Dale,
I would hate to be an industry analyst. The fab relationships add a great deal of complexity to the financial models. The article you posted was interesting. I like the following passages...
Despite the SanDisk and SIS moves, industry executives don’t expect many fabless companies to begin making their own chips. Most can’t afford it. And even those who could aren’t likely to want the technical challenge of running their own fabs. “It would be a real distraction,” says Ron Leckie, CEO of Infrastructure, an Irving, TX-based market research firm. He adds that fabless companies who try to do their own manufacturing risk venturing beyond their areas of expertise, as well as being caught with excess capacity when industry conditions change. “If you’re a fabless guy, and you get too aggressive, you can really get burned,” he says.
“You’re definitely swimming upstream going that direction,” agrees UMC’s Ballingall. Indeed, he notes, the overwhelming trend among device makers recently has been to shift production from their own fabs to the foundries.
Motorola, for example, aims to get half its sales from products manufactured outside the company by the end of 2001. Wayne Nesbit, vice president and director of worldwide external technology for Motorola’s Semiconductor Products Sector, says his group is now working with 10 different foundry suppliers and getting almost 40% of its revenues from outsourced products. Motorola used to rely on foundries mainly for overflow capacity, then bring production back to its own factories when demand slowed. But nowadays, he says, “foundries are a fundamental part of our capacity.”
It’s easy to see why foundries want to work with the big, integrated device makers. While it can take dozens of fabless customers’ products to fill a new fab, a customer like Motorola can potentially equal that with a single order. As Nesbit notes: "It would be very hard for TSMC to be investing $5 billion this year [on expansion] if Motorola were not a customer of theirs."
There are risks, though, in depending on just a few large customers. Foundry operators worry that device makers—despite assurances that they intend to be long-term customers—may cut back their orders during future industry slumps. As a result, some foundries are asking them for what might be termed “share-the-pain” agreements, requiring both parties to share equally the burden of future over-capacity. Most device makers and foundries are unwilling to discuss the existence or terms of these types of agreements.
[snip]
A less-risky group of customers is emerging, however, among device makers that are no longer adding to their own fab capacity. The ranks of these so-called “fab-light” companies are somewhat nebulous, since many haven’t publicly disclosed that they’re winding down their manufacturing businesses. Still, industry officials say there are a growing number of integrated device makers that, like their fabless counterparts, are fully committed to using foundry suppliers.
SanDisk is in a position where it must balance the economy of a strictly fabless charter (with its inherent limitations on available production capacity) with anticipated growth. Management has made the decision that the a purely fabless arrangement will not suffice. There will still be a strong fabless commitment, but blended with FlashVision LLC. Fortunately for SNDK the windfall from the UMC investment and the proceeds from the last two secondaries will give them the available capital to do a limited expansion. I see the $400 million dollar investment as a risky, yet necessary expenditure. The only other option is to become (and remain!) a "major customer" at several fabs, trust that production is adequate to fulfill demand, try not to get squeezed out by a more ambitious competitor (like Motorola transitioning down to a "fab-light" model), lock in long term contracts, pray there is no bidding wars for the fab's resources and hope that your not asked to "share the pain". This hardly seems to be an entirely risk-free endeavor.
SanDisk may be in a unique position where it (and Toshiba) can project that all flash product will be accounted for without needed to contract with others to round out the full potential of the fab investment. Obviously the projections for the growth in ultra-high density flash would be critical in making this assessment.
In the meantime SanDisk is collecting royalties on the flash memory it cannot manufacture itself. It has a large stockpile of cash and will generate about $70 million in royalties for 2000. I hope net income also mirrors this figure. The $400 million dollar foundry investment seems to be well within their budget if you look at things on this scale. I'm not saying it will be easy cutting the check, but at least you know the check won't bounce. And while more conservative investors in SanDisk may prefer a continued (strict) adherence to the fabless model and a dividend check, I am comfortable with the investment. As Eli says, execution will be the biggest risk to SNDK investors. I think it is crystal clear why he feels that way.
Finally, don't forget Toshiba. Toshiba's substantial manufacturing experience is also insurance that they won't get in above their head in running the fab. Certainly SanDisk won't be heading into unchartered waters with Toshiba at its side.
SanDisk is a like well-capitalized, hybrid "fab light" company that is approaching this designation from the opposite direction. It just doesn't have the same lenghty track record of some of these companies.
Ausdauer |
