These convergent forces are driving nearly everyone to seek business alliances with specialized third parties. Nowhere is this new reality more evident than in the tripartite relationship that today exists among the chip designer, the third-party library vendor, and the semiconductor foundry. But how do all the parties build a sustainable relationship where the models are self-reinforcing?

We at TSMC believe that the ideal business model for the alliance is a "pay-for-performance" model that enables the chip designer to reach volume device production in the shortest period of time. That means creating a business model that makes libraries available early in the technology process development cycle.

Promise or performance?

Until last year, the relationships among the three parties were fairly linear. The designer worked directly with a vendor who licensed a library to the designer for a set fee. Often, the chip designer ended up spending several months integrating the third-party libraries into the design flow, spending almost as much money on this process as on the design itself, and often delaying the completion of design tape-outs. Once it had licensed the library, the library vendor had little incentive to provide technical support or to further the integration process. The designer would then turn the design over to the foundry, with the customer/vendor relationship continuing down its narrowly defined path.

But the complications of deep-submicron designs have changed all that. Today, the relationships among the three parties are, by definition, much more collegial. The win-win business model is inadequate under this paradigm, making the win-win-win business model essential. Enter pay-for-performance.

The objective of the pay-for-performance model is to ensure that all parties look at a long-term, volume-based goal. When a designer enters a long-term commitment with the foundry and library vendors, it makes the foundry and the library vendor an integrated extension of the designer's manufacturing and delivery operations.

In this model, third-party library vendors create and preverify libraries in parallel with process technology development. This shortens design time and tunes a library to a specific wafer process. Such up-front collaboration paves the way for chip makers to move completed designs into production. The foundry can simultaneously promote quality standards to increase confidence in the integrated design flow.

Let's work together

Under pay-for-performance, the chip designer still works directly with the third-party library vendor for licensing and technical support, while the foundry works to ensure the availability of a wide variety of libraries at the earliest stage of development. The foundry assesses each library vendor's offering then provides chip designers with a listing of value-added, process-specific library options that best fit their particular implementation.

In addition, the foundry can prequalify the library portfolio, narrowing the choices down to three or four options. Once the chip designer selects a library vendor, the foundry facilitates the project to maintain schedules and ensure support. Most important from the library vendor's perspective, the foundry would manage and monitor a royalty program, which funds the library vendor's growth and promotes a relationship between library vendor and the foundry. While the foundry might manage business and technical programs with the library vendors, each library vendor determines its own business model (such as licensing, maintenance, or NRE fees) with the designers.

For the chip designers, the model creates highly motivated library and foundry partners. The foundry receives a steady flow of wafers to fill the fabs. But more importantly, the semiconductor industry benefits from a continuing stream of leading-edge designs.

Kurt Wolf is director of marketing at TSMC North America. He previously worked for 12 years at Advanced Micro Devices.