This week, we’re publishing a whopper of a case: The Rise and Fall of Data General (members only). This took some time to put together, because the first half summarises the development of the computer business from the 40s through to the 50s.
Just wanted to add that the conclusion I make at the end of this piece is very much informed by the various discussions members have had in the forum over the past couple of weeks. I want to thank folks for debating bubbles with me.
Would love to go through the 1980’s part of this history in more detail. Was there anything that Data General could have done that would have made a difference?
The story of this internal challenge and eventual triumph was the subject of Kidder’s The Soul of a New Machine. Unfortunately, Eagle — which launched as the Eclipse MV/8000 in 1981 — was a pyrrhic victory. IBM’s introduction of the PC that same year kicked off the personal computer revolution, which marked the beginning of the end of Data General — along with DEC and most, if not all of the other minicomputer companies.
The Eagle team disbanded shortly after its launch. Over the course of the 80s, Data General’s earnings and market share shrank in the face of an industry-wide shift to smaller cheaper personal computers. DG’s stock followed. This was despite the fact that the Eagle raked in over $1B in sales in 1984 alone.
I love this era (and I love “Soul of a new Machine”). Here’s my gut response.
Microcomputers were an easy substitute for people who wanted a low-cost, low-performance, single user computer, and it’s easy to see the way that the IBM PC stole that market from the minicomputer manufacturers, but I’m not convinced that it was microcomputers that fundamentally upset the minicomputer market. IBM was very sensitive to selling PCs that they saw as competitive with their Mainframe business (their hesitation to release a 386 PC was part of how they lost their grip on setting "industry standard” behavior for PCs), and DEC continued to find success with the PDP-11 and VAX all through the 1980s.
The greater impact came from network-first client/server systems like those built by Sun (or even Bloomberg). These client/server systems were able to benefit from all the competitive innovation in the minicomputer market without having to race to the bottom in pricing against other IBM Compatibles. Could Data General have successfully launched a workstation-esque down-market complement to their minicomputers?
They tried! And it didn’t work out:
Data General entered the UNIX market late because of an internal report that warned that it could not compete with its proprietary solution. Its software offering was the DG/UX operating system. Unfortunately, it based its hardware offering on the Motorola 88000 which, as mentioned previously, was not a success and was eventually withdrawn, leaving Data General to base its next product in 1995 on the x86 series from Intel. This was already a crowded market and the product was not ready until 1996, according to its 1995 SEC report. Thus Data General missed the PC and the workstation S-curves, joined the UNIX RISC computer system market late and backed the wrong RISC processor.
(quoted from a Digital Equipment Corporation case study/Ph.D thesis: https://pure.uva.nl/ws/files/2790157/176138_Goodwin_complete_thesis_formatted_final.pdf)
Likewise. I found a yellowed paperback copy of the book in a common room in 1998 and was transfixed. I still think of it as what Cedric would call a concept instantiation of ‘hero culture’ and the limits of hero culture’s effectiveness inside a company. It dissipated my lingering FOMO over being just slightly too young to participate in the then ongoing dotcom boom
This was one of the first books I read after a long dry spell from my mid teens to mid twenties of not reading for personal development (I.e, not reading things that weren’t assigned coursework); another was “Zen and the art of motorcycle maintenance”. Both helped me to develop a deeper appreciation for the “art” side of engineering. It’s so full of anecdotes and anti patterns and is far from a roadmap for success, but it’s neat to just see a group of motivated people build something interesting.
Does anyone happen to have a way of accessing Data General’s annual reports in the 1970s? Via Mergent or something like that? I’m curious about the annual the gross margin trajectory, i.e., how long the 80s-% margin was sustained. Or, any report on minicomputers in the 70s and 80s that would otherwise shed light on this progression.
I don’t have remote access to Mergent, but have been able to access it as an on site guest at university libraries
If no one else has access, and you can’t get to a university library, Interlibrary Loan is worth a shot - not for the reports themselves, but for photocopies. (Being willing to pay copying fees helps)
This was one of the more annoying things about researching this case. After much Googling, I found US libraries with copies of their annual reports (Purdue; Library of Congress) — both far beyond my reach.
But I didn’t even think I could get their annual reports. At a basic level, I wanted DG’s stock price performance. How did the market evaluate them over the course of the 70s? (Note that DG went public in 1969). There’s some indication of how the market evaluated them in the 80s and 90s (which are in the sources of the case: Forbes article one, two) — but I am curious as to how the market treated their outperformance in those heady times, even given the backdrop of 70s-era stagflation.
