It’s funny how the same reading of  history leads to different conclusions. The young investor in the 1840s Punch cartoon above stands in a back alley outside the Capel Court stock exchange asking a purveyor of dubious scrip how to honestly make £10,000 in railways. It is the end of a technology hype cycle in which the modern-day equivalent of $2 trillion was pumped into an investment bubble.  The picture on the right is a desolate and economically insignificant outpost connected by some of the 2,148 miles of railway capacity that entrepreneurs built during the British railway investment mania of the 1830s. The conclusion is that early investors in British railway companies were played for suckers.

The mania probably started with an announcement in the May 1, 1829 edition of the Liverpool Mercury:

“To engineers and iron founders

The directors of the Liverpool and Manchester Railway hereby offer a premium of  £500 (over and above the cost price) for a locomotive engine which shall be a decided improvement on any hitherto constructed, subject to certain Stipulations and Conditions, a copy of which may be had at the Railway Office, or will be forwarded. As may be directed, on application for the same, if by letter or post paid.

HENRY BOOTH Treasurer Railway Office, 25 April 1829

The Liverpool and Manchester Railway was not the first railroad in England, but the competition drew enormous interest.  Contestants used everything from “legacy technology” — horses on treadmills — to lightweight steam engines that could reach up-hill speeds of 24 miles per hour. The legacy technology defeated itself when a horse crashed through a wooden floorboard. It did not hurt that Queen Victoria declared herself “charmed” by the winning steam technology.

Business innovation  — ticketing, first-class seating, and agreements allowing passengers to change carriers mid-trip — was rapid and fueled as much by intense competition as by a chaotic, frenzied stock market in which valuations soared beyond any seeming sense of proportion, causing  John Francis in 1845 to despair: “The more worthless the article the greater the struggle to attain it.” When the market crashed during the week of October 17, 1847 — in no small measure due to to the 1845-6 crop failure and potato famine — and established companies failed, financiers like George Hudson were exposed as swindlers. Thomas Carlyle demanded public hanging.

The collapsing bubble is not the end of the story. Between 1845 and 1855 an additional 9,000 miles of track were constructed.  By 1915 England’s rail capacity was 21,000 miles.  British railways had entered a golden age. The lesson that observers like Carlotta Perez and others draw is that there is a pattern to technological revolutions:

  1. Innovation enables technology clusters, some  of which transform the way that business is done.
  2. Early successes and intense competition give rise to new companies and an unregulated free-for-all that leads to a crash.
  3. Collapse is followed by sustained build-out during which the allure of  glamor is replaced by real value.
  4. This leads to a golden age that results in more innovation as lives are structured around the new technology.

This is a Schumpeterian analysis of innovation that is reflected everywhere, but particularly in the economics of the new technologies of the late twentieth century.  The stamp of the the 1840s British railway mania can be seen in Gartner’s technology hype cycle and in nearly every discussion of the 2000 dot-com collapse.  It is an analysis that is a special problem for angel and other early-stage investors because there is no real guide to tell you when the bubble will burst. Unless you are George Hudson, what investor will find the risk acceptable? A rational early investor will steer clear of technologies that radiate this kind of exuberance.

But what really happened to all that investment in the 1830s? I was amazed to see the recent article by my long-time colleague Andrew Odlyzko at the University of Minnesota who analyzes the British railway mania example and concludes that the early investments did quite well:

The standard literature in this area, starting from Juglar, and continuing through Schumpeter to more recent authors, almost uniformly ignores or misrepresents the large investment mania of the 1830s, whose nature does not fit the stereotypical pattern.

Andrew enjoys taking contrary — often cranky but always well-thought out–  positions on conventional wisdom, so I approached his article with cautious interest.  After all, I thought I knew a little about the railway mania episode.  I had used it myself to illustrate innovation cycles. Like most people, I had focused on the disaster of the 1840’s, so I was drawn immediately into Odlyzko’s argument that during the mania of the 1830’s,  “railways built during this period were viewed as triumphant successes in the end.”:

After the speculative excitement died down, there was a period of about half a dozen years during which investors kept pumping money into railway construction. This was done in the face of adverse, occasionally very adverse, monetary conditions, wide public skepticism, and a market that was consistently telling them through the years that they were wrong.

In other words, the end result of the wildly speculative exuberance of the  1830s was the “creation of a productive transportation system that had a deep and positive effect on the economy.” Investors saw great returns. A shareholder in London and South Western Railway (LSWR) who in 1834 paid a £2 deposit on a share worth £50 and who paid all subsequent calls (totaling £95.5) would have watched the investment grow to 2.31 shares valued at  £200 by mid-1844 and would have received in 1843 alone £4.62 in dividends — a 9.68% annual return.  This defied the more rational demand and cost forecasts:

at the start of the period…in June 1835, such investor would have paid £10, and seen the market value it at £5.5. In fact, over most of the next two and a half years, the market was telling this investor that the LSWR venture was a mistake, as prices were mostly below the paid-up values.

