CEO, Alchimer, Massy France
The 22nm node might become known not for the technological breakthroughs it inspires, but for something decidedly less glamorous: diminishing returns.
At a high level, the transition to 22nm does not require any major new wrinkles; no big technical roadblocks need to be cleared (though there will be plenty of "small" issues that won’t seem small to the engineers who have to resolve them). And it seems clear that the elite group of chipmakers (Intel, TSMC, Samsung) have the resources to spend their way to success at 22nm, while smaller rivals should be able to identify profitable market niches and develop technology to pursue them.
But at this juncture, chipmakers large and small face a very tricky calculation – namely, are you better off going to market with a shiny new 22nm design, or with a stacked-chip approach using well-depreciated technology from a previous node?
Historically the math has always come out in favor of the most advanced manufacturing process, but the analysis today is not clear-cut, especially with 3D packaging technology about to come of age. Indeed, it seems likely that big, smart companies will probably come to different conclusions, depending on their markets, their competencies, and their relative optimism or pessimism.
Keep in mind that end users really don’t care whether their smart phone is powered by a 28nm chip, a 22nm chip, or a hamster running in a wheel, as long as it works. Memory manufacturers have a great history of achieving needed capacity levels with clever engineering – the solutions don’t have to be elegant as long as they get the job done.
The bottom line is that the traditional approach to running a semiconductor business is more open than ever to creative thought and new ways of looking at challenges, because the risk-reward equation is less clear-cut than it has been. In short, we are seeing diminishing returns, which will have an effect on the decision-making process.
And while money is always central to that process, time is nearly as important. Getting your product out the door quickly is an excellent offset to risk, so we can look for companies to look closely at new solutions that can be easily adopted and qualified.
In one sense, what we’re seeing today can be traced back to the memory bubble of the mid-1990s, when the industry’s risk-reward balance was amped up by a period of underinvestment, followed by the advent of Windows 95, which had much larger memory and processor requirements than earlier operating systems. DRAM prices and margins skyrocketed, and companies raced to build new fabs, some of which were ready for action just in time for the inevitable market downturn in 1997-98. Siemens, Motorola and Texas Instruments all threw up their hands and incurred major charges to exit the DRAM market, and the image of brand-new fully equipped fab buildings sitting idle was burned into the brains of every strategic planner from Frankfurt to Hsinchu.
In the ensuing period, there has been a winnowing of the wafer fab community and stratification into elites and others, as mentioned above. This is largely a result of new attitudes towards the delicate balance between opportunity and exposure – attitudes that are, in general, much more conservative than those of 20 years ago.
Looking ahead, there is every reason to think that the necessity of making a profit will drive the world’s remaining chipmakers to exceedingly rigorous analyses of the choices they must make, and the potential rewards, of new technology nodes. These are the sorts of conditions in which sacred cows can become an endangered species.
Yes, there will always be people driving around in Bugattis. But for the broad semiconductor industry and its supply chain, the diminishing returns of the 22nm node could prompt a time of reconsideration, and an embrace of creative solutions that work from a time and cost perspective.
Steve Lerner, CEO, Alchimer,S.A., Massy France, [email protected]