(Business 2.0) – The most important software executive you've never heard of works for a company you certainly have heard of--except you've probably never heard it called a software company. Unlike other top software execs, Steve Mills, head of the software group at IBM, has no charitable foundations or tennis tournaments named after him and captains no America's Cup racers. In a field legendarily populated by nonconformists and visionaries, Mills seems like a throwback to tech's buttoned-down past--a 29-year veteran at a place where, not so long ago, it was an act of rebellion to show up to work with anything other than a white shirt under your tie.

By now every first-year business school student can recite how former CEO Lou Gerstner rescued IBM in the 1990s by transforming it from a company that sold computers to one that sold computer advice. Fewer people noticed that an equally remarkable rebirth took place in the company's software business. Today IBM's fortunes ride, far more than most people realize, on Mills and the products he oversees. If his $13 billion division were a stand-alone company, it would be the second-largest software firm in the world. Though it accounted for only 16 percent of Big Blue's $81 billion in 2002 revenue, it generated 31 percent of the company's pretax income and 78 percent of its bottom-line growth.

What's most important, though, is the role Mills plays in the future. Information technology's next productivity miracle, if there is one, will most likely come out of enterprise infrastructure software, or "middleware"--IBM's software specialty. (In IBM's world, middleware is the layer of software that sits between the operating system, which talks to the machine, and the applications, which talk to end users.) Every buzzword used to describe the corporation of tomorrow--"real-time," "agile," "transparent"--is just shorthand for this technology's potential to tie together an organization's scattered pockets of knowledge and share them over the Internet with managers, customers, partners, employees, vendors, and anyone else who can benefit from them. When Mills's boss, CEO Sam Palmisano, talks about IBM's own vision of "on-demand" computing--in which companies buy digital muscle only as needed, with no more fuss than buying electricity from the local utility company--he is talking about a future made possible largely by Steve Mills and his team.

That is, if it's not made possible by Microsoft instead.

The Redmond giant knows that the next great opportunity lies in the guts of corporate computing, where IBM lives. The inevitable confrontation has the aura of a title fight: the monopolist of the 1990s dueling with the monopolist of the 1980s over the first huge prize of the 21st century. "What's at stake," says Corey Ferengul, vice president of tech research firm Meta Group, "is who controls corporate computing."

It's a contest that pits technology's richest, most ferocious competitor against a taciturn corporate lifer and an organization that a few years ago was regarded as a joke by most of the industry. The amazing thing is, it figures to be a pretty good match.

Steve Mills is not a tall man, and neither his waistline nor his hairline do much to disguise his 51 years. In person, he has the air of a debater who believes he's already won the argument, even if his opponent has yet to concede. People who've worked with him describe him as meticulous and driven, a loyal but demanding boss, and such a passionate advocate for his products that he could, as one puts it, "convince you that IBM database software cures erectile dysfunction."

Perhaps the best insight into the man's character hangs on the wall of his office in the software group's headquarters in Somers, N.Y. It's a poster of the film The Shawshank Redemption, and it captures the moment when Tim Robbins bursts into freedom, having tunneled his way out of prison--with a handheld rock hammer. For Mills, it's an allegory for a career spent peddling some of technology's most abstruse products and negotiating the industry's most byzantine bureaucracy. "Brilliance is an after-the-fact perception created by success," he explains. "There's no fast path, no magic formula. It's all hard work and perseverance."

Be that as it may, anyone charged with digging out IBM's software business in the 1990s might have asked for a backhoe rather than a hammer. Thanks to the company's monopoly in mainframes, the division actually generated nearly $11 billion in revenue--almost three times what Microsoft was pulling in at the time--but by any measure, it was headed for disaster. Virtually all of its products were designed for IBM big iron; the division didn't even have its own sales force. As for focus, there was none. The group listed 4,000 often-overlapping products created in 30 different laboratories around the world. Arguably, the only growth products in its portfolio were internal rivalries, wasted resources, and bewildered customers.

When Gerstner arrived in April 1993 and began to defibrillate the company, Mills was running the division's Santa Teresa Laboratory south of San Jose. In August he was promoted to run Software Solutions, a business unit that sold database and application-development software. After barely a month on the job, he called his top managers to a meeting at a Greenwich, Conn., hotel and told them the software's subservience to hardware was going to end. They would focus on their database software, called DB2, and they would develop it not solely for IBM boxes but also for the fast-growing market of computers running Unix and Windows. Mills concluded by saying he wasn't interested in the Big Blue custom of building consensus. "You can go my way or the highway," he said. To secure funding, Mills went straight to Gerstner and promised he'd turn software into a business that could stand on its own. "At that time, for someone of his stature in the company to take that much personal risk was unheard of," recalls a former IBM executive. Gerstner agreed to spend billions and back Mills when he sought help from IBM's research labs and consultants. DB2 came out for Unix computers by Sun Microsystems and Hewlett-Packard in 1995 and for Windows operating systems a year later.

