Paul ,AMD - FAB 25 Problems Continue To Drain Cash; A Race For Survival
marketinvestor.com
AMD - FAB 25 Problems Continue To Drain Cash; A Race For Survival
07:39am EST 12-Feb-98
Fab 25 continues to have problems ramping sufficient volume of K6 Based on several industry resources, analysts believe AMD continues to have severe problems yielding the K6. This means that wafer yields, are well below 50%. The problems are sufficiently severe, that several large customers have made contingency plans in the event that AMD is unable to supply sufficient volume. Given that NSM is constrained by its IBM foundry relationship and is only building to satisfy existing demand, analysts believe this means that lost K6 business for AMD will go directly to Intel. Analysts believe the persistent problems are a combination of accelerating the ramp for a new product, on a new process, with new equipment. Unlike Intel or NSM, the equipment set being used to ramp volume production for the 0.25 micron process in Fab 25, is not the same as the set used in the Submicron Development Center. Intel refers to this as a "copy exact" strategy, while NSM calls it a "square beaker" approach. The closer the equipment set is across fabs, the easier it is to transfer process technology, since the learning curve acquired in the development fab can be applied within the production facility. Otherwise, a fab may need to "relearn" much of the process, which takes time and resources.
Contract pricing for K6 products is relatively high in part of Asia reflecting severe supply constraints, especially for Asian OEMs who already have socket 7 motherboard inventory and must use a K6 or old Pentium processor in order to ship. Since Intel has cut back production of its old Pentiums so it can transition to the Pentium-II, this leaves AMD to resolve this supply problem.
Intel distancing itself from AMD, compounding current competitive pressures Any other year, AMD's problems would be critical but not fatal. This year, however, is a major architectural transition point for the industry as the PC market decides between extending the older Socket 7 platform or converting to Intel's Pentium-II based Single Edge Cartridge (Slot 1) platform. Due to AMD's inability to produce adequate supply, momentum is certainly on Intel's side. Analysts believe that once an OEM or customer chooses the Slot 1, the odds of reverting back to a lower end Socket 7 are remote. AMD's own efforts for a Slot 1 equivalent, called Slot A, probably won't appear until 1999 and will involve a different bus interface protocol (based on an Alpha 21164 processor) that would make it mechanically compatible, but electronically different from an Intel platform, and therefore not in violation of Intel patents.
In addition, Intel's product roadmap is both broad and aggressive. While much press has been made over the higher speeds Intel is expected to introduce this year (up to 450MHz), Intel is also shoring up the lower end. Analysts expect Intel to introduce sometime in April a new 266MHz Pentium-II without L2 cache - code named Covington, priced around $150. Due to its stripped down features and volume ramp on a 0.25 micron process, it will probably cost $70 to manufacture and package (versus $105 for Pentium-IIs being produced on Intel's older 0.28 micron process and the $50 for AMD (when it gets to 0.25 micron). This should make it affordable for most sub-$1000 systems. While Covington will probably be slower than a 233MHz Pentium-II with 512k cache its $45 savings in price should appeal to the cost-conscious segment. Intel also plans to introduce a Pentium-II with 256k of integrated L2 cache on the chip itself; matching AMD's current plans for a its own integrated cache solution in 2H98, the K6+. This all spells bad news for AMD, especially given their stated pricing strategy of at least a 25% discount to equivalent Intel products. Analysts believe that in order to retain large design wins and stimulate future acceptance, that discount may be closer to 40-50% lower than published Intel prices for the current Pentium-II processors.
For this comparison, analysts "built" three PCs for approximately $1000, using standard parts and prices obtained from the January 30 edition of the Processor circular. Each configuration included 32 MB of EDO DRAM, 2.0GB IDE hard drive, a 16X CD-ROM, 33.6 fax modem, and a 14" SuperVGA monitor. The critical components for comparison were the CPU and the motherboard (including chipset and L2 cache). A K6 and motherboard (512k cache)was priced at $267, while a Pentium-II 233MHz (with 512k cache on the cartridge) cost $418 - a 36% difference. However, if analysts replace the Intel CPU with a Covington Pentium II at 266 MHz with no cache (assuming equivalent performance to the 233-MHz with a L2 cache) the CPU and motherboard would be $308 or only a 13% difference with the K6 solution. Clearly this would put further pressure on AMD to lower price.
Long-term viability beginning to be questioned; two quarters of cash left The longer it takes Fab 25 to ramp, the higher the probability that AMD may not be able to fully recover. Last April, when the K6 was introduced, AMD said that its future depended on the successful ramp and adoption of the K6. Two quarters late in its ramp, with no immediate resolution in sight, analysts believe that this high risk strategy may be resolved too little, too late. AMD's continuing problems at Fab 25 continue to drain cash, leaving AMD with roughly two quarters to turnaround its K6 problems. In addition to Fab 25, AMD also needs to resolve financing issues for Fab 30 in Dresden, and analysts believe that the German State of Saxony is exerting pressure on AMD to resolve this financing soon.
Other news that AMD may be retreating to the bunkers. AMD announced last week that it was canceling appearances at four investment conference over the next two months, so they can focus on Fab 25 issues. Analysts also believe that Vin Dham's departure last fall was probably not an isolated incident. There have been other executives who have departed recently, including the director of marketing for AMD's Communications Products Group, Mike Wodospian, who left for Level One.
Underlying break-up value limits downside to current stock price Even if AMD is unable to resolve the K6 ramp this quarter, investors should be cautious about assuming a free fall in the stock. Analysts believe there would be several parties that would be very interested in each of AMD's divisions, based on their underlying value:
* Fujitsu's joint venture with AMD (FASL) has been very successful, generating $25 Million in after tax profit for AMD in 1997. Assuming a similar comparable as Atmel or Cypress, the Non-Volatile Memory group should be worth 1.5 times 1997 revenue, or $1.1 Billion.
* Vantis revenue was down 3% in 1997 and is expected to grow modestly in 1998. Applying a 2x revenue multiple, Vantis should be worth at least $0.5 Billion.
* The Communications business is still a solid segment that could get anywhere from $0.5 - 1.0 Billion.
* The Computational Products Group has two tangible assets: Fab 25 and its designs. The embedded group has not been a strong growth business, but the AMD ELAN core has been used extensively within the communications market. In addition, the K6 design and size is impressive. A company who needs an advance fab could find a bargain in Fab 25. Any company that wanted to integrate an x86 MPU core could certainly utilize the K6; or a company that needs to fill capacity could also pick up the K6 as a stand alone. Even the access to microprocessor designs is highly valuable, as evidenced by bidding last summer for the patents of the now defunct Exponential (rumored to have been won by S3).
Intel in turn, bought Digital's semiconductor business, including an older fab for $700 Million.
|
