#422 1970 · Amdahl Corporation (Gene Amdahl) · Computer hardware / mainframe manufacturing
Gene Amdahl beat IBM by building the one thing IBM could never stop making its own customers want: compatibility
the problem
IBM's own compatibility promise to its customers was something any rival could also honor
background
Gene Amdahl had been the chief architect of IBM's System/360 in the 1960s, the mainframe family whose defining promise was that a customer's software would keep running as they upgraded to bigger, faster machines within the same family — a commitment IBM had built its entire enterprise customer base around trusting. By 1970, that installed base was IBM's greatest asset: no competitor could dislodge it by building a better computer, because customers had years of software and institutional investment tied specifically to IBM's own architecture, and switching meant starting over.
When Amdahl left IBM in 1970 to start his own company, he wasn't trying to build a better, different mainframe — a path RCA and General Electric had already tried and abandoned, unable to convince customers to rewrite their software for an incompatible machine. Amdahl's own three-week internal analysis of how to compete head-on with IBM concluded there was another way in.
what everyone would do
Try to build a completely different, better computer architecture and convince IBM's mainframe customers to switch to it — the approach RCA and General Electric had both already tried and failed at, since IBM's customers had years of software and institutional knowledge built specifically for IBM's own machines, and no better architecture was worth abandoning all of that.
what they saw
IBM's greatest strength — the promise that a customer's software would keep running on every future generation of IBM hardware — was also, structurally, a promise anyone could match. The compatibility didn't belong to IBM's hardware itself; it belonged to a published interface IBM had to keep honoring for its own installed base's sake. If a rival built hardware that honored that same interface, IBM's customers could switch suppliers without touching a line of their existing software, and IBM could not retaliate by changing the interface without breaking every one of its own paying customers first.
the move
Amdahl Corporation built the Amdahl 470, first shipped in 1975, engineered to be fully plug-compatible with IBM's System/370 — able to run the exact same software and connect to the same third-party peripherals — while offering better performance at a lower price than IBM's own comparable machines. Fujitsu funded the venture and built the first units, and Amdahl further benefited from an antitrust settlement that required IBM to make its mainframe software available to competitors, a legal opening RCA and GE had never had.
why it works
By making the Amdahl 470 fully plug-compatible with IBM's System/370, Gene Amdahl let customers run their existing System/370 software and use third-party peripherals without any of the switching costs that normally protect an incumbent — the entire cost of moving to Amdahl was the price of the hardware itself. IBM couldn't defend its market simply by breaking compatibility, since doing so would force its own customers into the exact costly migration that had made switching to Amdahl unattractive in the first place, and Amdahl was further helped by an antitrust settlement between IBM and the Justice Department that required IBM to make its mainframe software available to competitors, removing a legal barrier RCA and GE hadn't had cleared for them.
the payoff
Because switching to an Amdahl machine cost customers nothing beyond the hardware price itself — no software rewrite, no retraining, no lost peripherals — Amdahl succeeded where earlier IBM challengers had failed. By 1979, Amdahl Corporation had grown to more than $200 million in annual revenue and captured 22 percent of the global mainframe market.
where it breaks
It only works while the incumbent's compatibility commitment stays genuinely locked in — once IBM could credibly move customers to a new, incompatible architecture, as it eventually did with 64-bit systems, the plug-compatible strategy lost its foundation, and Amdahl's business, by then owned by Fujitsu, exited the market entirely in 2000. And it requires a rival with the technical capability to build hardware that faithfully replicates not just an incumbent's features but its exact behavior, since customers running production software have no tolerance for subtle incompatibilities a lesser clone might introduce.
what came after
The plug-compatible mainframe business Amdahl pioneered created an entire competitive industry segment that persisted for decades, eventually owned outright by Fujitsu, until IBM's shift to 64-bit systems around 2000 finally broke the compatibility link the whole strategy depended on. Amdahl himself is separately remembered for Amdahl's Law, a foundational formula in parallel computing for the theoretical speedup achievable from adding more processors.
references
- [1]Gene Amdahl, IBM mainframe architect then a rival, has diedComputerworld, 2015computerworld.com
- [2]Gene Amdahl, at 92; influential IBM designer later founded rival companyThe Boston Globe, 2015bostonglobe.com