There is growing agreement among the informed nervous that this is a bubble getting ready to burst. "I give it two to three years," Goertz said.

The original reason for design outsourcing was cost and availability. Huge U.S.-based integrated device manufacturers (IDMs) set the ball rolling, according to Rakesh Kumar, president of Electronic Integration Inc., a fluid virtual-ASIC company. Looking into India and Taiwan, the IDMs saw a growing pool of well-trained young engineers whose salaries and overhead were much lower than they would have been in the United States. Rather than move those designers to America, fighting for visas and creating higher costs, the IDMs created centers of expertise where the engineers were. They often seeded the design center by offering incentives to some of their own senior people to move to India or Taiwan to anchor a group of recent grads-so-called freshers. Then, having defined the skill set for the outsource group, the IDM could modularize designs so that a self-contained set of tasks could be dispatched to the remote facility, with a clean interface.

Werner Goertz sees a gold rush for designers in India.

But the nature of arbitrage is to restore balance to an unbalanced situation. Sure enough, design costs in India are rising. "You can see the trend already," Goertz said. "Salaries are going up; rents are rising. Design costs are moving toward parity with the U.S."

Observers on the U.S. side agree. In fact, the massive job losses in the United States have created a pool of skilled design contractors who are willing to bid aggressively on projects. "The sweet spot for component-level design is still North America," said Phil Heine, director of strategic business development at Avnet Cilicon, the design services arm of Avnet. "We have seen a number of design-outsourcing contracts go out for bid recently in which the low bidder was a North American group. Many of these groups are in fact design teams that were laid off by systems houses during the recession."

Some outsourcing relationships have shifted away from the focus on cost savings toward a strategy more like that of the initial IDM approach. Design teams in North America, stripped and battered by cuts, are reaching out not for cheap labor but for teams with particular skill sets: algorithm or software development, RTL generation, block-level verification or physical design.

This approach emphasizes longer-term relationships between the core design team and the outsource shops. It also stresses project management at a distance of partitioned designs.

Charlie Cheng: Break the design up at traditional sign-off points.

But working with an ASIC vendor is not the only way to outsource physical design. There is a prejudice that design teams outside the United States are most capable of the early stages of the flow, say, through RTL verification. But this is a poor generalization now, said Goertz. He pointed out that the IDMs have for some time emphasized building physical-design expertise in their non-U.S. design centers, creating a pool of senior people who can be the core of new physical-design groups.

The partitioning can also be done along logical or physical boundaries: An outsource group may be brought in to do the complete design-and perhaps even integration-of blocks for which the local design team lacks resources or expertise, said Steve Kawamoto, ASIC-marketing manager at NEC, which sees outsourcing approaches used so frequently now that it is building a network of certified third-party design centers to serve its ASIC customers. But Kawamoto cautioned that in this sort of partitioning, success depends on clear division of responsibilities, clear expectations and ruthless management of the change process.

There is no clearer example of that thinking than Open Silicon, a global fabless ASIC company spun off from what was Intel Microelectronics, the ASIC arm of the processor giant. Open Silicon president and CEO Naveed Sherwani described a company that was built on a foundation of outsourcing.

Naveed Sherwani: India has necessary cost structure.

The first aspect was the more obvious. "To live on much smaller margins, we had to reduce our direct-labor cost per chip from about $450k to $100k. That necessitated the kind of cost structure we could only find in India," Sherwani said.

But the second was more strategically significant and may well survive the drift to labor-cost parity. Open Silicon's business is build on a fixed methodology: a Magma-Synopsys flow that is not altered for special circumstances.

This methodology is partitioned into fixed discrete tasks that are tracked by an elaborate work-flow system, enforced by a project manager in the United States. The tasks are split between the customer and Open Silicon, with the back-end design further partitioned among seven separate expertise groups in India. Each of these groups, in turn, is headed by a senior designer, usually from the original Intel team. The core designer is surrounded by a cluster of engineers hired from college.

The entire system-from recruiting to formation of the seven design teams, the pipelining of designs through the teams and the open, database-driven task tracking on an individual level-is tailored specifically for Indian culture and the Indian education and employment scene. "Frankly, if I'd had my choice of any place on Earth to implement this flow, it would have been in India," Sherwani said. "It fits the culture perfectly."

This may be the model not just for success but for survival. Forget exploiting low cost or farming out trivial jobs; design globalization is not about colonialism. Rather, build a design flow around the long-term skills that lie in North America and elsewhere. Explicitly create a partitioning and project management style that can operate efficiently across cultural, spatial and temporal divides. And stay focused on the strategy.