So by acknowledging, albeit humorously, the IT nightmare that their customers face, and by articulating the industry's most ambitious vision for fixing it, IBM is attempting to make lemonade out of today's lemons. And in so doing, it is threatening to leave its competitors in the dust.

How IBM Plans to Conquer the IT World

Although it sells fewer units than smaller competitors in most of its markets, not to mention the fact that its prices are almost always higher while its technology rarely leads the pack, no other technology company can assemble as comprehensive a set of products as Big Blue: services, software, hardware and financing. Everyone else has a hole in his bucket. EDS ($US11.7 billion in outsourcing revenue in 2002 versus $US16 billion for IBM) lacks in-house hardware and software divisions, as does IBM's main consulting rival, Accenture.

Microsoft focuses solely on software. Only Hewlett-Packard comes close to having IBM's breadth, but HP's services and software divisions are much smaller.

Of course, all of IBM's major competitors have partnership agreements with smaller vendors to fill out their offerings, but for the risk-averse CEO, IBM is the closest thing to a one-stop shop in IT. And if you have a mess, you want a janitor who knows how to clean up every bit of it.

IBM's goal is to leverage its breadth in order to stop selling IT hardware and software and start selling business capabilities enabled by technology.

This vision is not new, nor are any of the technologies behind it. What's new is the effort to make it consistent across the vast global empire of IBM. All IBM sales representatives - from its server division to software to outsourcing - are supposed to sell on-demand along with their own products. On-demand is supposed to drive the development of all of IBM's software and hardware with the goal of making them capable of mixing with and managing applications from other vendors. The coordination challenge is huge. For example, employees in IBM's grid computing unit spend 50 per cent to 70 per cent of their time doing marketing and education inside IBM itself, according to Ian Baird, vice president of marketing and sales operations at Platform Computing, which is one of IBM's primary partner companies in grid computing.

The effort to develop the technologies necessary to fill out on-demand is no less ambitious. In 2000, IBM created a series of internal divisions, now called Emerging Businesses Opportunities (EBOs), that focus, among other things, on developing new technologies for the different pieces of on-demand, including utility (essentially pay-by-the-drink IT), autonomic (software that automatically diagnoses and fixes computer problems), grid (pooling computers to form a single virtual entity), business process integration and Linux (IBM's answer to providing a single OS that can run on any type of computer, from a mainframe to a PC).

IBM says it will spend much of its $US5 billion R&D budget on the EBO technologies in 2003. The heads of those EBOs report to the different business units that sponsor their efforts (autonomic is funded by IBM's software division, for example) and to IBM's central strategy organisation. The emphasis is on progress - not profits - but the mission, according to Thomas Bittman, research vice president of server strategies for Gartner, is for each new EBO to bring something tangible to the market within two to three years.

These efforts have already borne fruit. IBM's Websphere product is the top-selling application integration software, according to Gartner, and IBM's integration of Linux into many of its hardware and software platforms is a direct result of the Linux EBO's efforts. If there's a weakness here, according to Bittman, it's that the EBO system favours ideas that spring from within IBM over those from outsiders - a weakness it will have to address if it hopes to fill out its on-demand offerings. Meanwhile, IBM says it will spend $US4 billion this year to acquire vendors to complete its on-demand portfolio.

To some critics and competitors, IBM is luring customers into on-demand - and away from competitors - before the capabilities are ready, trusting its ability to fill in the gaps later. "They talk about freezing the market and adding new capabilities over time," says Joe Hogan, vice president of managed services for Hewlett-Packard. "We prefer to wait to announce things until they are available." (Soon after this interview, HP announced its own version of on-demand, called Adaptive Enterprise, that has as many - or more - holes to fill.) The general manager of IBM e-business on-demand, Irving Wladawsky-Berger, insists that IBM clearly explains the current state of on-demand to its customers, telling them: "Here is the vision; we'll get there incrementally." He acknowledges that on-demand is a long-term vision but says IBM is better able to solve the problems than anyone else. "The question is," says Wladawsky-Berger, "Do you start from zero or do you start from a strong base? We think we're starting from a very strong base."

Deconstructing On-Demand: What's Real, What Isn't

CIOs who'd like to keep their jobs during the next two years must add yet another task to their already overburdened schedules: managing their CEOs' expectations for IBM's on-demand package, and HP's Adaptive Enterprise and Sun's N-1 strategy, among others. In the current spending climate, CIOs who jump the gun on any of those technologies - or who let their CEOs and CFOs fire it for them - will pay dearly. The technologies for making what you have more efficient are untested, and most require consulting help. In today's economic environment, CIOs need to make sure that any experiments they undertake have clear business payback, and that's tough to do when the technologies are so immature. Here's some of what IBM's offering.

