Intel Faces Critical Delays Thanks to Supplier's Failure
By Jim Seymour
Special to TheStreet.com
5/10/01 3:39 PM ET
JP - JS gives an example of how money can't always buy you fab-lov'n. This from the bottom of the article sums it better than a roadkill commentary - d:
"Increasingly, Intel looks less and less like a high-tech star and more like Jimmy Breslin's The Gang That Couldn't Shoot Straight."
And when it rains, it pours...
eWeek (formerly PC Week), the top trade paper for IT managers, is out with a story about a major stumble for Intel (INTC:Nasdaq - news - boards). This could be important for investors ... as well, of course, for Intel customers that are original equipment manufacturers and for other Intel constituencies and stakeholders.
Basically, a key piece of machinery required to make the next generation of Intel CPUs -- chips due to customers this summer and fall -- was supposed to be delivered by Intel's supplier in April. Not only was it not delivered then, its maker is now saying it hopes to have it ready "sometime in the fourth quarter."
Badda-bing.
There's more to the story than that, of course, and you need to know more to assess the likely impact of this delay for yourself. It takes a little background to understand this, so bear with me as I work through it. I'll take out as much of the techno-tedium as I can, I promise.
The big production revolution in semiconductors over the past few years has been the move to so-called "sub-micron" technology. The term refers to the width of the tiny, signal-carrying electronic "wires," or more properly "traces," on the silicon base that forms a chip. By moving to ever-thinner traces -- now well into the thinner-than-a-micron range -- semiconductor manufacturers have been able to make far more dense chips, which draw less power, and run faster ... and, critically, cooler.
The current wave of production in key chips, such as CPUs (Pentium, Athlon, etc.) is handled with .18-micron technology. At this size, a trace is about 1/200th the thickness of a human hair.
But moving to the next-generation CPUs, such as Intel's Tualatin, and especially to the eagerly anticipated market-leadership product 2 gigahertz Pentium 4 expected in the second half of this year, requires .13-micron technology. Even Intel's 1 gigahertz Mobile Pentium III CPU, already in short supply due to low yields (the ratio of working, salable chips available from the total number of chips on a disc of silicon), is, as eWeek points out, pushing the envelope on .18-micron technology.
So Intel has been gearing up in a big way for moving to .13-micron technology for its new products. But the key machine for producing .13-micron silicon for Intel is the Micrascan V, which comes from the Silicon Valley Group (SVGI:Nasdaq - news - boards). The very high numerical aperture (VHNA) step-and-scan Micrascan V is one of the wonders of the semiconductor-manufacturing business.
Fellow Mark Bohr, director of process architecture and integration at Intel, has said often that the single biggest problem for the company in producing better, faster chips is lithography -- the system of "printing," or "depositing," those narrow traces on silicon.
If it sounds odd to you that the gating factor for semiconductor improvements is something as basic as what amounts to a high-tech variation on ordinary offset printing ... well, truth is stranger than fiction. Bohr is absolutely right: We are now treading on the edge of physics with .13-micron traces, walking the line between science and science fiction.
The Micrascan series of chip-lithography machines has been a smashing success since the early 1990s. The new Micrascan V is Silicon Valley Group's crowning achievement, the machine people thought they couldn't make.
Turns out they can't. Yet.
When Intel ordered their new Micrascan Vs, SVGI committed to an April, 2001 delivery. Intel wanted to have as many as four of its fabs ready to produce .13 micron chips this year.
But SVGI proved unable to deliver working machines. Now Papken der Torossian, chairman of SVGI, says they won't de delivered until Q4. (Could that slip even further? Sure.)
This has been a stunning blow for Intel. The company can't risk falling back to wider-trace technology, especially after the embarrassment of its faux pas with the1.13 gigahertz Pentium III chip last summer -- the chip it announced, sampled ... then couldn't deliver in quantity.
This is, after, all, a highest-of-the-high tech company.
And yet it can't back away from its commitment to get .13-micron-based chips in its customers' hands this year. Brave words still come from Santa Clara: "We're on track to begin shipping products this year on .13-micron."
But not without huge and unanticipated expenses. There is another path to getting to "point-thirteen" chips: the phase-shift machines produced by Numerical Technologies (NMTC:Nasdaq - news - boards). I've heard engineers call the trickery used by Numerical's phase-shift-mask equipment to produce .13-micron parts on .18-micron equipment "shadow technology." At a minimum, it's a less direct answer than Intel was counting on.
Intel is investing heavily in licensing Numerical's technology -- including a patent-license swap last month that surely hurt. And apparently Intel already knows that even with the unexpected investment in Numerical's solution, they won't be able to come close to hitting production targets this year for chips based on .13-micron designs.
You could argue that the market is perfectly happy with present, slower CPUs, that delivering these new, faster CPUs is not critical for Intel. But at a time when it is reporting disappointing quarters and falling demand, and seeing erosion of market share to Advanced Micro Devices (AMD:NYSE - news - boards) -- when Intel's credibility with its OEM customers, analysts and perhaps computer buyers is eroding -- failing to deliver reasonable quantities of new chips announced as long ago as these were is a serious stumble.
To be polite about it.
This isn't entirely "new news." Intel apparently began briefing analysts at the end of April on its problems. But the investing community as a whole hasn't known about this latest stumble.
Increasingly, Intel looks less and less like a high-tech star and more like Jimmy Breslin's The Gang That Couldn't Shoot Straight. |
