The x86 Duopoly & CPU FranchiseNarrow moat
Advanced Micro Devices (AMD) — moat facet
The one part of AMD that is owned, not earned: a two-company license to make the world's PC and server brains, which kept it alive when nothing else did.
The bedrock of AMD's moat — the thing that kept it alive through the lean years and anchors it today — is its position as one of only two companies on earth permitted to make x86 processors. The x86 instruction set architecture is the software-hardware contract that virtually all Windows PCs and the vast majority of the world's servers are built around: the operating systems, the applications, the entire installed base of enterprise software assumes an x86 chip underneath. And through a web of cross-licensing agreements dating back decades, only AMD and Intel may legally build those chips2. This is a genuine, durable structural barrier, and it is the widest single element of an otherwise narrow moat.
The importance of the duopoly is best seen in what it prevented. Through the years when Intel outspent and out-executed it, when its stock traded near two dollars and bankruptcy was whispered, AMD was never actually displaced, because there was no third x86 maker to take its place — the license itself was a moat that held when nothing else did. A newcomer with unlimited capital still cannot simply enter the x86 market; it lacks the license, and the installed base of x86 software makes any non-x86 alternative a costly migration rather than a drop-in substitute. That scarcity is why AMD's CPU business, even at its weakest, retained real value.
Within the duopoly, the story since 2017 has been one of relentless share gains, powered by the Zen architecture. Before Zen, AMD was a distant, declining second3; since Zen, it has taken share from Intel in every segment, generation after generation, as it executed a disciplined annual cadence of improvements while Intel stumbled. The most valuable gains have come in the data center, where AMD's EPYC server processors have climbed from a rounding error to roughly 40% of the market1 — a stunning reversal in the highest-margin, stickiest corner of the CPU world, where reliability and a multi-year track record matter most and where AMD has now firmly established itself as a first-choice supplier rather than a risky alternative.
The franchise spans three end-markets. Server CPUs (EPYC) are the crown jewel: high-margin, sticky, and the beachhead from which AMD sells its data-center GPUs too. Client CPUs (the Ryzen brand) power desktops and laptops, a large but mature and cyclical market where AMD has built a strong premium brand among enthusiasts and gained laptop share. And the architecture cadence itself — Zen 1 through Zen 5 and beyond, delivered on a reliable yearly rhythm — is the engine that keeps the share gains coming.
Two cautions apply. First, this is a duopoly, not a monopoly: AMD shares the x86 franchise with Intel, which remains the larger player with deep enterprise relationships and could yet recover its footing. Second, and more profound, the x86 architecture itself is no longer unchallenged. The Arm architecture — long confined to phones — has broken into laptops (Apple's own silicon) and, increasingly, the data center (Amazon's Graviton, and other hyperscaler chips), threatening the very foundation the duopoly rests on. The x86 license is a real moat, but it protects a castle whose surrounding territory is, for the first time in decades, being contested. Still, x86 remains dominant in servers and PCs today, AMD is taking a growing share of it, and the duopoly license remains the most durable advantage the company owns.
Widening. Zen's disciplined cadence has taken share from Intel generation after generation, above all in the high-margin server market — the franchise is getting stronger, even as Arm nibbles at x86 itself.
The CPU thesis is share taken from Intel in the market that matters most, and EPYC's climb from a rounding error to ~40% of servers is one of the great share shifts in chip history. Watch it: the gains came from a window of Intel weakness, and a resurgent Intel or Arm servers could stall or reverse them.
Source: Industry data (Mercury Research et al.) ↗- Third-party estimateEPYC's server share: from a rounding error to roughly 40%.Mercury Research x86 server-CPU share estimates — EPYC's climb from <1% (2017) to ~40% by 2025 — 2017-2025 · publ. quarterly · source ↗
- ReportedThe x86 cross-license limits the architecture to AMD and Intel.AMD–Intel x86 cross-license agreement (renegotiated in the 2009 settlement) — only the two companies may legally build x86 processors — 1976-2026 · publ. November 2009 · source ↗
- ReportedThe Zen architecture launched in 2017, starting the share-gain run.AMD Zen architecture launch (Ryzen, March 2017; EPYC, June 2017) — the start of the share-gain run against Intel — 2017 · publ. 2017 · source ↗