 50 account for only 5%. And of all manufacturers, only three companies have annual revenues greater than $10 billion. Moreover, the same patterns hold true for both foundries and semiconductor equipment suppliers. Given this landscape, “textbooks would predict consolidation” across the supply chain through acquisition of smaller- and medium-size companies, Manocha said.
A host of other challenges, most of which have arisen only in the last decade, also threaten all but the largest firms, Manocha pointed out. Chief among these is the rising cost of fabs. In the 1990s, 200mm fabs cost about $1 billion, he noted, but today a new 300mm facility costs some $5 billion. Beyond that, prices for equipment are soaring; EUV scanners for 32nm technology, for example, will cost about $40 million apiece, he said. Also, today’s processes entail high costs associated with high-k and low-k material integration, software development, intellectual property protection, product design, test, DFM, and DFX — factors that did not exist in the last decade.
R&D and design costs have also skyrocketed in the past decade. In 2000, total R&D expenditures were about 12.7% of total industry revenues; that is seen topping 17% in 2007, and reaching nearly 20% of revenues by 2011. Moreover, if the development cost of a single 65nm or 45nm product is tens of millions of dollars, and if the revenues over the life of the product need to be 10x the R&D investment to achieve ROI, many products will have to generate revenue streams of hundreds of millions of dollars, hardly trivial in a fragmented application space.
Added to these challenges, the semiconductor business cycles will continue to undergo extreme fluctuations. Some reports claim that the demand swings will be less pronounced than in the past, Manocha said, but while it’s true that the semiconductor demand curve will now begin to track the same general pattern as the electronics equipment growth, the industry’s swings will still be an order of magnitude more pronounced than the fluctuations in the electronic industry, as well as in the world’s GDP.
Toward a new manufacturing model
So what’s the best strategy for survival in this complex semiconductor ecosystem? For one thing, it’s clear that “companies can no longer afford to play alone,” Manocha said. In fact, NXP has recently adopted an “asset-light” foundry model under which it will outsource all advanced CMOS chip production beyond 90nm to foundry giant TSMC. The foundry partnership addresses Manocha’s bottom-line concern that manufacturing has become extremely difficult with respect to capital intensity. Indeed, it may be possible for mid-tier manufacturers to raise $5 billion to build a 300mm, 100,000 wafers/month fab to compete against high-volume chipmakers bringing in >$10 billion a year, he said — but when medium size companies cannot fill 300mm megafabs with product, it will lead to massive losses that they don’t have the depth of resources to absorb.
While NXP is driving the homogenous foundry partnership model with its asset-light strategy, Manocha’s personal view is that in the coming decades this kind of partnership will evolve into a heterogeneous model to include many more players, including not only foundries but also other IDMs, fabless companies, materials and equipment suppliers, hardware and software designers, EDA firms, and even assembly and test companies and other elements in the IC supply chain.
Manocha’s vision is that this heterogeneous approach will become a complex business model that extends the type of alliances that have proven successful in managing R&D costs into the realm of manufacturing. In R&D, forming alliances among diverse players to share the costs makes sense, he said, and in manufacturing it will make even more sense going forward.”
In one possible scenario, such collaboration could take place between IDMs, fabless companies, and foundries, so that the IDMs and fabless companies could provide product designs and product load, or demand, for the foundries, while the foundries could provide process technology, Manocha explained. The partnership could also extend to equipment suppliers to provide the tools and ensure that they are performing to the best of their ability, and even to leasing, venture capital, and private equity companies to provide funding.
An added advantage from this kind of model is that a balanced portfolio of companies working together can complement each other to offset business cyclicality. A manufacturing alliance with a microprocessor manufacturer, a memory company, and several niche product players, for example, could counter demand cycles so that if one business were down one quarter, others might still be strong, so demand as a whole would not fall below the break-even point, Manocha explained. In addition, the participating companies also can benefit from economies of scale. NXP’s experience with joint ventures has been that partnerships are more effective in limiting utilization swings, and hence show better profitability over time, he said, pointing to TSMC with its staggering profit margins as a prime example.
The overall message is clearly written on the wall: “It’s time for a paradigm shift in the manufacturing business model to bring together heterogeneous companies to share the risks and potential rewards,” Manocha said. If you don’t share the risk, there’s hardly any reward, he said, but if you do share the risks, and if the alliances are managed and nurtured properly, the result can be highly rewarding for all members. — P.L.