In fact, Compaq had purchased the remains of DEC because of significant assembling facilities, its wide distributional network (in 98 countries) and that cross-licencing agreement with Intel (in particular, allowing to manufacture 8-processor Profusion servers). As it seemed to be, the division developing the Alpha architecture wasn't welcome really: Compaq produced desktops, notebooks, workstations and servers based upon Intel's processors for a very long time, also paid certain attention to AMD's processors. Therefore, Compaq established an alliance with Samsung to develop the architecture in June of 1998. To mention, DEC and Samsung signed an agreement in February of 1998 which granted to the latter a full access to all DEC's Alpha-related patents, a permission to manufacture the existing Alpha processors and even to design and produce their derivatives. A new company was incorporated mutually called API (Alpha Processor Inc.) with the purpose of promoting the architecture. Well, some people seemed to make right conclusions after having a look at DEC's history. In the summer of 1998, EV6-based systems entered mass production stage featuring the best price/performance ratios compared to other competing products available on the market. Serious problems with the future Itanium by Intel were reasonable enough to conclude that the situation described would remain unchanged in the near future. However, not only Samsung manufactured EV6 those days. There was Intel with its Fab-6 in Hudson bound by the final agreement with late DEC.
 
Year 1999 was unsuccessful to Compaq because of falling sales on the market of personal computers. The most frequently named reason was an underestimation of possibilities given by the Internet to promote and sell computer hardware. Unlike Dell which adapted its business model and offered computer equipment priced most attractively among all top brands. Compaq's CEO, Eckhard Pfeiffer, resigned after a financial disaster in the 1st quarter of 1999. Trying to reduce losses, Compaq started to minimise its presence in certain areas and that affected Alpha systems. In May of 1999, an assembling line of AlphaServers in Salem (New Hampshire, the USA) was announced to be shut down soon.
 
On the 23rd of August 1999, a notorious event took place: Compaq announced to discontinue participation in development of Windows NT and stopped to supply this OS with Alpha systems of its own. In fact, it also laid off almost all people (about 120 programmers) from former DEC's Western Research Laboratory (DECwest) who worked on this project. Accordingly to Compaq's statistics, among all preinstalled OSes on newly shipped Alpha machines Tru64 UNIX held a share of 65%, OpenVMS — of 35%, and Windows NT — just about of 5%. So, there was no reason to keep flogging a dead horse. A week later, Microsoft announced in return that there would be no Windows 2000 for Alpha released, even though the RC1 (Release Candidate 1) was ready by that moment. Considering a fact that Microsoft together with Motorola and SGI discontinued any support for the PowerPC and MIPS architectures respectively in 1997, the future of "the universal OS" appeared to rely on a single computer architecture. Of course, if to discount IA-64 which failed on the workstation market soon and never got to desktops or notebooks.
 
In December of 1999, Compaq and Samsung signed a memorandum to support the leadership of the Alpha architecture in the near future. Both sides agreed to invest 500 mln. USD in this endeavour. Samsung was obligated to spend 200 mln. USD while developing and tuning new technological processes, and Compaq was supposed to spend 300 mln. USD on a further development of server solutions and Tru64 UNIX. Apart of that, during the same month Compaq and IBM agreed that the latter would manufacture Alpha processors using a copper-conductor CMOS technology of its own upon completion. However, Samsung was assured to remain the primary supplier of Alpha processors in the future. The year passed not so good for Compaq illustrated well by a price per share delta: from 51 USD in February to 28 USD in December. Although many analysts stated it could be worse.
 
Y2K passed for Compaq quietly. Samsung wasn't able to tune its 0.18µ process unlike IBM which started to supply EV68C to Compaq in small quantities though. Overall, the market had to enjoy considerably slower 21264A (EV67) processors. Development of 21364 (EV7, also known as Marvel) was in progress still, though 21464 (EV8, also known as Araña) was mentioned here and there. The fall of dot-coms affected Compaq's shares which dropped in price to 15 USD per share by December, i. e. for 44% since January. Could sound strange, but that was a good result because other companies which depended more on e-commerce lost much more: Gateway — 75%, Apple — 71%, Dell — 65%. Dot-coms themselves were either bankrupts or close to that. For example, Yahoo.com lost 95% and Priceline.com — 97%.
 
