The Process At TSMC That No One Talks About

We’ll probably never know the full extent of the story, but it appears the board of directors at TSMC didn’t just rubber stamp Morris Chang’s title when it named him chairman and CEO last month.

Chang was always chairman, but he wasn’t CEO. He had bestowed that title on Rick Tsai several years ago, who was Chang’s heir apparent. Tsai was named president of the new business development organization last month, reporting to Chang.

While Chang is almost synonymous with the creation of the foundry business, market founders don’t always end up as the long-term leaders. Steve Jobs at Apple is a rarity, and even he got pushed out of the company once. Most of the other companies that created a new business sector didn’t finish first, and most of the other leaders who started companies in the tech business—at least in the last several decades—didn’t survive.

The foundry business is a particularly tough one. Foundries talk about service, depth of offerings, established processes at advanced nodes and half-node jumps. But the bottom line is it’s still differentiated on price, which is measured in cost per chip, yield (a measure of the stability of processes), time to market for top customers and the ability to avoid respins. And for the most part, that business takes place at older process nodes where the factors are the only differentiators.

That’s at least part of what’s driving the restrictive design rules at the front end of Moore’s Law. Getting chips out the door, on time and in decent enough yields, is a marketing tool, if nothing else, and if restrictive design rules are necessary to accomplish that then foundries will insist on them. The other driver is lithography, which makes designs even harder to tape out successfully because the laser beams used to etch the silicon right now are much wider than the line widths between the wires.

So what really happened at TSMC? Consider the following series of events. First of all, TSMC is having yield problems at 40nm while its competitors are churning out chips at 45nm. Second, the Common Platform folks have 40nm technology ready and waiting when it becomes competitively necessary to offer it—namely when TSMC begins offering it. And third, TSMC is now feeling the pinch of the Common Platform’s growing ecosystem and SMIC’s rollout of 45nm process technology.

Add those three factors up and it’s no wonder that TSMC is changing its top management. The return of Morris Chang to CEO apparently wasn’t just an honorary title. It was a necessary change.

–Ed Sperling