Dead End: OS/2
Abstract
OS/2 was technically better than Windows. That is not a contested thesis, it is the consensus of contemporaries and historians. It was more stable, more secure, real multitasking on consumer hardware. And it still lost completely. The story of OS/2 is a lesson in why technical superiority is often the weakest argument in a platform war.
The Partnership: IBM and Microsoft, 1987
In 1981, IBM had made the mistake of handing Microsoft control over the operating system of the IBM PC. MS-DOS was IBM’s product, but Microsoft kept the licensing rights and was free to sell it to anyone. Compaq built IBM-compatible PCs. Microsoft licensed MS-DOS to Compaq. IBM lost control of its own standard.
IBM set out to correct that mistake. In August 1985 IBM and Microsoft signed a Joint Development Agreement; with OS/2, the next PC operating system was to be under IBM’s control, with Microsoft as a partner in a clearly defined role.
OS/2 version 1.0 (December 1987) ran in the protected mode of the Intel 286, with real memory separation between processes. MS-DOS ran in real mode: all programs shared the same address space, and one crash could take down the entire system. OS/2 was architecturally safer.
The problem: version 1.0 had no graphical interface. Presentation Manager (PM), OS/2’s GUI, only arrived with version 1.1 (October 1988). It was complex to develop for, poorly documented, and demanded more hardware than most users owned.
The Split: Microsoft Bets on Windows
Behind the scenes a tension grew that Microsoft decided to resolve in 1990.
Windows 3.0 (1990) was MS-DOS with a graphical shell. It had no real process separation and was technically inferior to OS/2. And it sold phenomenally: two million copies in the first six months, about ten million by the time Windows 3.1 arrived two years later.
Microsoft drew the conclusion: users wanted a GUI, not a technical architecture. Windows gave them a GUI on hardware they already owned. OS/2 demanded more RAM, more CPU, more money.
In 1990, Microsoft acted on it. The development team under Dave Cutler, which had been building the next-generation OS/2, was redirected to Windows NT, a new operating system with a modern architecture but a Windows face. The partnership with IBM unraveled in 1990; the divorce was final by 1992.
IBM was left with OS/2 alone.
OS/2 Warp: The Last Attempt
IBM did not give up. OS/2 Warp 3 (1994) was a serious effort to win the consumer market. IBM bought television advertising and ran a large TV campaign in 1994 that positioned Warp as a consumer-friendly operating system. IBM built its own internet service (IBM Internet Connection) into the operating system before Windows did.
Warp was good. More stable than Windows 3.1, real multitasking, protection against crash cascades. Developers praised the architecture. In certain niches (ATMs, cash registers, kiosk terminals) OS/2 remained dominant for decades. Cash machines in many parts of the world were still running OS/2 in 2010.
But Windows 95 (1995) made OS/2 obsolete on the consumer desktop. Windows 95 brought a 32-bit system (for most applications), a reworked GUI, and Microsoft’s entire distribution network. Above all: Windows 95 ran MS-DOS programs. OS/2 could do that too, but users did not know it, and the compatibility story was complicated.
Warning
The application death trap:
OS/2 suffered from the classic chicken-and-egg problem of every platform. Users did not buy OS/2 because there were too few applications. Developers did not write OS/2 applications because there were too few users. Microsoft actively sharpened this problem: developers enrolled in Microsoft’s programs received Windows development tools early and OS/2 tools late. OS/2’s technical superiority helped nothing as long as Lotus 1-2-3, WordPerfect, and most business applications were developed primarily for DOS and Windows.
IBM tried to market Win-OS/2 as the answer: a Windows compatibility layer that ran Windows 3.1 applications in a virtual machine. It worked, but it confirmed Windows as the real standard. Whoever needed Windows applications ultimately needed Windows.
The Technical Legacy
OS/2 disappeared as a consumer product. Its technical concepts survived by detours.
Dave Cutler’s team at Microsoft used concepts from the OS/2 work for Windows NT (1993), the operating system that founded the NT line from which Windows XP, Windows 7, Windows 10, and Windows 11 descend. OS/2’s architectural DNA lives on in modern Windows.
eComStation (2001) and ArcaOS (2017) continued OS/2 under different branding, licensed from IBM, maintained by small companies, used by niche users and industrial systems. ArcaOS is still actively developed and supports modern hardware.
The actual lesson: OS/2 did not fail on bad technology. It failed on bad marketing, an expensive sales channel, Microsoft’s decision to dissolve the partnership, and the momentum of Windows 95. Platform wars are decided by network effects: whoever reaches critical mass first wins, regardless of technical superiority.