Hitachi revamps DRAM strategy
By Robert Ristelhueber
EE Times
(09/22/99, 2:18 p.m. EDT)
SAN JOSE, Calif. — With the horrendous, only recently ended meltdown in memory chips still fresh in its mind, Hitachi Ltd. has decided to sharply pull in its horns and adopt a high-end strategy to cut its exposure to any future crashes. The conglomerate has reorganized and refocused its DRAM efforts on the high-end workstation and server market, skipping a 128-Mbit SDRAM and planning rollouts of double-data-rate SDRAM and Direct Rambus parts for late this year.
"We believe we've defined a very tight, crisp focus in DRAMs," said Ron Bechtold, vice president of sales and marketing for Hitachi Semiconductor (America) Inc.
That is reflected in a dramatic shift in Hitachi's DRAM end markets. In the second quarter of last year, 66 percent of its DRAMs went to the PC industry, and only 10 percent to the server/workstation segment. By the second quarter of 1999, that situation had reversed itself almost completely: only 21 percent of its DRAMs went into PCs while 23 percent went to workstations, and 37 percent to servers.
Although shifting from PCs has cost it market share in the overall DRAM business — Hitachi now has only about 7 percent of the world market compared to about 16 percent a couple of years ago — it has helped restore the company to profitability.
Hitachi has adopted a strategy along the lines of other Japanese companies, including Toshiba Corp. and Fujitsu Ltd.: narrow the memory product focus, and increase the accountability within the group. Before this past March, DRAM design, production and sales was reporting into a memory group that also had responsibility for other chips including SRAMs and flash. "Nobody was responsible for he health of one product," Bechtold said. Today, all those functions report to a dedicated DRAM division.
Measures of success
Part of the high-end strategy has been to push the technology envelope in processes, architectures and packaging, he said. Hitachi probably has a greater percentage of its DRAMs being built on a 0.18-micron process than any supplier, and has industry-leading soft error rates, Bechtold said.
The company has been out front in the development of stacking techniques such as tape carrier and double-density packages, he said. It also has been pushing double-data-rate (DDR) parts and will begin sampling a 256-Mbit device in the fourth quarter based on a 0.18-micron process. Current plans call for sampling a 128-Mbit DDR device based on 0.18-micron technology late in the first quarter next year, and sampling a 0.15-micron 256-Mbit DDR device in the third quarter of 2000.
Just deciding to stay in DRAMs wasn't a simple matter. Hitachi Ltd. lost $2.8 billion in fiscal 1998, and half of that was attributed to the semiconductor group, which was heavily weighted in DRAMs. But in the end, it was determined that DRAMs are an important driver of process technology and manufacturing efficiency, Bechtold said.
The company feels good about its decision to skip the 128-Mbit SDRAM generation, believing that it would be a short-lived product. It also has been holding back on development of Direct Rambus chips until that market sorts itself out. It expects to begin production of a 144-Mbit Rambus chip in the fourth quarter, with a 128-Mbit device to follow in the first quarter of 2000, he said.
Hitachi and NEC Corp. are still hashing out the details of their DRAM collaboration, announced in June. To cut costs the companies are expected to form a joint venture to handle future development of DRAMs.
eet.com |
