Alpha Status Report, April 17, 1992 Background The introduction of Digital's Alpha RISC architecture comes at an awkward time for the company. In January and again in April 1992, Digital announced its first-ever operating losses. The company has been involved in the continuing turmoil of downsizing and restructuring. Several high level executives have recently departed. And critical products and key supplier relationships, announced within the last 24-36 months, seem to have quietly been shelved without clear explanation. Specifically, these include ACE/MIPS and the VAX 9000. (For more details see also Aberdeen's Market Viewpoint -- Digital Mainframe Systems: Casey at the Bat, April 1992). No matter how powerful Alpha-based products eventually turn out to be, two issues are becoming apparent. First, Digital finally decided that it could not survive as the only major systems supplier without its own RISC. Digital was forced to turn to Alpha after the failure of the VAX 9000 at the high-end and after suffering delays and lack of performance with partner MIPS on the R4000. Second, in Aberdeen's opinion, Digital's management and field organizations have thus far been handling the Alpha migration issue badly. Rather than presenting realistic expectations for system and software delivery to the client and prospect base, management panicked in Q1/92 and accelerated Alpha marketing programs. This had the inevitable result of acting as a powerful anesthetic on VAX orders. Although the VAX 6000-600 family, introduced last October, has industry leading TPC-A price performance characteristics, these systems had only been in volume delivery for four-to-five weeks before the Alpha blitz completely neutralized sales efforts for the quarter. Recent Financial Results Digital has now compiled the unenviable record of two consecutive quarterly operating losses amounting to nearly half a billion dollars. This has occurred despite reductions in headcount with worldwide employment now at approximately 116,000. Briefly stated, the major problems are these: ** Digital has no products at the high-end with which its field can compete with IBM. Customers interviewed by Aberdeen find the multiprocessing VAX 6000 too complicated to tune for data center production applications as an alternative to mainframes. The AS/400 and HP 9000 continue as popular mid-range alternatives to DEC VAX. ** Digital's growth in services revenues -- which generally includes not only hardware field service, but also systems integration, software development and network development -- is limited to approximately 300 DEC "named accounts." While service revenues continue to climb, the growth rate is declining. Even the largest 300 customers only have so much budget to allocate. And Digital has finite personnel resources with which to pursue the market opportunity. ** It appears that Digital's workstation unit shipments grew in the first quarter calendar 1992. However, Digital is still having a rocky time with negative market perceptions concerning their status as a legitimate PC supplier. The company's relationship with Apple has cooled in view of IBM/Apple alliances. And Digital's increasingly warm relationship with Microsoft is still based primarily on product futures -- such as the NT operating system. Pathworks, the PC integration software included in NAS (Network Applications Support) is, however, popular. Pathworks is based on Microsoft's LAN Manager. ** The MIPS R3000-based DECstation 5000 workstations are designed with a daughter card to provide an efficient upgrade to the R4000. These have sold well for DEC -- especially early in 1992. Many workstation market customers seemed to continue to buy on an opportunistic, project-driven basis. Thus the Alpha upgrade and transition issues that were so damaging to the VAX 4000 and VAX 6000 order rate, do not appear to have had as much impact on workstation sales -- yet. Workstation clients continue to appreciate Digital's strength in networking. Digital also seems to appeal to the price sensitive segment of the market. This is consistent with the high unit volume shipment results -- with poor margins. ** We believe, however, that Alpha will likely contribute to even more customer confusion about Digital's workstation products and strategy in the coming months. Alpha's VMS (read VAXstation category) leanings -- in the near term -- will be unappealing to an Open Systems and Unix prospect. Lack of binary compatible applications with MIPS will be another strong negative. Next Steps Digital Equipment will attempt massive damage control at the upcoming DECWORLD which takes place in Boston from April 27 to May 15th. We expect that as Digital continues to attempt to accelerate Alpha products -- and intensify marketing programs -- that customers returning from DECWORLD will have received a heavy sell on the benefits of Alpha as well as more details about specific products and shipment timetables. Although Aberdeen does not believe that this will result in a buyer stampede, it is possible that Digital will stabilize the current financial situation by offering some kind of no-risk upgrade programs to create an early stream of orders and cash. But VAX 6000-600, not workstations, will be the focal point of this campaign. We believe that the workstation group will likely emphasize that with the R5000, it's business as usual. The Facts, Ma'am, Just the Facts Here is Aberdeen's realistic, April assessment of the Alpha situation. Alpha's architectural design center is focused toward high-end systems. This means simply that we are likely to see more products emerge from the Alpha program in the high-performance workstation segment and multi-user systems arena than in the $5000 per seat commercial workstation space. The MIPS R4000, much like the Sparc Viking chip, has a design center closer to the low-end of the market. Alpha, like RS-6000 and Precision Architecture, is positioned up the power curve. Therefore, customers should expect that the native benefits of the Alpha architecture are most readily obtained in more powerful packages. Specific to this point of higher-end system design, Digital's Alpha produces 30 watts compared to the 15-watt PA7100 chips from Hewlett-Packard. Another outcome of this design issue, is that although its MIPS relationship has obviously deteriorated, Digital will continue to not only support its MIPS-based workstation customers, but also likely continue with current plans to use the R4000 to aggressively compete with Sun Microsystems at the low-cost-per-seat end of the workstation market. Aberdeen believes that the combination of two factors, (the high-end Alpha design center and the revenue potential of an existing DECstation 5000 workstation base) suggests that R4000 workstations will be sold by Digital throughout 1993 using its OSF/1 operating system release. Clearly, this means that Digital cannot have binary compatibility with high-end Alpha systems -- at least until OSF/1 porting is completed for Alpha in mid-to- late 1993. Aberdeen also expects that when the Alpha dust settles late in 1992, Digital will attempt to salvage ACE with its MIPS-based machines, as well as with Intel-based, Microsoft NT commercial workstations. Digital's Hudson, Massachusetts semiconductor facility is currently producing 83-MHz chips with an advanced CMOS process. However, these N-VAX chips running at 83-MHz are not based upon the Alpha architecture. Instead, N-VAX chips from Hudson are the most advanced implementation of the VAX architecture and are used as a CPU engine in Digital's very aggressively-priced VAX 6000-6XX. Hudson is also responsible for Alpha. The first Alpha microprocessor produced at the facility, the 21064-AA, is a 200-MHz product. It was used to showcase Digital' s Alpha project at the International Solid State Circuits Conference in February. 