That internal triumph came as IBM was making its single most important decision of the decade: to embrace Sun's Java programming language. IBM was among the first to realize that the future lay in linking the islands of incompatible hardware and software that held most of every company's information. A group of strategists and software executives, including Mills, considered various options, from writing their own open language to doubling down on IBM's floundering desktop operating system, OS/2. A third option was to sign on to an unproven language called Java that had just been developed by Sun.

For IBM traditionalists, the idea was apostasy. Java was designed to run on any operating system--and, thus, any hardware, including that of other manufacturers. Also, since the new platform would need bulking up to handle corporate data loads, embracing Java would mean that IBM would have to commit thousands of its own engineers to improving a competitor's product. Throughout the debate, Mills calmly worked the halls, focusing the conversation on the common goal--linking disparate systems together. Eventually the opposition realized there was no other way. "We wished we had created Java," Mills says, "but Sun did. We went with it anyway."

From there on, the might of IBM was arrayed behind the new standard, making Java essentially the lingua franca of enterprise computing. Big Blue spent hundreds of millions of dollars helping its rival develop the industrial-strength version of Java, known as J2EE. IBM's programmers now develop most new software on the platform, as do 51 percent of programmers everywhere. "You can't underestimate the role IBM played," says Ted Schadler, a software analyst at Forrester Research. "It was the first to start contributing intellectual property and ideas and developers; it made Java its corporate language."

Having adopted Java, IBM was now free to team up with other Java-compliant vendors. For example, why shouldn't IBM consultants recommend J.D. Edwards financial applications to customers, as long as J.D. Edwards recommended, say, DB2 to its clients? There was just one hitch. Among IBM's 4,000 tangled software product lines were applications that competed with just about every potential partner. None of them made much money, but each had a constituency within the company. Mills, by then No. 2 to software group leader John M. Thompson, didn't care. "The last time I checked, people get angry if you compete with them," Mills says. His logic: IBM was no good at applications; it could make more money teaming up with partners; therefore, it should dump the apps business. By 1999 he'd gotten his way.

The effect was electrifying. Three and a half years after quitting applications, IBM has cross-marketing agreements with 9,000 application vendors. Last year, partnerships drove an estimated $12 billion in additional sales across the company, compared with $3 billion in 2001. "I look at IBM and us as a virtual software vendor," says Lenley Hensarling, group vice president at J.D. Edwards, IBM's largest applications partner. "We work together like one company, offering a consistent value to our customers."

One other benefit: Fleeing applications left Mills free to concentrate on the category in which IBM has the most credibility: middleware. Databases, transaction processing, computer-to-computer messaging, systems management--the complex but invisible tasks handled by middleware are the kinds of jobs for which IBM has been writing software since the mainframe days of the 1970s. "In a way, you can say we've been building middleware for 30 years," says Ambuj Goyal, one of Mills's top lieutenants. "It's in our genes." That's a profitable heritage to have. Aberdeen Group estimates that the middleware market, which had $78 billion in sales last year, could reach $115 billion in 2006. In the current tech-spending drought, those are astonishing numbers.

When Thompson was promoted to vice chairman in 2000, Mills inherited the top job at the group he had done so much to rebuild. (Thompson retired in September.) Instead of thousands of software products, Mills now mainly oversees four complementary lines of Java-based middleware: DB2, which stores and retrieves data; WebSphere, which helps applications work together over the Internet; Lotus, which handles Web-based collaboration and e-mail; and Tivoli, which manages networks. (See "IBM vs. Everyone," page 82.) Mills now has a small army of salespeople: The software group has 10,000 of its own, plus those who work at its 9,000 strategic partners. There are also the sales teams from IBM's other divisions--the legions in hardware and the 150,000 consultants from IBM's services group--whose products often come bundled with IBM middleware.

When those resources are working in concert, Big Blue can be irresistible. When JP Morgan Chase agreed in December to outsource all of its computing to IBM--a $5 billion deal--it said IBM's size and stability were important factors, as had American Express when it gave IBM a similar $4 billion contract in February 2002. But JP Morgan also said IBM's on-demand computing technologies--built on IBM middleware--were equally important.

