The two most influential technology corporations in the world agreed to build the future together—and tore up their alliance so badly that one of them lost the personal computer market forever.
1980. Bill Gates signed a contract with IBM to supply an operating system that Microsoft didn't have yet: they had to license it first, then buy it from Seattle Computer Products along with the rights to 86-DOS. The deal looked modest—Gates paid $75,000 for full rights. But the agreement with IBM had a clause that determined the balance of power for two decades: Microsoft retained the right to license the system to other manufacturers. IBM got an operating system for its PC, Microsoft got a monopoly on software for the entire emerging clone industry.
By the mid-eighties DOS had become the de facto standard, but the architecture had hit a ceiling. Single-user mode, no memory protection, dependence on real processor mode—all of this made the system unsuitable for corporate tasks. Intel released the 80286 with protected mode, opening the path to multitasking and reliability, but DOS couldn't use it. Business demanded a system that didn't crash from one frozen application and could work with multiple programs simultaneously.
1985. IBM's headcount stood at 405,000 employees—a giant corporation controlling the corporate mainframe market and confidently moving into the personal computer segment. Microsoft had just over a thousand people, but owned a critically important asset: the source code of the operating system on which millions of computers depended. Both sides understood: to hold the market, they needed a new platform. That's how the agreement to jointly develop OS/2 was born—a system that was supposed to displace DOS, cement IBM's dominance in hardware, and ensure Microsoft's dominance in software.
The partnership looked natural. IBM brought experience creating corporate systems, engineering resources, and control over standards. Microsoft brought knowledge of the PC market, startup flexibility, and understanding of how to write systems for mass users. OS/2 was conceived as the successor to DOS, designed for 80286 protected mode: preemptive multitasking, memory protection, stability. A system for business that doesn't forgive failures.
But cracks appeared at the start. IBM saw OS/2 as a controlled platform for its own hardware, where control over APIs and licenses would let them dictate terms to manufacturers. Microsoft saw in the project an opportunity to cement its monopoly on operating systems for all PC-compatible machines. Both sides wanted the future, but envisioned it differently: one as a vertically integrated ecosystem, the other as a horizontal platform with Microsoft as the sole software supplier.
December 1987. The first release of OS/2 version 1.0 shipped (announced back in April, deliveries began ahead of the promised date). The system worked exclusively in text mode, required a minimum of 1.5 MB RAM (when the standard PC configuration had 640 KB), supported only the 80286, and couldn't run most popular DOS applications in protected mode. Multitasking was there, memory protection too, but to users accustomed to DOS, the system seemed slow, demanding, and alien.
The corporate clients they were betting on weren't rushing to migrate. Migration meant buying new hardware, rewriting internal applications, and retraining personnel. Developers looked at OS/2 cautiously: the API differed from DOS, documentation was raw, and development tools were half-baked. IBM promoted the system as an industry standard, but the market wanted compatibility with existing programs, not a standard.
A year later the Presentation Manager graphical shell appeared—a response to Windows and Macintosh. The interface was well thought out, but only worked on 80286 in protected mode, required even more memory, and conflicted with most DOS drivers. OS/2 was becoming technically stronger but moving further from the mass market. IBM was building a system for the future, but users lived in the present, where tens of thousands of programs were written for DOS and hardware upgrades cost money.
Tension grew inside the alliance. Microsoft insisted on supporting the 80386 and moving to 32-bit architecture to use the full potential of new processors. IBM believed the 286 was far from exhausted, and rushing to 386 would fragment the platform and scare off corporate clients. The dispute wasn't just about technology: Microsoft wanted to accelerate system development to capture the market before competitors, IBM sought to control the pace of evolution, retaining the right to approve specifications.
In parallel, Microsoft continued developing Windows—a graphical overlay on DOS that didn't require protected mode, ran on most existing machines, and launched all old programs without rewriting. Officially, Windows was positioned as a temporary solution for the mass market until OS/2 matured for wide use. But each new version of Windows became more functional, and OS/2 more resource-hungry.
May 1990. Microsoft releases Windows 3.0—and the personal computer market flips. The system ran on any PC with 80286 and above, supported extended and protected modes, had a well-designed graphical interface, and ran the entire base of DOS applications. The first million copies sold in the first months, and the system ultimately sold ten million. Developers who had been writing for DOS for years got an API that let them add graphics and multitasking without completely rewriting code.
This wasn't sabotage of the joint project, but a public strategic pivot. Microsoft stopped pretending that Windows was a temporary measure. The company invested its core engineering resources into developing a system that competed with OS/2 for the same developer and user dollars. Bill Gates explained the decision pragmatically: the market voted for compatibility and accessibility, not architectural purity. OS/2 might be technically superior, but Windows worked here and now.
IBM perceived the launch of Windows 3.0 as betrayal. The corporation had invested hundreds of millions of dollars and thousands of man-hours in joint development, expecting OS/2 to become the only next-generation system. Instead, their partner deployed a competing product, using developments and knowledge gained during joint work. IBM lawyers studied the contracts looking for grounds to sue, but formally Microsoft hadn't violated anything: obligations for OS/2 were being fulfilled, and Windows was developed in parallel and wasn't prohibited by the agreement.
