Which area of corporate expenditure today costs the average business 40% more than it did in 1991? Astonishingly few people get the answer right: IT. Overall, computers cost UK industry some 1.3% of turnover in 1991, reckoned management consultancy Price Waterhouse in its annual IT review for that year. Yet in the latest review, five years later, the comparable figure was 1.8% of turnover - an increase 'broadly consistent across all industry sectors,' notes Price Waterhouse IT strategy partner David Henderson.
Indeed, at least for people-intensive businesses such as professional services firms, 'IT has overtaken property costs, and now forms the single largest expenditure after people costs,' according to a senior partner in another Big Six consultancy. Nor is the culprit the once-hefty cost of running huge corporate mainframes or servers. Not only has the operation of many of these been outsourced on advantageous terms to facilities management companies, but the initial purchase costs have also fallen. 'Mainframe manufacturers have learned to live on smaller gross margins, resulting in lower prices,' says Joseph De Feo, until recently the high-profile IT director of Barclays Bank, and now president and chief executive of IT industry standards body The Open Group. 'What's more, their designs now incorporate lower cost CMOS (complementary metal oxide semiconductor) devices, rather than the more expensive technologies of earlier generations.'
In fact, the humble office personal computer or PC must shoulder a good proportion of the blame. It's a conclusion that will surprise some: PCs, after all, are cheap. But peer a little closer at the supposedly low-cost PC, and some frightening costs emerge - although for many businesses they will be buried deep in the detail of the management accounts. Every time a new version of a piece of software is released, for instance, a user or a technician has to install it on every single PC which uses that software - and some companies own tens of thousands of PCs. Some large organisations, complaining that they have little control of either the timing or the resultant benefits of such upgrades, are beginning to describe the process as 'vendor push'.
Combined with lost productivity as users try to figure out how to make their machines work, support - or trouble-shooting - is a large part of any company's IT costs. Many large firms now find it quicker and more efficient to employ their own technicians to provide this service rather than rely on software vendors. And if the technicians can't solve the problem remotely, they will have to make a personal visit, obviously adding further to the costs. And, of course, as companies' IT infrastructure grows more complex, more and more employees need IT training in order to be able to use IT effectively.
'Twenty years ago, IT costs were all in one place - a data centre - and were easier to monitor,' says Harry Knightbridge, group director of information services and technology at Guinness plc. Now, an increasing proportion of IT costs are incurred and accounted for at departmental level. With departments buying their own hardware and software, and sending people on training courses from their own budgets, it is harder to keep track of exactly how much expenditure is IT-related.
The size of this blind spot is frightening. Take PC-based systems, chiefly those deployed at departmental level, operating either on a standalone basis or in departmental client/server workgroups. A study recently commissioned by Compaq Computer Corporation found both that a third of respondent companies could not even estimate the additional cost of supporting a PC in relation to its original purchase price, and that the vast majority of those which did venture an estimate thought that the cost to be just a small fraction of the original cost of acquisition. A full quarter of companies thought that PCs cost less than 20% of their price to support, with only 4% prepared to concede that such costs might exceed the original purchase price.
Under the circumstances, a certain myopia is admittedly understandable.
Hardware vendors are happier to point to falling per-unit prices, rather than to the increasing number of units they are selling or the growing costs of upkeep. Similarly, software vendors point to falling prices and added functionality, but fail to draw quite so much attention to the increasing frequency of software upgrades.
In fact, typical lifetime PC support costs are at least 400% of the original price, says the Compaq report, drawing on estimates from IT analysts such as the US-based Gartner Group. Gartner's analysts estimate that the total cost of ownership of a networked corporate PC may be as high as $13,187 (£7,800), although the cost for the average business is likely to come out closer to $9,092. Other analysts agree, placing the annual cost between $8,000 and $10,000. US semiconductor manufacturer Intel - whose products lie at the heart of both PCs and large servers - estimates the cost at $8,300. Set this against the $2,000 it costs a US client to buy a sensible PC with workable software programmes, and the scale of the problem is clear.
The spread in the above estimates is not only due to measurement inconsistencies, but also to the difficulty in defining what exactly a corporate PC is: machine standards vary widely, and so do the tasks that they perform.
Thus whilst Intel estimates an average cost for business as a whole at $8,300, the company reckons that its own costs are around $9,300, partly because of the complexity of the tasks that its 50,000 desktop PCs and 5,500 design workstations perform, and partly because many of them are hooked up into its own global network.
The PC is a very versatile animal, and few companies deploy units in exactly comparable ways. Some machines use standard software, either on a standalone basis or linked together in simple departmental workgroup.
Others form part of a global or national corporate network, running complex software or performing complex tasks. Some run in businesses that don't have, and don't need, mainframe functionality. Others run in businesses that do need mainframe functionality, but have downsized this to a client/server configuration, while others run PCs and mainframes, with varying degrees of linkage between the two. Training needs differ, user-support needs differ, maintenance needs differ, configuration needs differ, and the number of computer programs in active use differs. Each contributes to the cost of ownership, although variably. Even the operating system makes a surprising difference. Analysts expect that Windows 95 will reduce typical support costs by $1,180 a year, equivalent to almost $6,000 over a five-year life.
