October 26, 1998, Issue: 1132
Section: News
--------------------------------------------------------------------------------
Why Intel put money into Micron
Jack Robertson
Intel Corp.'s $500 million investment in Micron Technology Inc. is prompting a fair amount of armchair quarterbacking, as observers explore the range of market dynamics that such a large cash infusion could unleash.
The investment, which grants Intel the rights to 6% of Micron's stock, is overtly directed at winning Micron's support of Direct Rambus DRAM, an emerging architecture that Intel is positioning to serve the high-end PC market next year.
A Rambus licensee for some time, Micron nevertheless had kept its Direct RDRAM development program on a slow track. Because the technology is forecast to enter high-performance-computing markets first, the Boise, Idaho, company was content to watch from the sidelines until Rambus chip sales reached volume levels.
Micron's apparent changeof heart could be a boon for Intel, according to observers, particularly because at least one market research house had been predicting a serious shortage of Direct RDRAM chips in 1999.
In addition to shoring up the supply chain, Micron could apply its vaunted cost-cutting technology to drive down Direct RDRAM prices and speed the architecture's adoption across the lower-cost PC market, observers said.
As previously reported, 64-Mbit Direct RDRAM chips are expected to carry a 15% to 30% premium, which could discourage a quick market start. But with the industry's top five memory vendors solidly behind Intel's next-generation memory strategy, Direct RDRAM could spur yet another pricing war if any one company tries to grab early market share, analysts said.
What's less clear, according to observers, is the effect Micron's support of Direct RDRAM will have on the company's other new memory R&D efforts.
Some believe Intel took its equity stake, in part, as a preemptive strike against NEC Corp.'s Virtual Channel Memory (VCM), a lower-cost, high-speed DRAM core the company is proposing as an open industry standard for next year's mainstream PC market. Sources suggested that Intel wanted to nip VCM in the bud, especially since the technology has won the early support of Siemens Microelectronics Inc. and three Taiwan-based memory chipset makers.
Additionally, while VCM, as a DRAM core, could actually operate in conjunction with the Direct RDRAM memory interface, observers have speculated that Intel does not want its technology message confused by the introduction of another architecture.
Analyst Jonathan Joseph of NationsBanc Montgomery Securities Inc., San Francisco, said the 133-MHz speed of NEC's VCM doesn't approach the 800-MHz clock rate of Direct RDRAM, but could push back the market adoption of the Intel-backed wideband memory chip beyond Intel's planned 1999 timetable.
"The Intel investment brought Micron into the Direct RDRAM camp, but equally important, probably kept them out of the Virtual Channel Memory field as well," Joseph said.
Intel, sensitive to a pending antitrust probe by the Federal Trade Commission, has denied trying to influence Micron's pursuit of any alternative high-speed DRAM architectures. "Micron is free to develop any DRAM technology it wishes," an Intel spokesman stressed.
Indeed, Micron is expected to continue developing SLDRAM, a wideband-protocol memory that the company earlier championed as a competitor to Direct RDRAM. With a potentially lower cost structure, SLDRAM is aimed at price-sensitive PC and workstation/server memory markets, where Direct RDRAM initially may be out of reach.
Until the Intel investment, Micron's roadmap had only marginal funds available for early Direct RDRAM development and capital spending, according to observers. Now, Micron could follow a dual wideband memory path, aiming Direct RDRAM at high-performance desktop PCs and workstations, and SLDRAM at servers.
However, analyst Thomas Kurlak of New York-based Merrill Lynch & Co. Inc. doesn't believe Intel's cash infusion will do much to change the shape of the global memory market. "Micron already was upgrading the former Texas Instruments fabs and had adequate money for [capital expenditures at] its own [Boise] fabs," Kurlak said, referring to Micron's recent purchase of TI's memory operations. "Micron will continue its DRAM strategy regardless of the Intel investment."
Micron forecasts $800 million to $1 billion in capital spending next year, but most of this will be spent moving 64-Mbit chip production at its Boise fabs to linewidths of 0.18 micron and less. The company will spend another $1 billion over three years to convert three former TI fab complexes to 0.18-micron processes, but this will be largely funded from $750 million in heavily discounted financing obtained from TI and former TI partners.
If it is to enter the Direct RDRAM arena by the third quarter of 1999 as planned, Micron will need to draw on Intel's equity stake to make major capital investments, including the creation of a chip-scale packaging (CSP) operation, which is required by the new memory technology. Micron has a limited internal CSP effort for its fledgling SRAM and flash memory programs, but very little is tailored to DRAM.
Direct RDRAM will also demand new and costly testers to deal with the pipeline memory architecture, on-chip logic control, and chip speeds that are six to eight times faster than current 64-Mbit PC-100 devices. Micron said the Intel money will help it install Direct RDRAM production testers far sooner than originally planned.
Copyright ® 1998 CMP Media Inc.
|
