The Silicon and Industrial Revolutions

Posted by Dietrich Vollrath on August 13, 2020 · 20 mins read

You can see this post as a way of thinking about anything other than …. whatever real life has going on right now. And a warning that I posted this without a real close edit, because otherwise it would be another month before it saw the light of day.

I read Making Silicon Valley by Christopher Lecuyer this summer. It’s a fairly straightforward business history of the origins of Silicon Valley, stretching back in time prior to my own vague notion of “something something Fairchild Semiconductor”. I would not recommend it as a casual read, it is very much a “and then this happened” slog through company origins, contracts, and inventions. That said, it is addressing some real historical questions about the role of military contracts and the presence of Stanford in fostering the Silicon Valley of the late 1970s/eary 80s.

I found the book far more interesting for the somewhat casual asides about what motivated various company founders and inventors, and what led to the concentration of so much activity in that area. As source material for thinking about agglomeration and spillovers, the book has a lot to say. Beyond that, I found myself thinking that Lecuyer was implicitly making a very Joel Mokyr-ian argument that it was ideology and culture that drove this particular silicon revolution.

It also reinforces this idea (which I tend to associate with Mokyr as well) that these kind of industrial revolutions are dependent on small networks of capable tinkerers, and not on either the average level of human capital or on a few heroic innovators. While all the names you’d expect are there (e.g. Shockley, Moore), my impression of their importance was significantly deflated by Lecuyer’s book.

A overly simplified history

To give you an idea of the scope of what the book covers, I’ll give you my own 10-cent summary. It starts with the formation of local vacuum tube companies in the San Francisco area in the 1930s, in particular Eitel-McCullough. Those two were amateur radio hobbyists, and starting making tubes to service that community. This company became a central node of the network of capable mechanics and engineers that would evolve over time and spawn most of the important firms and innovations.

World War II created massive demand for vacuum tubes of all types. By the way, I am committing significant technological heresy here by using “vacuum tube” as a catch-all for the products Eitel-McCullough were producing. But this review isn’t about the technology itself, but the process that led to it. And to add another “by the way”, if you read the book you’re going to want Wikipedia open at all times to look up what the hell a klystron is.

Firms connected to Eitel-McCullough as suppliers, like Litton Engineering, emerged in this period as well. They in turn sponsored new firms like Varian Associates, and what you really had by the early to mid 1950s was “Vacuum Valley”. What will be important for the story below is that these firms required a host of specialized equipment that was provided by smaller firms in the area, and thus the Vacuum Valley had at this point a tightly integrated network of engineers available to be leveraged in other ways.

Vastly over-simplifying, vacuum tubes are just electrical switches, and so it became possible to think of other ways to do electrical switching, like using semiconductors. This is where Shockley Semiconductor appears in the story, although others like Hewlett-Packard were also working on semiconductors. Regardless, you now get a second wave of expansion as semiconductors become the electrical switch of choice, and the network of firms that grow out of Shockley includes Fairchild Semiconductor, which becomes its own massive node in that network. The most recognizable descendant of Fairchild is Intel. All of these firms took advantage of the engineering network of the Vacuum Valley, not just to provide equipment but to make crucial innovations.

The larger point to take away from this tiny history is that the network preceded the breakthroughs, not the other way around.

Iron, steel, and silicon

There is no question that some individuals played outsized roles in the development of Silicon Valley. Eitel, McCullough, Litton, Varian, and others in the first phase built firms from basically nothing. William Shockley won a Nobel prize for a reason. Fortunately for your iPhone, he was also a major asshole, but we’ll get to that. The guys who left Shockley to found Fairchild included people that would have an argument to be on a Mt. Rushmore of Silicon Valley: Robert Noyce, Gordon Moore, Jean Hoerni. If you were looking for “the” innovation that created Silicon Valley, then Hoerni’s creation of the planar process for creating transistors in silicon is probably it. If you want an analogy with the original Industrial Revolution, then Shockley is roughly Newcomen, and Hoerni is roughly Watt.