Andrew Odlyzko is a seasoned mathematician who knows better than try to prove a general principle by example.  He says as much in his paper. On the other hand, railway mania has been used for years as an illustration of an innovation cycle, and  Odlyzko has a very different reading of history. The conclusion that is usually drawn from the Railway Mania may lead markets and investors astray because it seriously misrepresents actual patterns. The whole point of a cycle — hype, innovation, or investment mania — is that it can be used as a risk-averse template for rejecting sales pitches that start with “This time is different“.  But that does not mean that it is never different.

There was a birthday celebration of sorts last week.  From the October 29th edition of  ABC News Science & Technology:

While the actual date of the Internet’s birthday is somewhat debated, many say that the Internet was born 40 years ago today at the University of California, Los Angeles, when a computer to computer message was sent for the first time from the UCLA campus to Stanford.

At the time, Leonard Kleinrock and his colleagues were charged with developing the Advanced Research Projects Agency Network (or ARPANET), a government-funded research project in global computer communications that eventually grew into the Internet.

I thought it would be a good occasion to  reflect on how easy it is for Loose Cannons to get smashed by colliding worlds.

In the days before ARPANET, computer-to-computer communications were homogeneous, and computer manufacturers liked it that way. The very idea of not owning every aspect of a technology stack seemed to be ridiculous.  Where’s the value if you can get critical components from anywhere?  What if competitors start using the same suppliers?  Heads of business units hated the idea, but Loose Cannons kept proposing technical architectures that looked, well, open.  The idea was playing out in many ways in many companies.

At IBM, two architectural revolutions were simultaneously  underway. We now know that they were related. In the summer of  1980, IBM executive Bill Lowe prepared to brief  the company’s Management Committee on development plans  for a personal computer:

It was a dangerous place to be.  The Management Committee — or, given IBMers’ fondness for acronyms, the MC — ruled on issues that couldn’t be resolved at lower corporate levels, so going before the committee was, to IBMers, like going before the Supreme Court.  It was actually rougher because the top IBM executives who sat in judgment were known to be brutal, especially if they thought someone was wasting their time.[1]

Bill Lowe had been beaten up by the MC before, but this time Lowes’ plan to use outside suppliers drew polite questions from MC members who expressed some concern about turning over even partial control of any of their businesses to “outsiders.” What Lowe and the vast majority of IBM engineers didn’t know was that earlier in the year the MC had received a  forecast for global PC sales that showed a peak market of 80,000 units in that began to rapidly decline in 1984 as the specialized customer  need for computers was satiated:

IBM had already been embarrassed by early missteps in the PC market but the corporate culture was focused on mainframes and services.  Problems might be created by opening up the hardware and software architecture of personal computers, but

The general attitude…was that you don’t have big problems in small markets, and we thought the personal computer was a very small market.[1]

The MC might have been more inclined to turn its attention to a market that had real legs.  Like, say, networking.  Ed Hendricks was an engineer at IBM’s Federal Systems Division in San Diego.  Hendricks had helped design VNET, at that time the largest computer network in the world.  VNET was  IBM’s internal corporate network, linking IBM mainframes at scientific data centers.  By 1980, VNET was a global asset with hundreds of  hosts in North America, Europe and Asia.

Meanwhile, ARPANET was growing into the Internet, and Ed Hendricks was interested in how IBM’s technology would continue to prosper when the world started connecting IBM mainframes to large UNIVAC computers, HP mini-computers,  PC’s, and supercomputers from Cray or Control Data.  Hendricks became an industry player in this arena, collaborating with my colleague Larry Landweber at the University of Wisconsin as the expansion of the ARPANET began in earnest. Ed  Hendrick’s IBM Internet Gateway Project was aimed squarely at insuring that IBM mainframes would not be stranded in a world in which they could only talk to each other:

The objective of this project is to begin to bridge the gap between IBM computer systems and network technology predominant among government agencies, conractors and universities.  More specifically, we are working to develop according to DOD standards the technical capacity to interconect networks of IBM computers and systems to similar but different computer networks used by government agencies and their affiliates.

Hendrick’s website preserves the sometimes heated but  thoughtful and deep technical discussions — involving Hendricks,  the legendary Jim Gray, and MIT’s Jerry Saltzer, among others —  that took place througout 1980 about the relative merits of ARPANET and IBM’s networking strategy. For reasons that are still unclear, IBM decided to move the Internet Gateway Project to IBM Research in Yorktown Heights, New York, an effort that Hendricks calls “screwy.”   Hendricks along with team members Gerot “Mike” Engel and Dale Johnson planned to spend a week at Yorktown Heights, getting comfortable with IBM Research’s Systems Laboratory, their proposed  new home:

…the Systems Laboratory was created to focus more directly on perceived business needs. Consequently, Systems Laboratory projects are evaluated and prioritized on the basis “leverage” they exert on the software product line…by design, ninety-five percent of the work carried out in the Systems Laboratory is so closely related to strategic product development that it cannot be discussed outside IBM.

Shocked, the Internet Gateway team concluded:

…a project such as ours which is intended to establish internet communication compatible across differing systems…could not be carried out under such guidelines.  Our overall reaction…was that the ARPANet Internet Gateway project could not have been started within the Systems Laboratory.