200-MHz represents peak performance for Alpha at the present time. Aberdeen still expects that early Alpha-based systems will be powered by microprocessors running at approximately 150 MHz. These 150-MHz parts are already in production on the line in Hudson. But specifics -- based upon facts -- about yield as well as details on packaging are not available at this writing. Digital has announced that it anticipates being able to deliver the 21064-AA in quantities of 1000 for $1,650 each beginning in July 1992. Aberdeen believes that volume production of a 150 MHz part has an 89% probability by August 1992. Perhaps the most critical, if least discussed, issue concerning Alpha and its impact on both Digital and Digital's customers has to do with software conversion. Alpha was not intended to be binary compatible with VAX/VMS. While Digital presumably could have included a "VAX compatibility mode" in Alpha, this would have penalized performance for what in retrospect would likely have been a few-year transition issue for users. The first issue to keep in mind with Alpha software, concerns the fundamental difference between Reduced Instructions (RISC) and Complex Instructions (CISC). With all RISC implementations, software compilers take on an increasingly important role. It is the RISC compiler, after all, that must execute in software what previously was handled by the microprocessor in CISC implementations. From an Alpha performance perspective, it is safe to say that no one yet knows how well designed or how well tuned Digital's compilers will be. Furthermore, first generation Alpha is superpiplined and superscalar -- a complex design under any circumstances. Alpha contains primary cache on the microprocessor and uses off-chip, secondary cache running at 66 MHz. Digital has no demonstrated experience in this cache and memory management software relative to RISC. A second critical transition issue has to do with other types of systems software. Digital has maintained that Alpha is operating-system neutral, yet according to announced schedules, VMS will appear before OSF/1. The company's software development and porting teams are clearly focused on moving millions of lines of VMS code to the Alpha platforms as a first priority. Although Digital has highly automated this process with CASE tools -- probably more advanced than those available to Hewlett-Packard or IBM in the early days of their RISC projects -- porting VMS and its robust software environment includiing Rdb, VAXcluster, networking, and transaction processing is still an enormous challenge. Not only is the pace of this migration difficult to measure, the overall performance of resulting Alpha-based systems is still completely unknown. Aberdeen expects that the complete VMS software transition will take a minimum of two years. This means that it will be 1994 or perhaps even early 1995 before Alpha systems offer a meaningful high-end software environment alternative to current VAX customers and prospects. OSF/1 or micro-kernel-based OSF/2 release plans are even less well understood. Summary Aberdeen believes that Digital and its customers would have been best served by a more gradual transition to Alpha marketing doctrine. Now, however, the company faces the potential of a widening gap between a VAX that appears doomed to obsolescence and next-generation, advanced Alpha systems. Thus Alpha presents serious business as well as technology challenges to a company increasingly under siege. Competitors are certainly going to take advantage of Digital's blundering with undeliverable futures. Aberdeen believes that Digital's Hudson facility can produce quantities of 150- MHz Alpha microprocessors in 1992. These are already being shipped a few at a time to large customers in primitive systems packages for testing and evaluation. At this writing, it appears that microprocessor hardware stability (i.e. chip yields) is less of a problem than potential snags in software migration (i.e. compilers and robust VMS applications). We believe that Digital will introduce or showcase several products based upon Alpha microprocessors, beginning at DECWORLD later this month. These systems are likely to be predominantly packaged for the uniprocessor mid-range during the early stages of product roll-out Q1, 1993 for the earliest volume shipments. The earliest systems will run Open VMS, based upon release 5.4. Finally, software is the real potential Achilles Heel of the first phase of the transition. While Digital hopes most applications will require only a "recompile and run" investment of customer/ISV time and energy, this statement over-simplifies the enormous challenges and risks that lie ahead. The bottom line is that without software, Alpha -- even at 200 MHz -- is useless to Digital. Even if Digital successfully meets these software hurdles during the next 12-18 months, many other questions remain. Will Alpha scale for multiprocessing? Has the architecture been designed for floating point or integer? What is the real benefit of a 64-bit architecture for existing applications software? How will the systems perform? Aberdeen will continue to monitor each of these issues carefully over the coming months. It is clear from recent financial results that Digital no longer "has it now!" Digital has entered the RISC technology business much later than its fierce rivals, Sun Microsystems, Hewlett-Packard, and even IBM. While the company gets high marks for recent public relations campaigns that have created tremendous mindshare about Alpha, Digital now faces daunting challenges as it attempts to turn an architecture into fully functioning systems-level products -- and profits. Aberdeen cannot emphasize strongly enough that even flawless execution of this business plan will take 24-36 months. And during that time Digital will face tremendous pressure from customers and competitors about its confusing, piecemeal, point-product strategy.