Are such high levels of cost set to continue - or, worse still, grow?
It seems not. A flurry of initiatives in the latter part of last year indicate that vendors are starting to perceive cost of ownership issues as the new competitive frontier. In September, Intel president and chief executive Andrew Grove launched a Wired for Management initiative which sought to pool the resources of a number of vendors in order to reduce these costs. Arguing that a certain level of costs would always be associated with the PC's 'adaptability and flexibility for the future' ('the PC is a true Darwinian device that always adjusts to its environment'), Grove nevertheless acknowledged that the time had come to make inroads into the human cost of supporting this flexibility.
There is good reason for Grove's concern: Intel not only makes most of the microprocessors that power the world's PCs, but an increasing proportion of their motherboard (the printed circuit board within a PC), too. If the PC manufacturers feel that they should be competing on cost of ownership, then Intel had better start delivering. While Intel makes the world's most powerful PC microprocessors, other manufacturers offer devices which are almost as powerful. Should rivals be able to offer their product alongside a lower cost of ownership, then PC manufacturers will be presented with a compelling argument for switching.
Thus Grove joined other vendors in announcing better and cheaper ways of managing corporate PCs. Special instrumentation built into Intel microprocessors and motherboards would allow technicians to diagnose remotely - and sometimes cure - hardware faults on users' machines. Software and hardware would be remotely configurable. Network and software tools would enable support specialists to achieve a higher first-call problem resolution rate. The need physically to send people to users' PCs would fall sharply, perhaps by as much as two-thirds. Overall, Grove confidently looked forward to seeing the cost of ownership of new PCs 'drop by 15% in 1997'.
Other people have significantly higher expectations, notably Larry Ellison, the abrasive chief executive of Oracle, the specialist large-scale database company (and the world's second largest software company). Ellison's own net worth of $6 billion now places him at number five in Fortune magazine's annual listings of the US's richest individuals. A year ago, Ellison saw an opportunity to redefine the PC. Widely decried at the time, his network computer, or NC, sought to eliminate most of the costs of ownership and most of the costs of acquisition by placing all the complexity, as Ellison puts it, 'not on the desktop, but on the network - where it belongs'.
Ellison's thesis was simple. The chief reason that PCs are complicated to build, maintain and support is because they were originally intended to be standalone devices. Today, it's usually the case that a network exists into which desktop computers can be plugged, and automatically configured, and from which software can be downloaded. When software needs to be upgraded, the logic is to upgrade it once, on the central server.
Smart cards, like credit cards, store users' configuration preferences, so that any NC, anywhere, becomes identical to the user's own machine.
Best of all, by stripping out the components necessary for standalone operation, the NC should be substantially cheaper to build. As Ellison sees it, PC vendors' reluctance to grow up and tackle the cost of ownership has given Oracle - hitherto a mainframe company - a golden opportunity.
'The hard part of the NC is the network, and the server software needed to control and communicate with clients' computers. That's what we're good at - it's been our core competency for 20 years. Other people, such as IBM and Sun, will want to have a slice of the market but we think we're years ahead.'
Ellison is at his most scathing when attacking the upgrade spiral of 'fatware' which drives PCs to become ever more powerful. 'Look at Microsoft Word,' he sneers. 'It's got a 90% share of the market, and every year they launch another version stuffed with more features that I don't need, I don't use, and I don't want to pay for.' He comically derides Word 7's latest feature: the ability to vary the type of underlining used for highlighting spelling errors. Oracle's own 'stripped down' office suite software - codenamed HatTrick - is scheduled to ship in the first quarter of 1997, and will contain no such refinements. In common with all NC software, it will run quite happily on PCs as well, although the reverse is not the case.
As Ellison sees it, the network computer is a simple office productivity and communications device, offering basic functionality at a substantially lower cost: ownership costs should plummet, and purchase costs will be in the range of $500 to $1,000. Despite this, he confesses, 'the network PC won't replace the personal computer - just as the personal computer didn't replace the mainframe. They will happily co-exist, each machine doing the job that it is best at. Some people will always need a personal computer, loaded with the sort of software that most of us don't use.'
While the pundits ponder - and Price Waterhouse's Henderson is far from being alone in adopting a wait-and-see approach - vendors at least are putting their money where their mouths are. Office NCs have started coming to market, priced in the bracket that Ellison forecast. And the market for home-based NCs was given a fillip when US electronics giant RCA announced a $300 device, based around homeowners' existing TV sets and cabling.
Most pleasing of all - for Ellison at least - has been a complete reversal of the entrenched position once taken by Intel and Microsoft. Intel has now started to manufacture microprocessors for NCs, and Microsoft, which once famously dismissed the NC as 'dorky', has been forced to announce its own version of the NC. The tide, Ellison is convinced, has turned.
One way or another, it looks as though the cost of computing is unlikely to rise a further 40% over the next five years.