Holy crap this is super useful, and exactly what I wanted to know. Thanks! I did not find this in my original research; let me see if I need to update this case with these arguments in mind. How did you find this, @mgoodrum?
Count me in. I think Soul has the best executed, most thrilling description of debugging ever published — one whole chapter that was impossible to put down. Kidder is an amazing writer. What a gift this book was.
One thing that I didn’t include in this case, and that I think I’ll have to add, is that DEC’s VAX really sucked the air out of Data General’s original market. The Eagle didn’t make much of a dent in the VAX; DEC basically creamed the competition and hit its peak years (in terms of both profits as well as market share) in the 80s.
I didn’t include this because the case was already getting very long, and I wasn’t 100% sure why the VAX was so dominant. I’ll circle back when I have a better source.
Regarding the 80s, there’s a throwaway comment in one of the Forbes articles that goes:
Data General survived because it had avoided excessive debt. (Even at its nadir, the company had more cash than debt.)
This was actually quite interesting. It seems there are dozens of minicomputer manufacturers who died in the 80s in the face of DEC dominance, though I can’t find online traces of them. The more famous contemporaries that failed in the early 90s were Wang Laboratories (bankrupt in 1992), Prime Computer (also 1992), Apollo Computer (acquired HP in 1989, gradually closed down over 1990-1997). DEC itself was acquired by Compaq in 1998 (a botched acquisition that doomed Compaq), a year before Data General got bought for parts.
This was via a ChatGPT prompt based on a vague recollection:
Did Data General compete with Sun Microsystems in the 1990s?
If so what happened?
Generate 1000-2000 words in response
Link to sources for further reading
The case study was behind one of the few links that didn’t point to Wikipedia (which I find too vague in this case)
The ‘Generate 1000-2000 words’ bit is just a ploy to motivate the LLM to get into details and point to detailed sources - the stuff it generates is hit and miss
I got curious about Data General’s stock price over time, and I do have access to New York Times page scans via TimesMachine, so here are the New York Stock Exchange year end prices and annual highs and lows for Data General between 1973 and 1982:
| Year | Year High | Year Low | Year End Close |
|---|---|---|---|
| 1973 | 40 | 37.5 | 37.75 |
| 1974 | 39 | 10.5 | 14.25 |
| 1975 | 39.5 | 8.625 | 37.375 |
| 1976 | 60.75 | 37.375 | 45.375 |
| 1977 | 53.875 | 34 | 50.75 |
| 1978 | 71.375 | 42 | 69.5 |
| 1979 | 74.5 | 46 | 54.875 |
| 1980 | 87.75 | 52.75 | 65.875 |
| 1981 | 68.5 | 40.5 | 53.75 |
| 1982 | 55 | 20.25 | 39.75 |
Data General was first listed on the New York Stock Exchange circa December 26th 1973, so the range for 1973 only represents ~6 days of trading
During this period, fractional prices were listed in eighths of a U.S. dollar; given the vagaries of newsprint, some of these fractional values might be off
Given what was happening in the 70s, this is quite impressive as a performance. Thanks, @mgoodrum.
Also, I’ve read through the thesis (with the aid of Gemini Pro to pick out pages for me to read) … and am now thinking through how to update the case to make salient some of the conclusions of the thesis (which I 100% agree with). 
The short version is that all the minicomputer companies started out producing machines for OEMs, but as the 70s progressed, switched to vertical integration. This is where the ‘extremely high margins’ comes from. Vertically integrating and selling into end enterprises meant that they grew used to consulting-like engagements, maintained a huge sales force, and grew fat and happy on 80%-ish margins … only to find themselves unable to adapt when the emergence of PCs forced a change back to the low margin, high volume ‘sell to OEMs’ game.
Goodwin’s thesis is that it is the shift in business model that killed the minicomputer companies, not the development of PC technology (he points out, rightly, that DEC and others had a decent response to the PC product category). But they were hobbled by the change in business model, and then they shot themselves in the foot by switching to commodified open platforms like Unix (though to be fair, in DG’s case, did buy them a number of years in the 90s).
Edited to state that the picture is more complicated than expected; I was not as familiar with DEC’s history and a closer reading of the thesis paints a more complex picture.
This strikes me as crucial for understanding the case of DG. After reading the case my impression was that it was a failure of technical directions that ultimately killed them, but this seems relatively minor compared to the organizational challenges of switching to lower margin PC sales.
