The Hewlett-Packard and Intel partnership.

In June 1994, Hewlett-Packard (HP) and Intel announced an exciting new collaboration with ramifications for companies worldwide: a partnership to develop what has become known as the Merced chip - a new generation of core microprocessor technology ("architecture") different in kind from its antecedents. This article, reporting a conference speech given late in 1997, provides an eye-witness account of the collaboration as seen at that date by one of the key players. The speaker reflects on the reasons why the two companies got together, the strategic and organisational implications for HP - and the key factors ensuring that the partners continued to work together productively, if not always in perfect harmony.

The HP-Intel relationship goes back to late 1993. The first meeting between the two CEOs, Andy Grove and Lew Platt, to discuss establishing the relationship was in Lew's office in November 1993. The partnership was announced in June 1994, and we began our joint efforts to develop the new architecture - core technology for microprocessors - at that time though it was not until February 1995 that we actually publicly signed the alliance agreement that has been governing the relationship.

Hewlett-Packard's enterprise systems group has a core systems technology organisation that encompasses many aspects of basic systems technology. My specific job is to manage HP's transition from its current proprietary architecture basis, which is comparable to that of many other companies, to the next generation that HP is developing jointly with Intel. Because of that, I am also a member of the group responsible for managing the HP-Intel relationship. This article describes some of our four-year experience in managing this relationship, as well as explaining why the two companies are working together and what we intend to get out of it.

Why Get Together?

In the electronics industry, the basic architectures do not have a lifetime of more than 15 years. Before we approached Intel, we had already researched the next generation of architecture. We used our experience in moving to the RISC architecture in the 1980s, which we had pioneered, to determine what the next architecture after that would be like. We started with research and moved onto development activity to prepare ourselves for the next generation, which we expected to begin around the turn of the century.

HP and Intel each already had its own independently successful architecture. Why not just move to the next generation independently? The basic answer is very simple: huge and accelerating cost. In particular, we concluded that the next generation of architecture needed to be an industry standard. There were two main reasons for this. One was that the investment cost of developing a state of the art microprocessor is increasing at a dramatic rate, at about 1.5x times from generation to generation - building leading-edge microprocessors is now comparable in cost to the arms race. Second, the cost of manufacturing those devices is escalating. The cost of a so-called FAB, a factory that is capable of building microprocessors from a semiconductor process technology perspective, is about $2bn today and has a lifetime of about three years.

The fact that the entire microprocessor business, the primary driving force for computers, is becoming so extremely investment-intensive means that ever fewer players can afford to participate. Anyone who wants to be a player needs to be able to …