But littered throughout Lecuyer’s book around these individuals are references to all the innovations and improvements made by the network of engineers working in the Valley in these decades.

Let me just focus on Fairchild and their contribution to give you a flavor for what I mean:

Newly available historical materials … the shaping of silicon technology … was a group effort rather than the creation of ‘heroic’ individuals such as Noyce. (p. 130)

…their first task was to build a strong technical and management team. … The founders also recruited local electronics technicians. Those who had worked in the Peninsula’s tube industries brought with them knowledge of chemical handling, glass working, and vacuum techniques. (p. 139)

… designed the equipment needed for device development and eventual production, and outsourced its construction to local machine shops. (p. 143)

Fairchild’s engineers also developed a radically new planar component, the integrated circuit. (p. 155)

It was Last and his engineers … who made the revolutionary step of engineering and fabricating planar integrated circuits. (p. 157)

Like vacuum tube corporations, Fairchild Semiconductor was dependent on a highly skilled workforce to control complex manufacturing processes and design advanced products. (p. 163)

In addition, it [Fairchild] trained hundred of engineers and technicians in these new techniques. (p. 167)

I could go on and on with quotes like these from Lecuyer’s book. Every advance is shepherded into existence by a small group of skilled engineers, perhaps with a single individual named as the motive force or originator of an idea. But the success of all these inventions - transisitors, integrated circuits - is the continual series of improvements to production techniques and performance that were made by the unnamed engineers of the Valley.

Which is why my mind went to Mokyr’s work. In The Englightened Economy this is one of his themes. Chapter 3 of that book is about “Useful Knowledge and Technology”, and discusses the relative importance of captial-S science versus incremental improvement.

A few items from Mokyr to compare to the quotes regarding Silicon Valley:

But the spirit of mechanical improvement through better access to knowledge is symbolized by many other figures whose mechanical aptitude and ability to tease out every drop of economic value out of what they knew never ceases to astonish us. (p. 55)

Tacit artisinal savoir-faire, experience-driven insights, trial and error, and serendipity drove many of the eighteenth-century inventions, especially in mechanical engineering and iron and coal, far more than any solid scientific base. (p. 60)

It’s not that during either revolution (Industrial or Silicon) the people at work were blind to scientific principles, but knowing them doesn’t necessarily get the efficiency of your steam engine any or yield from your silicon wafers any higher. For that you need to do the thing, and by “you” I mean a small cadre of skilled “mechanics” for lack of a better term. Mokyr uses the example of John Smeaton, known for no specific invention in particular, but who developed much more efficient water mills, among his many very-important-but-not-notable accomplishments.

But do not take the Silicon or Industrial revolution to mean that it is the mass of workers who are responsible for these improvements.

… the Industrial Enlightenment was not the realm of a few heroic inventors and engineers, but neither was it a mass phenomenon that included the working class. It was a minority affair, confined to a fairly thin sliver of a technological elite of well-trained and often literate men. (p. 57)

This is something that Lecuyer doesn’t speak to head on, but it’s lurking in the background of his entire book. Silicon Valley was built by a small group of engineers and technicians that revolved around this geographic area and its corporations. The origin of that thin sliver of the well-trained was the vacuum tube industry.

As Mokyr notes, it is probably impossible to separate the causation between science and practical know-ho, or to separate the causation between flashes of individual insight and the boring accumulation of process improvements. Lecuyer’s book doesn’t provide proof one way or the other. I was just more struck by symmetric the two situations were in this respect.

I think the tendency is to lean on the stories of individual insights and singular inventions. They are better stories, after all. There is a reason that James Burke’s The Day the Universe Changed is such compelling TV. And I’ve always read Mokyr, despite his acknowledgement that there is a role for such individual moments, as pulling hard in the other direction to try and correct our overall outlook on what drove innovation, and ultimately economic growth, in the long run.