Maybe I am over indexing on this, but it completely changes how I categorize this case. After my first read, I was intuitively putting it with General Magic where the key difference is that they failed to find a product that landed after past success. So it was a case describing how in cases of rapid technological change, you must continually find authentic market demand.
With this thesis however, it changes to a case in organizational culture and incentives. Basically it feels like a case like Intel, but instead of pivoting from memory early and hard, they pivoted late and slow.
Curious if this matches with other people’s takeaways?
As a general aside, I think finding some way to incorporate more annual reports into the cases would be a very positive addition. That way when I am trying to put myself in the shoes of the people in the cases, I can stop and see what the actual financial metric look like at various points. I think this might be an even better way to get an intuition for reading financial statements.
We’re working on it 
As for the categorisation of DG — hold up first. There’s still one last piece that’s missing from this story, that I’m investigating: why did VAX suck so much oxygen out of their core market?
Bitsavers is an archive of original documents that covers this period of the computer business extensively
It’s a pure document archive optimized for replication rather than curation or reading guidance, so I asked ChatGPT to do a site search and surface documents on VAX’s market, highlighting stuff related to Data General
I sampled a few of the results, and they look like they might be useful, so here’s a link to the chat:
While looking for something else, I found Data General’s first annual report at the Internet Archive:
https://archive.org/details/Data2469_1969
The stock option plan, as yet unused, is mentioned in the notes, as well as the fact that the officers were contractually prohibited from receiving options
So … it turns out I’m wrong. I wasn’t as familiar with DEC’s history as I was with DG’s and it turns out Goodwin’s thesis ties together three different lines of argument, and is very much a response to Edgar Schein’s DEC is Dead, Long Live DEC.
I’ve read more of Goodwin’s thesis now. It’s … fascinating, and perhaps deserves to be its own case. But I need to read some of the material that Goodwin was responding to, like Schein’s book before I can do so. This history of DEC is more complicated because it was the market leader and had more resources, so therefore it was involved in more things (sort of like Meta or Apple is today).
As a result, however, I retract all comments about PCs disrupting vertically integrated minicomputers due to OEM dynamics. The market that DEC operated in was vastly more complex than I expected. For instance, it even had a search engine bet going in AltaVista.
I apologise for the error. I’m narrowing down a reason for the VAX dominance. I’ll update the case and post an update in this thread when it’s done.
Updated the case with the following section:
The story of this internal challenge and eventual triumph was the subject of Kidder’s The Soul of a New Machine. Unfortunately, Eagle — which launched as the Eclipse MV/8000 in 1981 — was a pyrrhic victory. The Eagle team disbanded shortly after its launch. IBM’s introduction of the PC that same year kicked off the personal computer revolution, which marked the beginning of the end of Data General — along with DEC and all of the other minicomputer companies.
The odd thing about this period was that the PC — whilst a very hot segment — was only a threat to the minicomputer companies in the 90s. In the 80s PC adoption was a slowly growing force. As a result, Data General was not immediately threatened by the scrappier PC manufacturers. The bigger problem was DEC’s VAX system. By the time Eagle launched, VAX had been in the markets for four years, and had gained a strong hold over DG’s core markets: the scientific and technical computing segment, as well as the commercial computing segment.
VAX’s early lead lead to two big — and ultimately defensible — advantages: the first, VAX’s multi-user, multi-tasking operating system VMS became a widely targeted platform for software developers. The second advantage, perhaps more important for enterprise sales, was the VAXcluster networking story that VAX launched with. The pitch to enterprise customers was that you could start with a few VAX machines in a cluster setup, and then add more to your VAXcluster over time, benefiting from shared resource coordination and reliable data sharing. Naturally, this acted as a form of lock-in. Approximately 6000 VAXcluster systems were in operation by mid-1987. Even today, the VAX is considered the most successful minicomputer system ever.
As a result, DG’s market share collapsed over the course of the 80s. DG’s stock followed. This was despite the fact that the Eagle raked in over $1B in sales in 1984 alone.
A footnote: Word of mouth in my corner of the internet circa early 1998 was “Google is AltaVista, but better.” And it was, for reasons that have been documented ad nauseam elsewhere
Stanford-era Google’s front page copied AltaVista’s early DEC-era search interface (screenshot visible on the AltaVista Wikipedia page) which helped make switching easy