They concluded that if the project was to have any chance at all of success, there would need to be a formal review of management decisions, what  IBM called the “Open Door” process.

March 14, 1981

John R. Opel, President IBM Corporation

Dear Mr. Opel,

This letter is intended to invoke the IBM Open Door Policy.  My purpose in requesting this Open Door is to seek clarification of the decisions which led to a situation where a project which is clearly critical to IBM’s future posture in the data communications industry cannot be pursued…Bureaucratic accomodation for only that which is in the strategic plan is a very dangerous posture to be in while the data processing and communication industry is rapidly evolving.

[My team and I] have been working to carry out a project to establish a capacity…to cooperate with the U.S. Government and University Computer Science departments in the evolution of techniques to interconnect dissimilar computer networks…There is essentially unanimous agreement that this activity promises important advances for IBM and for computer technology in general.

In September 0f 1980 we were notified by our management that this work could not be carried out…On each occasion when this qustion [of where the work could be carried out in IBM] was being escalated to the proper level, my management would insist that I leave the management issue to them and to concentrate my own efforts of the technical work.

Last week I was informed verbally that no sponsorship for this project could be found.  My manager asked where hie should look to find me a job. My position was…that inability to find organizational sponsorship for the project is not equivalent to a decision that IBM should not be involved in developing the capacity to interconnect IBM networks to government and university networks…to look for other professional opportunities now and give up attempts to pursue this technology…would be to let the company down….

Sincerely yours,

Gernot Engel

19 March 1981

Mr. Thomas J. Watson, Jr., Chairman Emeritus

Dear Mr. Watson,

My employment with IBM has been terminated as a consequence of recent management decision which are incompatible with my professional goals…I believe I am justified in requesting more thorough and explicit responses to the following questions:

  1. What “business needs required the termination of our ARPANET Interconnection Gatweway Project and the abandonment of the…professionals we had been dealing with?
  2. What factors prevented alternative organizational arrangements that would have allowed our group to continue its work within IBM?
  3. What is IBM’s posture regarding professional cooperation with the computer scientists working in association with DARPA…to establish mutual techniques for interconnection of dissimilar computer networks?…

Sincerely yours,

Gernot Engel

May 15, 1981

John R. Opel, President IBM Corp.

Dear Mr. Opel,

On March 4, 1981 I sent a letter to your office requesting clarification of a decision which cancelled the internet gateway project…Your office’s attempt to analyze the internet decision appears to be stalled because it was handed back to middle management….I can only conclude in this instance the Open Door Policy has failed. My recommendation to salvage the situation is that you give fifteen minutes of your time to receive a presentation on the internet project and attempt to evaluation for yourself the value of this project to IBM’s future.”

Sincerely yours,

Gernot Engel

May 19, 1981

Dear Mr. Engel,

I have reviewed the results of [the] investigation into your concerns.  Your disappointment with the decision to terminate the VNET/ARPANET project is understandable; however, I conclude the decision was properly based on the need to fund other Ad Tech projects with greater business potential…

I understand you are currently considering a return to IBM, and I hope you choose to do so.


John R. Opel

Number 1-81: September 11, 1981 MANAGEMENT BRIEFING


Organizations seem to have an irresistable tendency to codify successful practices in rules, instructions and controls which soon begin to take the place of judgement. When that happens, the result is bureaucracy.

IBM is not immune.  Earlier this year, reports from many sources indicated to me that a growing bureaucracy is affecting the performance of our business…corporate staff heads, group executives, and the division presidents are exploring ways to reduce unnecessary controls, rules and approvals in their areas of responsibility…We will succeed in that effort only if you managers, at every level of the business,k are willing to stand up and fight bureaucracy wherever you find it…If you have all the information to make a decision, make it…

[signed by John Opel, president]

John Opel stepped down as IBM president in January 1985 and chairman in May 1986.  He was succeed by John Akers, and he was succeeded by Lou Gerstner in 1993. Gerstner, the former CEO of RJR Nabisco, described his transformation of IBM in “Who Says Elephants Can’t Dance?”[2].  Most observers agree that critical to IBM’s turnaround that took it from a free fall in the early 1980’s to unquestioned market  leadership in computers, software and services was the dismantling of a remote, hierarchical management culture that squeezed innovation in political pincers.  By the time I took over the computing research directorship at the National Science Foundation in the late 1980’s, IBM had become a major player in the growth of the Internet [3]:

In the mid-1980s, NSF decided the time was right to try to link its regional university networks and its supercomputer centers together. This initial effort was called NSFNET.
By 1987, participation in the new NSFNET project grew so rapidly that NSF knew it had to expand the capacity of this new network. In November of that year, it awarded a grant to a consortium of IBM, MCI, and a center at the University of Michigan called Merit to create a network of networks—or inter-net—capable of carrying data at speeds up to 56 kilobits a second. By July, 1987, this new system was up and running. The modern Internet was born.


1. Paul Carroll, Big Blues: The Unmaking of IBM, Crown Trade Paperbacks, 1994

2. Louis V. Gerstner, Who Says Elephants Can’t Dance? Inside IBM’s Historic Turnaround, Collins, 2002

3. National Science Foundation, NSF and the Birth of the Internet,