Variable pricing. The only piece of on-demand that's real today and might not cost more in the short term is variable-priced computing. The concept is simple.

Instead of buying a new server that you may not use very much, you pay only for the processing power you do use. Software monitors the server and turns on more CPUs during peak periods and turns them off during slow ones. Vendors charge a monthly fee to keep the box in your data centre and then bill you based on the average amount of processing power used each month. This is not new technology; it is, however, a new pricing scheme. Call it leasing on steroids.

But it's not always a bargain. If, for example, you use most of the capacity of the new server every month, over time the fees can add up to more than what the purchase price of the server would have been. CIOs need a clause in the contract that caps the fees at the original purchase price of the server. And, of course, the billing schemes, like the technologies behind them, work only with one vendor. IBM cannot bill you for your HP server (unless you outsource it to IBM).

Thus, even the variable pricing model falls apart when you consider trying to manage the entire infrastructure that way. But it's a start.

A piece of American Express's $4 billion outsourcing deal with IBM Global Services is based on IBM providing computing in increments of CPU power and storage capacity rather than making Amex pay for new boxes and hard drives. "The key is flexibility," says Amex's Salow. "Any good business has its ups and downs. [Variable-priced computing] lets us flex up and down a little more with the business."

Grid computing. Variable pricing is helping drive the development of grid computing, which for years seemed to be an arcane technology with little value outside of supercomputing. With business relevance at hand, IBM is rushing ahead with its grid research, but it is not yet ready for the typical corporate IT infrastructure.

According to Gartner's Bittman, useful grids today require that the computing devices (for example, PCs) be alike. They are also limited to applications that are designed for heavy parallel computing (where the processing work can be sliced up into many bits and then reaggregated). Most general business applications, like ERP and CRM, for example, do not work that way.

Furthermore, CEOs don't want grids. They want cheap pools of computing power served up to them in the same way that electricity companies serve up power. But that is a very different, and much more complex, proposition.

"The concept of a plug-and-pay electric utility model for computing is appealing to anyone who's dealing with the kind of cost pressures we're facing today," says David Dibble, executive vice president of Schwab Technology Services for Charles Schwab. "But you peel back the onion even one layer and the analogy falls apart." Right now, building an infrastructure on the scale, security and fault-tolerance levels necessary for outsourcing companies to become the electric utilities of computing is impossible, Dibble says. "The PhDs who will do it are in grade school today, I believe."

A much more likely near-term scenario is that CIOs will build small grids inside their companies to save money and resources. Indeed, Dibble has successfully piloted a small grid computing environment inside Schwab with IBM, as have other financial services companies, like JP Morgan Chase. But grid is by no means a reliable route to reducing complexity and cost in most corporate infrastructures today.

Open source. IBM's on-demand vision depends on software to connect and manage the messy infrastructures of corporate computing, but it has rarely led in software development. Instead, it has relied on acquisitions and open standards to sell the software portion of on-demand computing. In fact, IBM is killing two birds with one stone by embracing open standards like Java and the open-source Linux operating system. Open source gives IBM a weapon against its only real competitor, Microsoft, and it makes IBM look good to the IT community.

IBM makes nothing on Linux itself, but it sells the hardware and services necessary to get Linux up and running. Similarly, IBM is playing the dominant role in funding and helping write standards like the Open Grid Services Architecture, which is an open-source architecture for building grid computing applications. Open standards for grid computing help IBM sell services, hardware and grid consulting engagements, three areas where Microsoft does not compete.

Yet IBM's approach to standards contains a risk for CIOs. If IBM's influence becomes too large in these standards organisations, other vendors may not cooperate. Furthermore, standards organisations are famously slow moving. Patience will be a virtue for CIOs who want fully realised products that adhere to open standards.

And IBM is not always the white knight it portrays itself as. For example, when the open-source world treads too closely to IBM's turf, IBM's goodwill quickly ebbs. When asked about open-source alternatives to IBM's DB2 database or its Websphere application server, for example, IBM executives bristle. "We're not working on [open source] for philanthropic reasons; we're going to make money," says Dan Frye, director of IBM's Linux Technology Centre, who leads 250 IBM developers assigned to help build out the Linux operating system.

Linux is not on-demand's operating system, however. Websphere is. IBM's application server is based on the popular J2EE standard but contains enough proprietary hooks to make critics wonder whether IBM is simply moving the old Windows versus OS/2 battle higher up the corporate infrastructure and lining up against Microsoft's new application architecture, .Net. Websphere is IBM's core technology for making the highest level promise of on-demand come true: business process integration.