In the beginning of 2001, Samsung started to manufacture EV68A in quantity, but the right moment had been missed already. Compaq planned to ship EV68C-based systems (GS-class AlphaServers) and to modernise those already in production. EV7 was still somewhere there when something happened not expected at all. On the 25th of June 2001 ("black Monday"), Compaq proclaimed to transfer all its server solutions from the Alpha to IA-64 architecture by 2004. In fact, it meant a surrender to Intel and Hewlett-Packard. EV8 was cancelled immediately, though some details about its internals were provided at a Microprocessor Forum in October of 1999, and EV7 was scheduled for release not earlier than the beginning of 2002. Afterwards, the Alpha Microprocessor Division had to be disbanded and most of its personnel should be employed by Intel. Samsung and IBM ceased to produce Alpha processors soon after. However, the situation became even more interesting later. On the 3rd of September 2001, Hewlett-Packard announced its intention to acquire Compaq which experienced certain financial difficulties and its price-per-share value was of 10 USD in December of 2001. The deal was approved by shareholders' meetings of both corporations as well as by the governments of the USA and Canada. The acqusition had been completed by May of 2002.
 
On the 21st of October 2001, API (renamed by that moment to API NetWorks) delegated all rights to service Alpha hardware sold by that moment to Microway, the largest [after Compaq] builder of Alpha workstations and servers, an old partner of late DEC. API itself left the market of Alpha products and concentrated its efforts on network technologies, data storage systems and development of the HyperTransport bus.
 
As a conclusion, it could be said that Compaq hadn't followed many of those mistakes made by DEC before, but it had screwed up the architecture anyway. High-performance Alpha systems based upon 21264A and 21264B hadn't hit the price tag of 2000 USD, and low-cost 21264PC had never appeared. A possibility of producing low-priced mainboards in volume using AMD Irongate had been ignored, and pricey DEC Tsunami (offered by Compaq for over 1000 USD per set in OEM lots) had left no chance for Alpha systems to enter the mid-range computer category. There were other companies who designed system logic sets for AMD Athlon, but they decided not to adapt them for 21264, though VIA had such an intention initially. The AMD Irongate-4 (AMD-762) north bridge appeared roughly half a year after Irongate-2 (AMD-761), i. e. in December vs. August of 2001, yet too late to shine in any mainboard design for Alpha. It could be used in 2-processor configurations and supported the same memory interface as Irongate-2 (64-bit 133MHz DDR SDRAM), thus was able to deliver an excellent price/performance ratio given a pair of inexpensive 21264B processors.
 
Although Compaq had managed Alpha to lose the workstation market. In fact, there were only two Alpha workstations produced by Compaq: XP900 (with a 466MHz EV6 and 2Mb of B-cache; code-named as Webbrick) and XP1000 (with a 500MHz EV6 and 4Mb of B-cache, later with a 666MHz EV67; code-named as Monet). They were based upon DEC Tsunami, though with a relatively narrow 128-bit memory data path. These machines had failed in competition with x86 workstations which were less powerful but also much less expensive. Eventually, their failure indicated the end of Windows NT on Alpha: Compaq's servers ran Tru64 UNIX or OpenVMS mostly, other vendors preferred to install Linux though. This issue could and should be counted against Compaq. For the record, DEC had fought for the workstation market desperately and even had achieved some success. For instance, if to consider good sales of DEC PWS (Personal WorkStation; code-named as Miata) with a DEC Pyxis and a 21164A (EV56) inside. Compaq hadn't achieved anything on this playfield but had lost everything quickly.