The changing foundry model
10/01/2003
Solid State Technology asked an analyst and an IC manufacturer for their take on the current foundry business model.
Strategic partnerships: Critical to foundry success
Bill McClean, president, IC Insights, Scottsdale, AZ
The IC foundry market will exhibit high growth over the next few years; however, the increasingly long list of new entrants into this market will make it very difficult for the participants to achieve a high level of profitability. The foundries that are more profitable will be those that stay at the leading edge of the process technology roadmap. Joint ventures, such as the recent partnership between IBM and Chartered Semiconductor, and/or significantly increasing R&D spending, could accomplish the goal of staying at the leading edge. Even being a half step behind in the technology race can spell financial disaster for a foundry.
More and more, the business landscape in the foundry segment is beginning to resemble that of the DRAM market in the 1980s. While the trend toward outsourcing IC fabrication to foundries is certain to continue, the expanding number of companies chasing a maximum of 100 points of market share almost assures an eventual competition-induced company shakeout.
The emergence of pure-play foundries will continue to have a major impact on future IC industry production. The 10.3 million MOS wafer capacity level of the pure-play foundries represented 17% of the total worldwide MOS capacity available in 2002, up from 9% in 1998. In 2007, IC Insights expects the pure-play foundries' share of total worldwide MOS capacity will increase to 30%. Furthermore, the total share of MOS capacity represented by the foundries will accelerate over the next few years as more IDMs (e.g., AMD, ST, Motorola, Toshiba, etc.) increase their outsourcing of IC production.
In the final analysis, the IC foundry business is going to continue to be an important and growing portion of the future IC industry — it is also going to be very competitive. Established pure-play companies like TSMC, UMC, and Chartered Semiconductor will be aggressive in protecting their market shares. Newcomers, such as 1st Silicon, Dongbu, Silterra, IBM, Powerchip, as well as the Chinese foundries (e.g., SMIC and Grace), are going to be equally aggressive in attempting to secure large customer bases.
While the Chinese IC market (as opposed to production) already represents a significant and growing share of worldwide IC consumption, it may be a very different story when measuring the success of China-based IC production, including the new foundries.
Given the limited capabilities of China's indigenous design houses, it is IC Insights' opinion that the near-term (i.e., through 2005) success of the Chinese IC companies such as SMIC and Grace is almost entirely dependent on their ability to take market share away from existing IC foundries (e.g., TSMC, UMC, IBM, Chartered Semiconductor, etc.). The difficulty these Chinese foundries are likely to encounter in attempting to gain significant IC foundry market share cannot be emphasized enough.
The most critical issues with respect to a foundry's future success will be its ability to develop long-term strategic partnerships with its customers and leading-edge processing technology. If a foundry can establish significant supply relationships with a few large IDMs (AMD, Agere, Motorola, etc.), as well as with some of the large fabless houses (Qualcomm, Nvidia, Xilinx, etc.), and keep at the leading edge of IC process technology, it should be able to survive the cyclical downturns of the IC and foundry industries and flourish in the cyclical upturns. By 2010, there will only be five or six foundries, out of the 20 that currently exist, to have successfully met this challenge.
The IC foundry business is going to continue to be an important and growing portion of the future IC industry — it will also be very competitive. |
For more information, e-mail Bill McClean at [email protected].
90nm and beyond: Why go it alone?
Kevin Meyer, VP, Worldwide Marketing and Services, Chartered Semiconductor Manufacturing, Singapore
The advances of process geometries and accompanying technology have generally been taken for granted. Each progressive generation prior to the 130nm node has been reached in a relatively orderly fashion. It's been a predictable progression from 1.0µm down to 0.15µm, and extrapolations have predicted the same steady progression for nanometer-level efforts.
The 130nm node, however, presented a bump in the process roadmap. Multiple variables, introduced by the transition to copper interconnect and the development of low-k dielectrics, compounded with issues related to smaller geometries, have kept some companies stuck at the node much longer than expected. We've seen the advertising "air wars," but the lessons learned at 130nm demonstrate the risk of making false assumptions based on marketing hype. Today, the lingering implementation challenges have caused the entire semiconductor industry to question whether it has reached an inflection point where the rules must dramatically change. Indeed, some question the viability — technically and economically — of 90nm.
For the record, 90nm — and beyond — must and will be achieved, particularly from the standpoint of inspiring new design activity and furthering the convergence of applications onto system-on-chip (SoC) solutions. But the questions are: At what cost and what impact to the traditional approach for semiconductor design and manufacturing?
Collaboration is now perceived as the key to bringing advanced technology nodes to market. Companies expecting to spend as much as $30 million for the development of their 90nm SoC designs will look to credible foundry partners who offer shared process platforms borne of true collaboration to ensure the predictability we saw with earlier generations. With masks costing upward of $1.3 million at 90nm, according to industry estimates, companies cannot afford multiple spins in getting a design "right the first time." They will also want to spread the risk and cost among all those that will benefit from the next generation.
The cost of developing and implementing at 90nm is drawing together what, a few years ago, might have been considered strange bedfellows. Only a select few companies are cautiously considered as having the R&D skill and practical partnerships needed to assure a robust 90nm process platform.
While collaboration is not new in technology circles, Chartered Semiconductor believes today's technology development issues require an even more intertwined strategic engagement model that involves multiple industry players. We're starting to see this with the emergence of clusters of companies or semiconductor constellations focused on 90nm initiatives. Chartered's own agreement with IBM to partner at 90nm and 65nm is a reflection of this trend. In this example, two companies that would seem to be competitive in some respects — both offer outsourced manufacturing services — are joining forces to solve the tough 90nm challenges. Other foundries have aligned themselves in similar fashion with IDMs and system houses.
Collaborative constellations, made up of fabless semiconductor companies, IDMs, and foundries, are a convergence of the perfected skills that each brings to the group. IDMs bring their long history and expertise in technology development, while fabless semiconductor companies bring their focus on product design. Pure-play foundries, which have an increasingly more significant role, lend their strong capabilities in process manufacturing and expanding leadership in advanced process development.
With more tightly coupled collaboration, there is also a growing recognition that openness and standardization are essential for success, perhaps even survival. The pure-play foundry industry and its constellation partners must acknowledge that it is no longer just about process, but is more about offering complete solutions with open access to design platforms (silicon-validated) enabled by third-party library, IP, and design tools/services. Because companies need more control over their design and manufacturing options, these platforms must ensure true design portability rather than lock customers into proprietary, inflexible solutions.
Going forward, business and technical standardization efforts driven by industry groups such as the Fabless Semiconductor Association, as well as those within and between emerging semiconductor constellations, must preserve customer choice. The goal is to provide foundry customers with proven paths from concept to silicon, while ensuring maximum flexibility, freedom of choice, and economies of scale. A greater sense of this as a shared mission from the boardroom, to R&D labs, and in accounting departments, will drive results. The benefits are obvious, and they go a long way in helping companies deliver better, faster, and cheaper products to market.
Clearly, the rules are changing at 90nm and beyond. Much more cross-pollination of people and ideas must happen between companies that may even consider themselves direct competitors. And in the end, those companies will owe their success not only to their technical prowess, but also to their skill in collaborating.
Collaboration is perceived as key to bringing advanced technology nodes to market. Openness and standarization are essential for success. |
For more information, contact Kevin Meyer at Chartered Semiconductor Manufacturing Ltd., 60 Woodlands Industrial Park D, Street 2, Singapore 738406; ph 65/6362-2838, [email protected].