For all of Big Blue's heft, none of Mills's product lines dominates its category. DB2, for example, owns only a quarter of the Unix database market; Oracle remains the big kahuna there. Even WebSphere, Mills's most successful line, probably only splits the market with rival BEA Systems's WebLogic. Why? The legacy of the group's awkward 1990s phase has been difficult to expunge. Some products still show signs of being pieced together from that era's jumble of product lines. "The fact is," says Dwight Davis, a software analyst at Summit Strategies, "they don't always work together that well." Also, since Mills's middleware was built for Fortune 500 companies' complicated IT departments, it can't be installed without a certain amount of tweaking--typically from IBM consultants at upwards of $200 an hour. Competitors ask cynically whether complexity isn't part of IBM's business model. "If the software weren't complex," asserts Shahin Kahn, the chief competitive officer at Sun, "they couldn't sell you the services of 25 consultants to help you get running."

Mills get prickly on the subject. "I've never met a customer who wants to buy software," he contends. "They want to deploy software, and they want to know who's going to be there after they deploy it." Still, he knows that his products' complexity is a problem. The fastest-growing market for enterprise software is in small and midsize businesses. Few such customers need industrial-strength middleware. Fewer still can afford to hire a platoon of consultants to make it work.

Which brings us to Microsoft. The behemoth dominates business computing among small and midsize companies with an approach that's the mirror image of IBM's. Rather than tying together many disparate applications solely with middleware, Microsoft creates a soup-to-nuts Windows environment, supplying everything from apps (like Microsoft Office) to middleware (like Microsoft Exchange) to operating systems (like Windows Server 2003), all of which work together right out of the box. You must use the Microsoft version of whatever software you need, but you know it won't take weeks--and huge consulting fees--to set up. 7-Eleven, a longtime Microsoft client, spent six weeks working on a J2EE implementation for a Web-based supply-chain system before giving up and going with Microsoft software to connect to its vendors.

Microsoft's answer to Java is a group of software technologies collectively known as .Net. One of Microsoft's traditional strengths has been its popularity with software developers. (The more people creating applications for your platform, obviously, the more desirable it is.) That carries over with .Net. Microsoft's .Net developer tools are winning over programmers, and that's already starting to have an effect. Despite a direct plea from Mills not to "drink their Kool-Aid," Siebel Systems--one of IBM's earliest and biggest Java partners--announced in October that it will make its applications available on .Net as well as J2EE.

Less than two months later, Mills spent $2.1 billion to buy Rational Software, a maker of development tools. The move was a clear attempt to lure programmers away from the Microsoft camp. This is a contest Mills can't afford to lose. "Win the hearts and minds of developers, and everything else follows," says Mike Gilpin, an analyst at Giga Information Group.

While direct competition between IBM and Microsoft has been sporadic, it will intensify. IBM is poised to move off its large-company power base and bid for midmarket business, where it will collide with Microsoft going the other way. It will be a battle between two very different visions of corporate computing: .Net vs. Java, out-of-the-box simplicity vs. breadth of selection. IBM is not Microsoft's only competitor, but it casts the biggest shadow. And it's the only one whose CEO has pledged to spend $10 billion during the next half-decade on an alternative vision. "In terms of footprint, cash, and product approach," says Goldman Sachs software analyst Tom Berquist, "IBM is the most legitimate competitor to Microsoft."

Eric Austvold, research director at AMR Research, puts his money on Microsoft. To win in the midmarket, IBM will have to lower its prices and make its software simple enough to be usable without consultants. "That requires a substantial rewrite of where IBM is today," he says. "It's going to be a lot easier for Microsoft to go up the ladder than for IBM to go down."

But Mills is working on it. In September he introduced WebSphere Express, a simplified version that can be installed in five clicks. DB2 Express, a slimmed-down database that runs on Linux and Windows, starts at $1,000. And in case the market had any doubts that the chiefs of Mills's four core product lines were operating in unison, all four have been careful to say publicly that "this is Steve's organization, not mine." Says Meta Group's Ferengul, "The uniformity of that message was striking" at a November IBM analysts meeting.

Getting your managers to sing in tune is not quite the same as getting your middleware in perfect harmony. But one blow of the rock hammer at a time. The campaign against Microsoft will be long, and Mills is a patient--and self-assured--man. "There's probably a path for Microsoft to be successful and profitable, maybe," he says. "But they won't be as dominant as they have been. Hey, life's a bitch."