Qimonda: evolution of a DRAM vendor

An interesting press evening with Henry Becker, president of Qimonda North America, and Jan Dierens, VP sales and marketing at same, developed into a round-table discussion of the evolution of a DRAM vendor. Qimonda, formerly the memory division of Infineon, is coming off a strong quarter and the executives were entertaining questions from a gathering of analysts and reporters.

The company, like everyone in the DRAM business, is facing an uncertain outlook for 2007. On the positive side, the continued development of server farms in support of Web applications is providing strong demand. This is somewhat offset, the executives said, by what appears to be a wait-and-see attitude toward Microsoft’s Vista operating system. Corporate IT departments seem to be holding off on the major expenditures that would be required to move to Vista, so the operating system’s fate this year is in the hands of consumers. Consensus at the table was that Vista was unlikely to provide a boost in personal computer sales, but could cause a ripple of increase in average DRAM content per new system, and some upgrade business.

Better news for Qimonda is the company’s increasing penetration of non-computer markets, such as consumer electronics and mobile devices. This is good for a couple of reasons. First, it’s much nicer business than the commodity trading-floor atmosphere in the server and PC worlds, where price and volume are in play most of the time. In the consumer world, systems vendors tend to give their memory partners solid 13-week forecasts and good visibility into the more distant future–an unheard-of luxury in much of the memory market.

The second reason, more strategically important to Qimonda, is that many of these design wins are for specialty parts, not commodity DRAMs. This is the basis of Qimonda’s view of the future. “Our goal is to have our share of revenue higher than our share of bits,” Dierens said rather algorithmically. In other words, the margins are in the non-standard parts.

But then, the problems are there, too. Becker said that Qimonda has already concluded that they need a different process platform to serve the specialty markets than the platform they use in the server market. This has led to two very similar processes side by side within Qimonda: 75 nm for specialty DRAM and 80 nm for commodity DRAM. The equipment sets are similar so there is some capacity flexibility, but the two processes are not entirely interchangeable.

A more profound change is taking place in the structure of the company itself, however. Becker said that in the market today, “the more we know about the customer’s application, the better the result will be for them.” Far from stamping out JEDEC-defined parts, the DRAM business increasingly requires the DRAM vendor to understand the end-system and to participate in architectural decisions in order to reach the best design. For an organization that has been streamlined to serve commodity products to powerful buyers, rather than to joint-develop with design teams, this is a cultural shift of tectonic proportions.

“We have to find business models that let us understand the customer’ system design,” Dierens said. “I think over the next few years you will see the profile of Qimonda change.” Indeed, the company Becker and Dierens are describing—consultative in relationships, rich in systems understanding and increasingly application-specific in product address—sounds a lot more like an SoC vendor than it does like a traditional DRAM house. Whether Qimonda can make that transition, whether it can do so while retaining the cost controls necessary to compete in the still-vital commodity markets, and where the change will ultimately lead in a world that keeps sniffing at concepts like intelligent memory and backing away from them, remains to be seen. No question, thought, that it’s a company looking ahead rather than sideways.