BOSTON — The private-equity buyout movement is just one sign of a tectonic shift in the semiconductor industry as businesses that have chased Moore's Law now grapple with "technology overshoot," two Harvard professors said last week.

Clayton Christensen and Woodward Yang told EE Times that designers deliver more technology than consumers can figure out how to use, forcing businesses to adjust from stem to stern—from how they design and manufacture devices to how they service customers and structure their corporations.

"There are some people seeing that just relentlessly pursuing Moore's Law is not giving them the return on investment they had previously seen," said Yang, an entrepreneur and professor in Harvard's Division of Engineering and Applied Science. "With the maturation of the semiconductor business, you're seeing other aspects . . . that are required aside from technology, things like finance, long-term investment, customer relationships."

Yang said technology overshoot happens, for example, when people no longer line up outside retailers to get the first new computer systems with the latest rev of an Intel processor.

"We build faster microprocessors and we're not sure why we want a faster one anymore," Yang said. "When that happens, people look not to the technology but other methods of trying to create value for the end customer."

Christensen, a Harvard Business School professor who has written best-selling books describing the effects of disruption—The Innovator's Dilemma and The Innovator's Solution—said the semiconductor industry is slowly beginning to look toward ways of operating in a new world.

To IPO or not?

The private-equity phenomenon is part of this transition. In large part, the movement of companies to be bought and taken private has arisen because capital is shifting toward investments with better returns. The traditional path of a startup--garnering millions in venture-capital money to get started, and then cashing out in an initial public offering after a few years--is largely closed off.

The cost for a startup to create an IC can be about $50 million. A 5x or 10x return on that investment takes much longer than it once did, and there must be a roughly billion-dollar end market to go after, assuming there is competition at the IC level. There aren't a lot of new billion-dollar markets popping up these days.

"The exit strategy for VCs has changed," Yang said. "The cost of doing an IPO has gotten quite a bit higher. [A private-equity exit] is a less-risky strategy, and you know whom you're dealing with."

In addition, there is so much late-stage venture money available now that the price differential between public equity and private equity is almost erased, Christensen said. "In many ways, the private market is disrupting the public market," he said. "Companies that used to need to go public [now say], 'Let's not go public. Let's bring in some private money.' "

If the financial structures and investor profiles are changing, so too is the very nature of what's considered valuable. In an emerging sector, quality is defined generally as functionality and reliability. But as a sector matures, quality is redefined as speed to customers, convenience and customization, Christensen said.

"What that means is that the kinds of innovation that will help you get traction and merit attractive pricing in the market are innovations that help you be faster in giving each customer exactly the circuitry they need," he said. "When other industries go through that transition, it's wrenching."

Faster, different

That means the entire industry, which considers itself fast-paced, needs to speed up some more--but in a different way. Christensen sees a world in which designs will broaden dramatically, but the volume per design will drop. Circuit design needs to become faster too, drawing on established libraries and a process in which components are assembled in modular fashion "so you can start to mix and match to be fast and flexible and responsive," he added.

Christensen and Yang have been working with industry leaders for more than a year on a road map for future electronics industry business models. Part of their work involves examining manufacturing processes with an eye toward figuring out how they can be speeded up.

"The old world, where a wafer could take 90 days to meander through a fab, worked when quality was defined as functionality," Christensen said. "But when quality is defined as 'give me exactly what I want exactly when I want it,' then a wafer has to get through a fab in a week."

The semiconductor equipment industry, then, needs to shift its focus from staying on the hairy edge of process technology and lithography to reliability and uptime. "You'd love to have the downtime be predictable," Christensen said. "Then you don't need to have wafers work-in-process in the flow to buffer unreliable equipment."

Mind-set shift

If the means to change 50-year-old habits in the electronics industry seem remote, the unimaginable has already occurred, Yang said.

"Intel sold its applications processor business to Marvell. Unthinkable," he said. "Intel entered the NAND flash business. Unthinkable. The Japanese have basically exited the DRAM market. Freescale was bought by private equity. There's something going on here."