During World War 2, collaboration meant working with the enemy, a behavior that could land you in jail, or worse. In today’s business environment, the term has acquired luster, but collaboration still means working with competitors who, if not exactly enemies, certainly aren’t necessarily a firm’s natural allies. The companies involved in collaborative efforts are often in fierce competition with one another for the exact same customer base. As a result, the ability to effectively collaborate demands a corporate culture that can juggle two conflicting views of how businesses should inter-operate.
There are two primary collaborative models within the semiconductor industry. The first is horizontal, which is when two or more companies in the exact same business cooperate in order to solve an identical problem. The second is vertical, which is when two or more companies in different parts of the semiconductor industry cooperate in order to solve a shared problem.
The most visible example of horizontal collaboration in semiconductor industry is probably the Common Platform, an effort where IBM, Chartered and Samsung jointly developed digital CMOS process technologies and advanced manufacturing for 90-nm, 65-nm and 45-nm technology. It’s been a successful effort, attracting numerous implementation partners from the EDA, IP, and design services industries. Those partnerships, in turn, have allowed foundry clients the valuable potential to source their chip designs to multiple 300-mm foundries.
Horizontal collaboration takes place in the EDA sector, too. For example, Cadence Design Systems and Mentor Graphics have collaborated to devise a common approach to verification using the SystemVerilog language. The resulting Open Verification Methodology (OVM), based upon an existing IEEE standard, is intended to lower the development cost of the suppliers, give end users more options and capabilities, and foster a broader set of verifications IP suppliers, according to Pankaj Mayor, Cadence’s group director of business enablement.
The advantage to horizontal collaboration is that it allows companies to reduce the costs associated with solving knotty technical and marketing problems. In the case of the Common Platform all three companies were faced with excessive costs increases relative to researching new process technologies.
While the Common Platform has proven successful, there’s still plenty of competition going on between the participants, according to EDA analyst Gary Smith. “What is actually expensive is process research not process development, which is not outside the budget of the normal semiconductor company,” he explains. “Process development part of their competitive advantage and will continue.”
In other words, to make a horizontal collaboration profitable, the firms involved must continually find ways to differentiate themselves from the competition. The same is true of horizontal collaboration
in the EDA segment, according to Mayor, cites OVM as a perfect example. “The fact that the two companies are collaboration on a standard does not mean that they are no longer competing in the
verification space [and] of course the companies have their own goto-market strategies and differentiate on product offerings,” he says.
The vertical model faces differs from the horizontal model because the participating companies are not always in in direct competition. For example, when Synopsys works with Intel, IBM or TSMC to support the design rules require to achieve acceptable yields at 65nm and below, it’s a natural partnership between constituents, each of which has a good reason to make the partnership work, according to Kevin Kranen, Synopsys’s director of strategic programs. “As designers and EDA firms work with the foundries, we discover design rules to cope with different effects that occur in the individual manufacturing process,” he explains.
However, while it’s true that Synopsys does not compete with the foundries, because it has multiple vertical relationships with multiple vendors, the company must still manage the effects of competition
within the foundry segment. “Each of these foundries, even within the Common Platform, have variations in its manufacturing process that must be reflected in our tools,” Kranen explains. Because competitive pressure drives foundries to differentiate their offerings, the EDA firms must make additional effort to accommodate those differentiations, making each vertical collaboration more of a challenge.
EDA firms can also run into resistance from the foundries when the programmers need proprietary information about those key differentiations, according to Rich Goldman, vice president for strategic development at Synopsys. “The foundries’ ability to compete is tied up with their ability to offer processes that are superior to their competitors,” he told the author recently. “They don’t want too much information about those processes to bleed into the rest of the market.” This, along with the requirement to support multiple processes and versions of the same processes, creates additional overhead for the EDA firms, according to Kranen. “It’s a good thing that there’s a relatively small number of process that require sophisticated EDA,” he says.
Competitive pressures often insert themselves into vertical collaborations when those collaboration include multiple partners from the same segment, according to Mayor. He cites the example of the Design For Manufacturing Coalition (DFMC), which consists of “a growing number of companies (some who are competitors) are collaborating to address the challenges of DFM/DFY in a holistic
fashion.” Mayor also cites the Power Forward Initiative, which consists of 26 companies from the foundry, IP, EDA, ASIC, design services and semiconductor equipment categories, who have come
together "to enable the low power design market by using their collective expertise to devise new methods. See Figure.
Figure: The Power Forward initiative requires collaboration between companies from a wide range of industry segments.
Because collaboration models always have a competitive undertow, managers and engineers have to put aside rivalries and view their jobs differently, according to Synopsys’s Kranen. He observes that, to be a successful collaborator, a company must follow three key guidelines:
Effective collaboration requires an ability to envision another company as both a partner and a competitor, according to Richard Wawrzyniak, senior market analyst for ASIC and SoC at Semico Research Corp. “There are always problems because nothing runs the way you think it will,” he explains. “There are examples where partnerships have fallen apart because the companies decide that it makes more sense to go their separate ways.”
Competitive pressures can render a collaborative effort ineffective, sometimes disasterously. A perfect example of this is the 1990s alliance between MIPS, Digital and IBM to support Windows NT on RISC. Rather than working together, the three firms squabbled over few Windows NT users who were willing to experiment. The resulting debacle drove Digital out of business, damaged IBM’s credibility as a CPU supplier, and helped drive MIPS out of the workstation CPU segment.
However, Wawrzyniak notes that, despite the risks, the advantages to collaboration can be huge. Gary Smith similarly notes that the Common Platform has helped IBM remain, along with Intel, an industry leader in manufacturing. He notes that other foundries, like TSMC, are contemplating similar consortia in order to remain competitive.
Collaboration isn’t simple, but for the companies can manage the process and move forward, both as a partner and a competitor, the benefits outweigh the effort required to make such relationships work.
Geoffrey James has written over a hundred feature stories for national publications including Wired, Men's Health, Business 2.0, SellingPower, Electronic Business Computer Gaming World, CIO, ComputerWorld, NetworkWorld, BNET and The New York Times. He is the author of seven books, and a frequent contributor to publications from Extension Media.