Who’s Calling The Shots Now?

By Ann Steffora Mutschler
Determining who makes the decisions in semiconductor industry is not as easy as it sounds. There is not a straight line of responsibility in today’s market due to changing industry dynamics such as the shift from the IDM business model to the foundry model.

“If you go back far enough, everyone had to manufacture their own chips. There was a substantial influence from the process side, but very little from the marketing side,” said Wally Rhines, chairman and CEO of Mentor Graphics. “There were always three dimensions—manufacturing, design and marketing. Manufacturing had a great deal of influence over what got produced; design had a lot of input, but marketing had a lot less influence. As it evolved, though, manufacturing had less and less influence because the processes became more standardized. In the industry’s midlife, more emphasis was placed on design.”

That has changed significantly. Marketing is now calling many of the shots in everything from how products get distributed to what features are included. The iPod is a classic example because it used what was at the time fairly standard technology. What made the product successful was the marketing.

However, the older model that decision-making and influence evolves through a company as a function of size and tenure continues to be true, said Jack Harding, chairman and CEO of eSilicon Corp. “The young company that is looking for the break-out product is driven by the technology team. The slightly maturing company or the growth state company is driven by the marketing team. And the financial team drives the mature company. Against that backdrop, I do think there are some subtle changes going on that are noteworthy. It’s been my observation that in the semiconductor business, due to the cost of developing a single chip, which can be $50 million to $100 million, the marketing team is participating earlier than it once did.”

The reason is the high table stakes for creating a new chip. “You can’t afford to be wrong,” Harding said. “I’m observing much more marketing sophistication in younger companies. That’s driven partially by venture capitalists that have been burned on wasted projects, but also just the maturing of the industry which is now populated by a lot more 50-year-olds than 30-year-olds.”

He noted that when he entered this business, most CEOs were in the mid- to late-30s, “but now it’s an older, more sophisticated group that is working within the confines of an industry where growth is, on a secular basis, shrinking and the ASP pressures are increasing dramatically, so it circles back to that ‘you can’t be wrong’ problem.”

Similarly, Brad Holtzinger, VP of worldwide sales for MIPS Technologies, has seen a shift from engineering being the dominant force to marketing now having a more prominent role.

“In the past it was purely about speeds and feeds—engineering was focused on tweaking the microprocessor cores to eke out the highest possible performance. This was especially true when companies made all of the technology internally, in some cases even designing their own cores using an architecture license. As more functionality is incorporated on today’s silicon, companies must buy more of their IP and focus on integrating it. This is where marketing gets more involved because the IP and software ecosystem gains such importance,” he said.

What happened to the manufacturing influence?
“In manufacturing, half the large chipmakers are fabless or fab lite these days,” Rhines said. “Memory companies are still true IDMs. Samsung and Intel still manufacture their own chips, while Qualcomm and Broadcom have no manufacturing at all. But even though the foundries have a great deal of influence on how things get built, the processes determine what can be designed and what companies can sell. There isn’t much flexibility of choice. Once a company locks in a design with a foundry, they usually lock in most of the derivatives, too.”

Hans Stork, group VP and CTO of the silicon systems business unit at Applied Materials, has a unique perspective on the shift to the foundry model, as he served as CTO and senior VP of silicon technology development at Texas Instruments when the company decided to outsource its process development (prior to joining Applied).

“When TI made its big strategic change to largely depend on the foundries for advanced logic technology three years ago, I often got the question, ‘So you want to be totally subject to foundries in terms of decision-making?’ My answer was, ‘No! We have so much volume that they will listen to us.’ If your biggest customer is 20% of your revenue, that’s what you do.” Essentially, this model puts the manufacturing in a follower, rather than a leader, role.”

“With so many products built in the foundries, and the foundries not being captive to the system or the component design houses, there is more of a distance, so they won’t have as much to say in terms of direct interaction,” Stork continued. “At the same time, you look at big companies like TI, Qualcomm and Broadcom, they have very dedicated teams even on the process level working with the foundries to ensure that they get what they need.”

Without a doubt, given that there are so few IDMs left, that fact has changed where decisions happen. And while it is particularly true for digital and consumer logic, Stork reminded that in the analog world, it is not at all the case. “Analog still very much drives by strong process-design interaction defining the features of the chip,” he saidd.

This marketing-driven trend is likely to continue, and possibly even expand.

“I see a growing and more critical role for marketing even in the youngest of companies,” Harding said. “The No. 1 reason why we see a chip we’ve developed with a customer fail is because they’ve over-featured it. They want to be all things to all people. You take that omni-functional part into the marketplace, inevitably there’s a market that says, ‘we want half of the features, not all of them.’ Now you’re trying to sell them a chip that’s twice the size that’s populated with features that are irrelevant to that particular marketplace.”

That’s particularly true in technology-driven companies where engineers put every feature on to the chip because it can be done and because they don’t understand what the market will want in the future.

“At 65-, 40- and 28-nm you’re going to see a much smaller appetite from the investor community for that type of ‘build it and they will come’ mentality,” said Harding. “The emphasis will be put on a marketing team that has the knowledge and wherewithal to limit the feature set to exactly that what is demanded, and thereby keep the waste to a minimum.”

–Ed Sperling contributed to this story.