Next / Previous / Shipman's Home Sweet Homepage / Site map

How Hewlett-Packard got into the computer business

Hewlett-Packard was dragged into the computer business by an act of rebellion by two software engineers and the connivance of their immediate management. They didn't originally plan to get into general-purpose computing.

I joined -hp- (that's how the Hewlett-Packard logo was supposed to be rendered in ASCII) in May 1971. I was hired by Mike Green, and Jerry Smith was one of the engineers who interviewed me. Although I heard this story from them nearly forty years ago, I sincerely hope that it is mostly true history and not too much folklore. Any errors, however, are probably due to my imperfect memory.

The genesis of -hp-'s computer business started, the way I heard it, at Union Carbide in the early 1960s, where they were beginning to look at automated process control. At that point Digital Equipment had some offerings, but Carbide figured it would be cheaper to design and build their own computer.

In about 1964, they had a working prototype. The architecture was based on the PDP-8, but with a larger word size (16 bits vs. 12 for the PDP-8). However, when Carbide's upper management got wind of it, they were not happy. During this period IBM made a practice of crushing any company that tried to compete with them, even substantial companies like RCA and GE and Xerox. So Carbide looked around for someone who would buy the design. Todd Poyner's memoir corroborates what I heard about the origins of this architecture in Union Carbide.

Around this period, -hp- was looking for Something To Talk To Voltmeters. That's the phrase I always heard: because -hp- was at that time largely an instrumentation vendor, they saw computers as an adjunct to instruments—in short, a computer was something that could talk to a voltmeter, plot a series of measurements, maybe even issue orders to a controller.

So -hp- bought Carbide's architecture. As I heard it, the transaction involved a set of schematics, the primary electrical engineer who had designed it, cash, and the transfer of a secretary in the opposite direction: from -hp- to Carbide because she wanted to move to New Jersey.

This machine became the -hp- 2116, the company's first actual computer product. The 2116 was an extremely solid design. In fact, because of an -hp- corporate policy in force at the time, you could actually shove one out the back of an airplane attached to a parachute and it would probably survive the landing.

The initial operating system was pretty bare-bones, but adequate for instrumentation work. However, it was clear that -hp- needed better software, so they hired fresh Stanford graduates Mike Green and Gerould "Jerry" Smith, who had master's degrees in EE with a substantial software background from what was then and still is a power in the academic software world.

Mike and Jerry were assigned to design a real-time operating system for process control applications. However, they were not happy with this quite boring assignment. Moreover, they thought that the 2116 would make a good timesharing system, at a price point that would open a huge market: small colleges, even secondary schools, small engineering firms and other customers that could never afford the room-sized machines common for the larger customers.

So, with the permission of only their immediate supervisor, in something like nine months they had written a complete operating system that would support sixteen Teletype ASR-33 printing terminals (110 baud, that is, a whopping ten characters per second). This was a highly specialized operating system: instead of a command prompt, each terminal spoke only a dialect of BASIC.

The demo was a 2116 ringing the bell on sixteen ASR-33s at once, probably something like this.

  10 PRINT "^G"
  20 GOTO 10
  30 END

This product was called the HP2000A. In a short time a two-processor version, the HP2000C, was developed; one processor handled only communications, the other execution of programs. This model sold until at least 1978, and was commonly seen in the original target market: educational institutions and smaller engineering firms.

This success, and their tendency to recruit great software engineers fresh out of Stanford, MIT, and Berkeley, allowed the development of the quite ambitious Omega machine, which mutated into the HP 3000, a timesharing system that was sold for decades.

At one point a software engineer named Gene Studlien (STEWED-lee-un), who was working on the early HP 3000 team, wanted to transfer to an -hp- division in Boston. I was hired to replace him, not from a Name School but from an obscure little nerd paradise named New Mexico Tech, but that's another story.

I heard later that there was another applicant who was more qualified, and came from one of the good schools. They chose me over him because he had just married and they were expecting a child. They figured they would get more work out of a bachelor. They called it right: I probably averaged at least fifty hours a week back then, often more.

But the Omega, the Alpha, and the HP 3000: that's a story for another day.


Next: Hewlett-Packard tightens the screws (1973)
See also: Industrial period I: 1971-1983
Site map
John W. Shipman, john@nmt.edu
Last updated: 2011/08/20 20:08:52
URL: http://www.nmt.edu/~shipman/write/memoirs/hp2000.html