AMD and the x86 Underdog

Zusammenfassung

For most of its history, Advanced Micro Devices was the perpetual second to Intel — founded in 1969 by Fairchild defectors, kept alive as a legally mandated “second source” for the chips IBM and others refused to buy from a single supplier. Twice AMD broke out of that role: in 2003, when its Opteron and Athlon 64 chips extended x86 to 64 bits and forced Intel to copy its design; and again after 2017, when CEO Lisa Su’s “Zen” architecture and a radical chiplet strategy turned a near-bankrupt company into Intel’s equal and Nvidia’s chief AI rival. AMD is the rare underdog that, twice, set the direction the leader had to follow.

Eight Men from Fairchild

Advanced Micro Devices was founded on May 1, 1969, by Jerry Sanders and seven colleagues from Fairchild Semiconductor — the same fountainhead that produced Intel a year earlier. Where Intel’s founders Noyce and Moore were physicists, Sanders was a flamboyant salesman, and AMD’s early identity reflected it: rather than invent, it would out-manufacture, making reliable second-source versions of other companies’ chips to tighter quality standards.

Its first product, the Am9300 shift register, shipped in 1970. Through the 1970s AMD made clones of popular logic and memory parts and, crucially, a clone of Intel’s 8080 microprocessor (the Am9080) — early practice for the role that would define it.

Second Source: Surviving in Intel’s Shadow

AMD’s existence was secured by a quirk of the IBM PC. When IBM chose Intel’s 8088 for the PC in 1981, it demanded a second source — a guarantee that the chip would still ship if Intel stumbled. In 1982 Intel signed a licensing deal letting AMD manufacture x86 processors. For most of a decade AMD was, by contract, the safety net of the entire PC industry.

Then Intel tried to close the door. As the 386 generation arrived, Intel refused to license the new designs, and the two companies fought a war of lawsuits through the late 1980s and 1990s over whether AMD’s license extended to microcode and newer chips. AMD largely won the right to keep making x86 processors, but the message was permanent: its supplier was also its mortal competitor.

The 64-Bit Coup

AMD’s first true breakout came in 1999 with the Athlon, the first time its chips matched or beat Intel’s at the high end — and the first x86 processor to reach 1 GHz, in March 2000.

The decisive blow came in 2003. Intel’s bet for 64-bit computing was Itanium, a clean-sheet architecture incompatible with existing x86 software. AMD made the opposite bet: extend x86 itself to 64 bits while running all existing 32-bit code unchanged. The Opteron (server, April 2003) and Athlon 64 (desktop, September 2003) shipped this AMD64 instruction set. It won so completely that Intel was forced to abandon Itanium’s claim on the mainstream and license AMD’s design — the 64-bit x86 in essentially every PC and server today is AMD’s architecture, not Intel’s. For a few years in the mid-2000s, AMD led the x86 market on merit.

The Near-Death Years

Success was squandered fast. In 2006 AMD bought the graphics company ATI Technologies for $5.4 billion — strategically prescient (it gave AMD the GPU expertise behind every modern game console and its later AI ambitions) but financially crushing, loading the company with debt just as Intel’s “Core” comeback and the disappointing Bulldozer CPU architecture (2011) sent AMD’s chips to the back of the pack.

Unable to afford its own leading-edge fabs, AMD spun them off in 2009 into GlobalFoundries, becoming a fabless designer reliant on foundries — eventually TSMC — to build its chips. By 2014 the stock traded near $2 and bankruptcy was openly discussed.

Lisa Su and the Zen Turnaround

In October 2014 Lisa Su — an MIT-trained engineer who had worked on chip technology at IBM — became CEO. She bet the company on a single new CPU core, Zen, and a manufacturing idea that competitors thought too risky: chiplets. Instead of one large, expensive monolithic die, AMD built processors from several small dies stitched together over a fast interconnect (Infinity Fabric). Small dies meant higher yields, lower cost, and the freedom to scale core counts far beyond what Intel’s monolithic designs could match.

Ryzen (desktop) and EPYC (server), both launched in 2017 on Zen, reversed a decade of decline. By exploiting TSMC’s process lead while Intel’s own fabs stalled, AMD took performance and efficiency leadership across PCs and the data center. AMD chips power the PlayStation and Xbox, and its Instinct GPUs (the MI300 line) made it the only credible challenger to Nvidia in AI accelerators. Under Su, AMD’s market value rose more than fiftyfold — one of the great corporate turnarounds in tech history.

⚠️ Dead End: Bulldozer and the Cost of the Wrong Bet

AMD’s near-collapse traces to one architecture. The Bulldozer family (2011) gambled on a design that shared hardware between pairs of integer cores to advertise high core counts cheaply — but each core was weak, and the chips lost badly on the single-threaded performance most software actually needed. For five years AMD had no competitive high-end CPU, hemorrhaged market share, and survived mainly on console chips and cost-cutting. The lesson is the mirror image of the 64-bit coup: in 2003 AMD read where computing was going better than Intel and won; in 2011 it bet on a future of many weak cores that did not arrive, and nearly died for it. The same company shows both how an underdog wins — and how fast a single architectural misjudgment can erase a lead.