Linux Predictions Revisited: From Torvalds’ Hobby Project to the Windows Desktop Reality

The useful way to revisit these 1990s claims is to ask what each one actually forecast: the size of a project, the fate of a kernel design, success on everyday PCs, or competition with established Unix products. Those are four different scorecards.

The hobby outgrew its original PC​

Torvalds’s August 25, 1991, post to the comp.os.minix Usenet group sought feedback on a project still taking shape. Its target was specific: Intel 386/486-compatible PCs. Calling it a hobby described the project as he saw it then, not a measured prediction about how many people might eventually contribute to it.

The hardware boundary certainly did not hold. Linux kernel documentation now includes architecture-specific material for ARM64, PowerPC, RISC-V, s390 and others alongside x86. That is evidence of a much wider reach than the original PC target—not a promise that every Linux distribution or device supports every processor equally. For Windows and PC readers, it is a reminder that Linux is a kernel spanning many kinds of hardware, not just a desktop operating system competing for a place on a laptop.

The verdict: The modest description aged badly as a forecast of scale; the initial hardware limitation aged even worse.

“Linux is obsolete” was an engineering argument​

On January 29, 1992, Minix creator Andrew Tanenbaum opened a Usenet debate titled “LINUX is obsolete.” His objections concerned Linux’s monolithic kernel design and its close connection to x86 hardware. Tanenbaum favored a microkernel approach, which places more operating-system functions outside a small central kernel.

Torvalds’s defense was more practical than triumphant: Linux let people run useful software on the hardware available to them. The thread also included contributors arguing that availability and the ability to work with familiar Unix software mattered. In retrospect, Linux’s spread shows that its original architecture did not make it obsolete. It does not prove that microkernels are inherently inferior; that would be a much broader technical claim than this history can settle.

The verdict: Tanenbaum identified real design trade-offs, but his prediction that Linux’s design made it a dead end did not describe what followed.

The desktop prediction remains the awkward one​

The recurring “Year of the Linux Desktop” joke survives because Linux’s success elsewhere does not answer the question Windows users usually mean: has it become the standard choice for general-purpose PCs?

Valve’s August 2026 Steam Hardware & Software Survey offers one carefully bounded snapshot. It reported Linux at 3.90% and Windows at 93.95% among surveyed Steam customers. Participation is optional, and Steam users are not a census of all PC owners. The figures therefore cannot settle Linux’s overall desktop market share, but they do show why progress in PC gaming should not be mistaken for Windows having been displaced.

For someone considering a move from Windows, the practical question is smaller and better: do the applications, games and peripherals you rely on work as you need them to? A platform can be a good fit for an individual without winning the whole desktop.

The verdict: Linux is a viable desktop choice for some users. The sweeping desktop-takeover prediction remains unfulfilled.

Microsoft’s Unix forecast was narrower than it sounds​

The leaked 1998 Microsoft memorandum later published with Eric S. Raymond’s annotations said Linux was “on track to eventually own the x86 UNIX market.” That is a notable assessment from a competitor, and it is much more precise than saying Linux would kill all proprietary Unix—or every other operating system. The memorandum was a forecast, not a market-share study proving when or how thoroughly that outcome occurred.

It also shows how far the conversation had moved in seven years. The 1991 question was whether a hobby project for an Intel PC would be useful. By 1998, the question inside Microsoft was how open-source development and Linux might compete in an established operating-system market. That change in what observers thought worth predicting may be the most revealing result of all.

The lesson for today’s Windows-versus-Linux debates? Specify the arena before declaring a winner. Hardware reach, kernel architecture, Unix competition and desktop adoption are not interchangeable measures of success.