With Websphere, IBM wants to convince application developers to stop writing their applications at the operating system level (Windows, Unix, Macintosh) and write them to Websphere and Java instead so that they can work on any operating system. But Websphere relies on Web services, a complex and incomplete set of standards, and Java, which, while powerful, can be difficult for programmers to work with.

IBM's Big Gamble . . . and Yours

With e-business on-demand, IBM is gambling with its most valuable asset: the trust of the market.

During the past 30 years, IBM has built a positive image for its brand equalled only by such marketing masters as Coca-Cola and Nike. In a 2002 survey of 240 companies by IDC, twice as many respondents identified IBM's Global Services outsourcing group as "best-in-class" as they did IBM's nearest competitor, EDS.

"CEOs and CFOs are all about reducing risk," says Rob Schafer, program director for research company Meta Group. "In the minds of CEOs, IBM reduces risk."

Tom Kegley, vice president of IT for North America for Swiss pharmaceutical and health-care giant Roche Group, witnessed the power of IBM's brand firsthand.

When Roche's diagnostics group was considering outsourcing Web hosting for about 100 Web sites in late 2001, the finalists were IBM and Genuity. Genuity's bid was lower than IBM's, and its service received high marks from big-name clients. But Genuity was having financial problems.

Meanwhile, IBM had been building a relationship with Roche's top executives since 2000, when it hosted them at IBM's Armonk headquarters and did a mea culpa presentation about Big Blue's fall from grace and its rise under then-CEO Lou Gerstner. The meeting created a bond, says Kegley.

So when Roche executives heard about Genuity's financial problems, the handwriting was already on the wall - even though Roche had had some service problems with IBM in other deals, says Kegley. "The director of marketing said: 'Do we want the devil we know or the one we don't know?'," recalls Kegley. IBM got the job last year, and so far the agreement has worked well, Kegley says. Genuity, meanwhile, declared bankruptcy last year and was acquired by Level 3 Communications in a fire sale.

The Roche story is a classic example of the "solution sell", perfected by the father of IBM, Thomas J Watson Sr. In 1914, Watson began hiring boatloads of salesmen to build relationships with customers and learn their business problems before trying to sell them the company's machines. Today, IBM's dramatically successful services arm, IBM Global Services, built by former McKinsey consultant Gerstner in the 90s, is trying to do the same thing with on-demand.

But leading with consulting and outsourcing can get IBM in trouble with its customers - as it has in the past. Consultants from IBM Global Services (as well as all the other major IT consultancies) oversold their customers on the capabilities of ERP, CRM and supply chain software, and the e-business craze brought another wave of ill-considered enthusiasm.

If IBM oversells on-demand and overpromises on the implementation time line, it could damage that close and, in comparison with most other big vendors, trusting relationship it has with its corporate customers. Worse, with outsourcing such a big part of IBM's business now, there is a danger that on-demand will devolve into a euphemism for simply getting rid of your IT by turning it over to IBM.

IBM's Wladawsky-Berger says the goal of on-demand is to make IT more efficient and more integrated without forcing CIOs to buy new systems - no matter whether IBM runs it. "Clearly, if on-demand required you to redo everything," he says, "that would be the dumbest strategy anybody ever could have come up with."

But it's not yet clear how IBM will accomplish an on-demand future that isn't more expensive than the past.

"IBM has a vast array of enabling technologies, but they have a lot of work to do from a product architecture and vision and marketing perspective to integrate it into a single on-demand message," says David Cearley, senior vice president of product management at Meta Group. "If IBM can better coordinate the pieces and the vision, then it has a very powerful message and will be a leader in this next generation of computing. If it doesn't evolve its message, and it continues to deliver fragmented products, and on-demand becomes nothing more than marketing, then its breadth will hurt it."

What's clear is that right now e-business on-demand is not much more than a slogan. CIOs who think it's something more are looking through those magic business binoculars, darkly.

If It Looks Like an Outsourcing Deal and Walks Like an Outsourcing Deal . .

Then it probably is an outsourcing deal, even if IBM calls it utility computing.

AMERICAN EXPRESS's $US4 billion deal with IBM Global Services in February 2002 was hailed by IBM as an example of utility computing coming of age. But the Amex deal and the other deals IBM has signed since announcing its on-demand strategy last year are simply outsourcing deals with a slight twist: variable pricing for some of the computing and storage power. Much of Amex's variable pricing involves IBM simply bringing in extra boxes and hard drives without turning them on or charging Amex for them until the company needs the extra power.