1990–1991. The Federal Trade Commission (FTC) began investigating the IBM and Microsoft collaboration for suspected anticompetitive market division: OS/2 for business, Windows for consumers. The formal pretext was a joint declaration from November 1989 in which the companies promised to "develop a consistent family of systems software" for the nineties—regulators saw this as market carve-up. But by the time the investigation gained momentum, the partnership was already falling apart. Microsoft finished work on OS/2 1.3, handed over the code to IBM, and focused on Windows.
IBM continued alone. The company released OS/2 2.0—a fully 32-bit system with 80386 support, improved multitasking, built-in support for running DOS and Windows applications in separate virtual machines. Technically it was a breakthrough: stability, performance, memory protection at a level unattainable for Windows 3.x. But the moment had passed. Windows already controlled the mass market, developers were writing for Win32 API, and corporate clients saw no reason to pay more for a system that at best was catching up to the competitor.
July 1994. Chicago, negotiations between IBM and Microsoft. Participants: Tony Santelli from IBM and Joachim Kempin from Microsoft. On the table—the future of relations between the companies and the terms of preinstalling operating systems on IBM machines, which still remained one of the world's largest PC manufacturers. Microsoft proposed a "frontline partnership" modeled on the Compaq deal: discounts on its products and better technical support—but conditional on abandoning OS/2 Warp. IBM responded with a counter-scenario: dual-boot on all computers with 8 MB RAM and above—OS/2 Warp with the option to boot into Windows 3.11.
The financial side was tough. Microsoft's royalty for MS-DOS tools was $2 per copy, for Windows 3.x—$9. This agreement was valid until September 1997, and each machine shipped without Windows meant a loss of nine dollars in pure revenue for Microsoft. IBM insisted that OS/2 Warp was a technically superior system that deserved a chance to compete on equal terms. Microsoft responded that the market had already made its choice, and refusing to preinstall Windows would only lead to users installing it themselves by buying the boxed version.
November 1994, Comdex. IBM officially notified Microsoft of its decision to preinstall OS/2 Warp on all PCs with 8 MB RAM or more with dual-boot capability for Windows 3.11. This was the last major attempt to change the balance of power. IBM launched one of the most expensive marketing campaigns in its history—the corporation ultimately spent over $1 billion promoting OS/2 overall, including prime-time advertising and large-scale campaigns for developers and retail.
The result was modest. Consumers didn't want to deal with two systems, developers saw no incentive to port applications to a platform with less than 10% market share, and corporate clients preferred not to mess with a system the main software producer didn't support. OS/2 Warp got positive press reviews for stability and performance, but laudatory articles didn't convert to sales.
Platform war economics worked against IBM. The more users chose Windows, the more developers wrote for it. The more applications came out for Windows, the less sense there was in switching to OS/2. Even built-in support for Windows programs worked against it: users ran the software they needed in emulation without thinking about native OS/2 applications. The system became an expensive overlay on Windows that ran slower than the original.
OS/2 was technically superior to Windows throughout almost its entire lifecycle. Protected memory prevented one application from crashing the entire system. Preemptive multitasking distributed processor time fairly, not allowing one program to freeze the interface. The HPFS file system surpassed FAT16 in performance and reliability. Built-in support for network protocols made OS/2 ready for corporate environments without installing additional drivers.
But the personal computer market doesn't work by the laws of engineering perfection. Windows 3.x crashed from an incorrectly written driver, froze when simultaneously launching heavy programs, and required rebooting several times a day. But it ran WordPerfect, Lotus 1-2-3, AutoCAD, and thousands of other applications users had already paid for. OS/2 offered better architecture but worse compatibility—and lost.
Dependence on DOS and Windows applications became a trap. IBM spent enormous resources on emulating someone else's platform instead of focusing on creating its own ecosystem. Each new version of Windows broke compatibility, forcing IBM to fix emulation. Developers saw that even the OS/2 producer was investing in Windows support and drew a conclusion: why write natively if emulation works.
The conflict within the alliance made the problem worse. IBM and Microsoft couldn't agree on a unified architecture, development priorities, and profit sharing. Each side pulled the project in its direction, slowing decision-making and blurring focus. Brand confusion—OS/2, OS/2 Extended Edition, OS/2 Warp—prevented users from understanding how versions differed from each other. Microsoft turned Windows into a clear product line with understandable improvements in each version. IBM released editions with different feature sets, prices, and hardware requirements, confusing even technically savvy buyers.
By the mid-nineties it became clear: OS/2 wouldn't win. IBM continued supporting the system until the early two-thousands, finding a niche in ATMs, cash register terminals, and industrial systems where stability matters more than mass adoption. But the dream of dominance in the personal computer market was buried. Windows 95 definitively cemented Microsoft's dominance, uniting DOS and Windows into a single system with 32-bit architecture, preemptive multitasking, and an interface that became the standard for twenty years.
The OS/2 story isn't a story of bad technology. It's a story of how control over the ecosystem, compatibility with the installed base, and the ability to manage developers turn out to be more important than architectural elegance. IBM didn't lose because OS/2 was worse. It lost because Microsoft understood: in platform wars, the winner isn't the best system but the one people write programs for. And while IBM was building the perfect operating system, Microsoft was building an ecosystem where it was profitable for developers, users, and hardware manufacturers to stay.