IBM chip chief vows to stick to merchant path
[Capacity/300mm stuff toward the end --Duker]
A service of Semiconductor Business News, CMP Media Inc.
Story posted 3:30 p.m. EST/12:30 p.m., PST, 4/9/99
By David Lammers and Ron Wilson
EAST FISHKILL, N.Y. ( ChipWire/EET) -- The new chief of IBM Corp.'s Microelectronics Division vows to continue the unit's aggressive expansion into the merchant market by leveraging the company's vast patent portfolio and in-house mask-making capability, while limiting DRAM exposure and being more open to foundry deals.
John Kelly, successor to Michael Attardo as the division's general manager, made his comments in a wide-ranging interview, his first major one since taking the reins last month.
Attardo took IBM from a purely captive semiconductor manufacturer to one roughly evenly divided between commercial and in-house sales. The merchant effort, which started in 1993 with DRAMs and SRAMs, now provides ASIC and full-custom designs to high-end OEM customers.
Kelly, 45, said the Microelectronics Division stands to benefit from creation of the IBM Technology Group, a new unit charged with stepping up capabilities in displays, storage, printers and network hardware as well as in chips, and moving that technology quickly out to the OEM or merchant market.
"We are taking our merchant thrust to the next phase, to be a solutions provider in targeted markets," Kelly said. Those markets include servers, storage, wired networking, wireless communications and a grab-bag category of consumer opportunities that IBM calls "pervasive computing." Also, a separate packaging and module effort will try to expand IBM's flip-chip technology in the OEM space.
Jim Vanderslice, brought back to headquarters to run the IBM Technology Group after heading up the San Jose, Calif.-based Storage Systems Division, is bent on increasing OEM revenue quickly. That push aims to smooth what has been a difficult transition at many large, integrated companies: wrestling propriety ASIC cores out of internal divisions so they can be used in core-based designs and sold to OEM customers.
A few years ago, the various ASIC design teams within IBM all had unique tools and process tweaks. That has largely given way to a common tool set and interface standards. A design that originates in the Storage Division can quickly make its way to an ASIC created for a commercial customer, Kelly said.
"All of our internal divisions work from a common platform. With the formation of the technology group, we are all on the same network," he said. "We all get paid according to the performance of the group, and so we are a lot more open than we used to be." Dataquest Inc. ranks IBM as the 13th largest merchant semiconductor supplier in the world, with $3.234 billion in sales.
Last year's sharp drop in DRAM revenue, as well the effect of the Asian crisis on large chip buyers like Qualcomm Corp., caused IBM's commercial IC sales to decline by 4.6%.
Standing by memory
Kelly, who spent nearly all of his career in the Microelectronics Division after earning a doctorate in materials engineering from Rensselaer Polytechnic Institute in 1980, said he will continue to shrink DRAMs, taking memory manufacturing to "minimal" levels, but will not give up on making memories. IBM has worked with Siemens Semiconductor (now Infineon Technologies AG) and Toshiba Corp. in joint memory manufacturing, and Kelly said IBM has the ability to switch both its internal DRAM capacity, as well as some capacity with an unnamed partner, over to logic.
"I'm still concerned about DRAM prices. The decline has somewhat stabilized over the past few quarters, but there is some recent weakness," he said. "Logic stands to grow, and from an IBM Microelectronics standpoint, we know we are taking share in every major account that we have. We expect the logic business to grow very well."
Kelly said IBM is taking a new tack regarding intellectual property (IP) and patent protection. "We have received some very large payments for our IP. Fortunately, we have not had to go to court. We'll be leveraging our IP and patent portfolio, not just for cash, but for other IP. And we'll be trading for additional revenue opportunities for the division." He said the operation is "in a very strong position, accounting for a third of the patents received by IBM every year. We want to leverage that into future growth and access to other IP, as opposed to pursuing patent-infringement cases."
Though Kelly was, of course, bullish about IBM's growth opportunities, he said the company will try to pull different "levers" to expand logic capacity.
"In terms of investments, we are working aggressively to try and put the capacity in place for the growth we have seen and expect. We are taking a multi-pronged approach: making investments in logic ourselves, in [the] Burlington, Vt. [fab]. Our silicon-on-insulator technology is ramping up there as we speak, as is our 0.18-micron CMOS capability. We are also actively converting memory to logic at IBM fabs in a number of locations, and will continue to do that fairly aggressively. Memory for logic is a good trade-off."
IBM recently struck a memory-process-technology licensing agreement with Nanya Semiconductor Corp. in Taiwan, and has a logic-technology agreement with Acer, both of which could supplement IBM's internal capacity. And Kelly said IBM is talking to the established foundries.
"We have joint ventures with partners, each of which has their own capacity. If they are not able to use it, we take more of theirs. Another lever is foundry offloading, which may be through traditional foundries, but we also are going to leverage some of the technology licensing we did with companies in Asia to license capacity back to us. That will probably grow in the future. We always make our first investments here in the United States, close to our development facilities, then offloading, a year or two later, as we need."
Asked when IBM will build a fab able to process 300-mm wafers, Kelly said, "The first thing that has to happen is the overall industry has to return to healthy growth. The timing must be such that we stay on a productivity/cost-reduction curve, and until we see the industry really on a recovery path, on a growth path, most companies will choose to sit tight and wait. I think it will be another couple of years before we see any substantial capacity in 300 mm.
"We set up a 300-mm development effort at our Semiconductor Research and Development Center here at East Fishkill, and while that work started, it has been slowed down until we are sure the industry is ready to make the move as a whole. Our intent is not to be out in the forefront with this; substantially, we want to move with the industry."
|
