Windows, Unix, Linux, SPARC, Pentium and Merced - there's nothing technologists love more than a debate about operating systems and microprocessors. But soon, those debates will be irrelevant. We have reached the end of computer history as we know it. Fifty years of CPU- and operating-system-centric computing is ending. Large organisations are on the threshold of an era that will be characterised by widespread diffusion of the computing function. In the next five years, the economics of computing will cause work to be dispersed to autonomous systems that will combine as needed for cooperation, without having to depend on specific operating systems and microprocessors to function at all.
According to IDC, this year's worldwide spending on IT will be $US1.6 trillion. At a conservative future annual growth rate of 9 percent, total IT spending for the next decade will be $30 trillion. That's simply not affordable. The demise of computing as we know it will be forced by the following:
In 1997, US spending of $660 billion on IT neared that of corporate wages and salaries ($732 billion) and was rising at a faster rate. That makes IT an attractive target for cost reduction when an economic slowdown occurs.
Forty-six percent (and rising) of business capital investment is in computing equipment. Now that corporate allocations to IT have been reached, growing needs for investments in energy, transportation and manufacturing must be met.
The "computer paradox" remains. There's no demonstrable relationship between computer spending and corporate profits. Executives can no longer tolerate that.
There's no technical reason why each of the 500 million corporate desktop computers estimated to be put in place in the next five years needs mainframe-like overhead such as 20 million to 60 million lines of code, gigabyte files and a managerial complexity that I estimate amounts to at least $3,000 to $5,000 each in excess overhead.
Despite all the talk about distributed computing and e-commerce, the reality is that the principal appliances of the computer age are the Wintel machines. Their architecture is an embodiment of centralised thinking. The operating system must be fully informed about every conceivable feature and event.
The extremely successful Wintel offerings are the result of valiant efforts to include everyone within their fold through cooperation, backward compatibility, software engineering tools, added features and options that accommodate increased variety. Although commendable, those efforts have led inevitably to a level of complexity that is manageable only at exorbitant costs.
The Wintel central control approach to computing is doomed to economic failure because it passes on to its customers huge administrative costs and enormous risks for uninsurable failures. Corporate management is searching for simple, low-cost and reliable solutions.
My bets are on these major innovations:
User devices without an operating system, memory, compilers, browsers and peripheral drivers. With high-speed corporate intranets, most office workers need only a panel driven from network servers that can benefit from the economies of reduced overhead costs, increased reliability and superior security.
Software applets that run on "virtual machines," which make every "client" device independent of processor technology and unrelated to the choice of an operating system. Because they can function without requiring support from a specific choice of technology, such software will also have a much longer economic life than today's software, resulting in an enormous improvement in life-cycle costs.
Software that can communicate, interact and support business services without central control that guides their every step.
The consequences of this transformation are far-reaching. Most of the code and all of the organisational forms based on central planning will crumble. Those who bet on the new architectures will prosper.