Windows + Intel = Wintel?
One of the most important partnerships in the global computer industry is that of Intel and Microsoft. Cheap PCs using processors (CPUs) from Intel, coupled with operating system (OS) software from Microsoft, have defined low--cost computation for most people. The partnership was dubbed `Wintel' -- Windows from Microsoft, on computers with CPUs from Intel. The two firms are at $400 billion dollars in market capitalisation -- more than twice the entire Indian equity market.
The rise of Intel and Microsoft is a remarkable episode in economic history. Both companies were fairly unimportant when IBM chose to back them for its launch of the IBM PC. Neither were the obvious, strong choice. Zilog and Motorola built better CPUs than Intel, and Digital Research did better software than Microsoft. IBM may have deliberately chosen weak partners to ensure control over core functions. From that accident has risen a remarkable domination of their respective sectors. Economists view this as an example where "initial conditions matter"; where small fluctuations in the remote past have enormous consequences in following decades. Ideally, capitalism should have a self--correcting nature, and these fluctuations should matter much less.
Intel and Microsoft have followed different paths in the decades that followed. Most engineers disrespect Microsoft and have a grudging acceptance of Intel. This attitude puzzles laymen : why is Microsoft excoriated while Intel is tolerated?
The answer lies in history. Both Intel and Microsoft started with weak technology. If we pick the late 1980s as a point of reference, both were far behind their rivals. Intel has done a remarkable job of overcoming its clumsy roots and coming close to the state of the art today. This transformation has come at an enormous cost, and could not have happened without steady monopoly profits -- hence this is unsatisfactory from an economic perspective -- but Intel today is technically respected. Software from Microsoft, in contrast, is roughly as confused as it was in the late 1980s. This violates a sense of fairness amongst engineers, who see Microsoft as a random fluctuation that has gone awry.
In addition, Microsoft's business tactics have been fairly offensive. In recent days, the US antitrust investigation into illegal acts by Microsoft has brought information to the public eye which has always been well known to industry insiders. Intel has done better in avoiding abuses of its monopoly power.
Through the 1980s and the 1990s, the Wintel partnership worked smoothly. Intel had the highest volumes of production of CPUs, and PCs were ubiquitous low--cost computers. In order to use PCs, there was no alternative but OSes from Microsoft. Microsoft kept releasing slower software, forcing users to require faster CPUs. As long as Windows worked only on Intel processors, Intel had an interest in supporting Microsoft. As long as non-Microsoft OSes did not run on Intel CPUs, Microsoft had an interest in supporting Intel. Anyone who chose one of Wintel tended to use the other. This symbiosis was essential; it is hard to see how $400 billion of market capitalisation could have been built if the two companies were unconnected.
The Intel of old was not enthusiastic about Unix. Unix is a multi--vendor OS that leaves users free to choose an OS vendor and a CPU vendor. Unix users can move to non--Intel CPUs, while a Microsoft user is locked in. Hence Intel promoted Wintel and underplayed Unix.
Now, the CPU business has a strange logic. Top--end processors are frightfully expensive to design and manufacture. It is typical to spend $500 million in designing a new family of CPUs, and early production costs in excess of $1000 per CPU are common. However, there is no alternative but to work on new high--end designs, for they later become mass--market commodity items. In order to have a slender edge in the high--volume CPUs of today, the firm needs to have invested massively in exotic CPUs two years ago.
What motivates this article is a possibility that the Wintel partnership may now be at a turning point. Two basic changes have taken place:
- For the first time, today, the vanilla Windows user does not need the costliest Intel CPU. All these years, there was a thirst for faster CPUs to do simple tasks given the inefficiency of Microsoft Windows. Today, for the first time, low--end Intel CPUs -- and their cheap clones -- are powerful enough for the normal tasks of PC users, even if Microsoft OS is used. If clones are viable, Intel's profit margins are low. Intel's market share has dropped dramatically from 84.3% of retail computers in August 1997 to 54.3% in August 1998.
- For the first time, non-Microsoft OSes have become available for Intel CPUs, in the form of several different Unix flavours, including Linux, BSD, Solaris, and SCO. Five years ago, users were forced to choose between cheap PC hardware (with Microsoft OS) or more powerful hardware (to run Unix). Today, Unix requires cheaper hardware -- a PC costing Rs.25,000, using Unix, is an adequate server for a 10--man office.
The first change puts Intel in a bind. To stay in the treadmill of CPU design, Intel needs customers who are hungry for speed. If the CPU business does not move fast enough into newer CPUs, Intel's profits will be wiped out by the clones. For years, Intel relied on users trying to do simple tasks with ever--slower Microsoft software to fuel sales of high--end CPUs. Today, that process is not working so well.
Where else can Intel sell high--end CPUs? Into servers; into large computers used in the corporate world; and into the ever--proliferating Internet. This is not Microsoft territory. Microsoft is trying to write a new OS, called "Windows NT", for these new markets; it is expected to be ready for use in 2001. A census was recently run of the 750,000 web servers in Europe. Unix has 75% market share; Microsoft has 23.4%. The market leader, Linux, does better than all Microsoft OSes combined.
Corporate computing -- as opposed to individual computing -- is dominated by mainframes, and servers using Unix and Novell. In these areas, Microsoft does not work as a vehicle for Intel. Intel needs to position Intel CPUs as a way to run Unix in order to compete with the CPU vendors who dominate this area, all of whom use Unix (Sun, HP, DEC, IBM and SGI). This is now possible, given good Unix alternatives for Intel.
This is perhaps why we now see some cracks in Wintel, with several announcements in recent weeks from Intel. Intel is working for better `device drivers' for Intel--Unix; it has purchased an equity stake in Red Hat, a Linux vendor; it will work to ensure that Unix runs well on future Intel CPUs. This is in contrast with the Intel of yore, where Windows was the prime vehicle to sell new CPUs.
Every `strategic alliance' has a finite lifespan. Wintel was one of the most successful and most important partnerships in the world. It has endured for fifteen years. Yet, as Andy Grove of Intel said on Thursday "We didn't go to the altar, never swore allegiance or anything like that."
Footnote: This article was written in 1998. Some years later, I saw an article which seemed to tell this very story from a "few years after" perspective: here it is.
Back up to Ajay Shah's media page