That's not the same thing as plugging into the wall and having computing as a utility flow down the wire, says David Tapper, senior analyst of networked infrastructure management services for IDC (US). The computing power that IBM provides Amex doesn't come from an enormous pool shared by many customers - as electricity would be - and a utility wouldn't be assuming management over old computing systems and 2000 of Amex's IT employees. Indeed, it's hard to imagine Amex or any of IBM's big, security-conscious outsourcing customers sharing a pool of computing resources for their hundreds or thousands of different apps any time soon. "There's a huge amount of trust that needs to be built before this can happen," says Tapper. "It's like saying: 'You're going to feed and clothe me, right? You're always going to be there, right?'"

Today, utility computing is not a new technology. It's a pricing scheme, and not necessarily a bad one. Just call it what it is.

How to Manage Your CEO's Expectations

What to say when he asks: 'Where are we on on-demand?' 1. Separate the hype from the reality. "It's an interesting theory, Bill, but right now that's all it is. You see, on this grid thing, even if it were possible (which it isn't), we don't want our applications sloshing around in a big pool with our competitors' applications. And when a computer heals itself, let me know so I can call the National Enquirer." Short answer: "On-demand's not ready for prime time, Bill, but I'll keep an eye on it, and I'll keep you informed."

2. Remind him about the nature of outsourcing. "See, Bill, right now on-demand's just an outsourcing deal, and outsourcing is a 50-50 proposition. Believe me, it's not like flipping a switch - bingo - you get computing power. But I'll keep an eye on it and keep you informed."

3. Hop on variable pricing. "This is something we're looking into right now, Bill. We're not going to buy any more servers. We've got enough servers. Now we're going to start buying by the CPU and the megabyte. We can do it only with one vendor at a time right now, but we're going to look into it with each of our major vendors to see if we can cut 10 per cent to 25 per cent of our hardware costs. We don't want to become completely dependent on one vendor though, because then they've got us you know where. I'll keep you informed."

4. Tell him you're joining a standards organisation. "We're going to be part of the solution on this one, Bill. I've assigned several of my guys to this group that's working on creating standards for Web services. Remember when we discussed Web services? Well, that's what's going to make this whole thing work - some day - and we're making sure that when the standards do emerge, they'll actually work with what we've already got. But that's long term, Bill. Real long term. I'll keep you informed."

5. Ask for his help. "Actually, Bill, you can help me here. The next time a vendor tries to sell you anything new, blow your top. Tell them how you already have 300 servers from six different vendors. Make them tell you how they're going to make their server work with everyone else's server without making us pay more. They owe us - they owe you - an explanation. "And how are the kids?"

How Practical Is Grid Computing?

Unless you are comfortable manipulating source code, not very.

GRID COMPUTING is following a well-worn path previously travelled by any number of world-beating technologies such as the Internet and open-source software.

They start in academia and gather strength in the obscurity of the research world. Then they're adopted by leading-edge businesses, from where they emerge into the mainstream. Grid computing has reached the third step. Those fearless Ferraris of corporate computing, financial services companies, are beginning to use grids for real business applications that will have real impact on customer service, internal productivity and computing costs.

At JP Morgan Chase's investment banking division, Steve Neiman, head of information architecture, is building a grid with the help of IBM and grid computing specialist Platform Computing that combines a number of different Morgan product applications, such as foreign exchange and equity, that used to exist separately. Before the grid, if employees wanted to sell these products in combination, they had to coordinate the computing transactions in isolation and combine them manually. But by rewriting the applications to run on a single grid architecture, traders can now coordinate all the products to deliver them faster, with fewer IT people needed to manage the applications. Neiman expects to see a 40 per cent reduction in overall IT costs for the applications.

Neiman can do this for a number of reasons. First of all, he can count on the support of JP Morgan Chase employees who know that better IT butters their paycheques. He doesn't have to worry about "server huggers", users reluctant to turn their precious servers over to a grid where someone else's applications and computing needs might threaten their own.

Most important, Neiman doesn't have to worry about getting access to the source code of the applications he's painstakingly rewriting to work with a grid because JP Morgan Chase writes all its own applications and has some of the world's best programmers. Most CIOs deal with packaged apps and wouldn't know how to begin to rewrite them to work on a grid. That's the showstopper for grid right now, says Neiman.

"Until the vendors write the applications for grid, widespread adoption won't happen," he says. "It's probably impossible to take someone's product and put it on a grid. You're asking for trouble. We tried that on a small scale, and we decided it was a very hard pull."