NoCal, baby

So Joel Mokyr is really smart, and the history of Silicon Valley is consistent with his views on how industrial revolutions work. But the other element of Lecuyer’s book that stands out is the pure contingency involved in Silicon Valley evolving where it did. Remember that in the 1930s-50s when the vacuum tube industry and then the nascent semiconductor industry were developing, there was no question that the beating technological heart of the U.S. was located in the Northeast. In 1948 you would have placed a more money on Rochester, N.Y. than San Francisco being the home of an innovative electronics industry.

What put this network of capable engineers in northern California rather than upstate New York. Have you been to either place in February? Lecuyer cites numerous examples of how it was simply a case of people preferring to live around Santa Clara rather than out east that helped Silicon Valley develop. The geographic advantages of Northern California didn’t cause smart engineers to emerge there, but they did ensure that the smart engineers that did emerge didn’t leave. If you were a sharp engineer in Indianapolis in 1950 and RCA came calling, you probably packed up and moved to New York. But when RCA (and Westinghouse and GE and ….) tried to poach these engineers from Silicon Valley - and they did repeatedly - these guys stayed put because Northern California is delightful. And at that point, not a housing nightmare.

This goes all the way back to the vacuum tube origins of Silicon Valley. Many of the engineers at Eitel-McCullough, Varian, and Litton ended up working in the East during World War II, either as employees of eastern firms, or as representatives for their California-based firms. But:

They longed for more congenial and familiar surroundings. The sentiment was not specific the Varians and their friends. It was shared by many western engineers who had moved to the East Coast in the late 1930s and during World War II. Most of these men desired to go back to the West and were afflicted by what was often called at the time “Californiaitis” - the intense desire to relocate to California.

Motivated in part by a similar attachment to California, and also to help care for his ailing mother, Bill Shockley decided to open up his semiconductor firm in Mountain View. This might have been just another example of NoCal keeping the local boys home, but a second example of contingency in the development of Silicon Valley came from his choice. Turns out Shockley was kind of an asshole to work for. He apparently tried to use lie detectors more than once on his employees, among other stories.

This led to a group of his top engineers - including Hoerni, Moore, and Noyce - to leave the company and set up on their own. They were funded by Fairchild Camera, and hence Fairchild Semiconductor was born. One of their great disagreements with Shockley was over the technical direction to take semiconductors, and it is arguable that had they not left, the arrival of integrated circuits would have been delayed or lost. You’ll have to leave that contention to someone who knows more about the actual engineering than me. But the combination of Shockley’s poor management style combined with the insistence of the Fairchild group that they stay in Northern California meant that the pool of engineers and knowledge stayed local to the area.

I think it is hard to understate how important it was that as new firms hived off and split from existing ones - including from Fairchild, which spawned numerous children - they stayed in Northern California. And while some of that was for obvious economic reasons like the availability of engineers, a lot of it was also based purely on geographic preference.

The origin of the agglomeration economies that we think lie behind the relative productivity of particular cities may well be random. Beyond that origin, Lecuyer’s book provides some ideas for how agglomeration economies work, and it is very much about the informal social networks formed that pass around tacit knowledge.

Beneath this atmosphere of intense and often ruthless competition was an undercurrent of information sharing. Engineers involved in MOS [metal oxide semiconductors] startups exchanged process and design data with engineers in other firms in an informal way. Most MOS engineers face the same difficult task of developing and stabilizing complex manufacturing techniques. Any information that would help them solve the numerous process issues that they were encountering was welcome. … A web of previous associations also facilitated information sharing. Most of the MOS engineers had worked for other corporations before …. At these firms they built friendships … These previous associations made it easy to contact engineers at other firms and ask their advice. (p. 275)

Bars also fostered the exchange of information among engineering groups. In the first half of the 1960s engineers and managers at Fairchild and other silicon corporations on the Peninsula had developed the habit of meeting after work at a local bar. (p. 275)

… the MOS community on the Peninsula developed a repertoire of process “tricks” that were known only in the area. … In contrast, MOS firms located outside Northern California were not plugged into these networks and did not benefit from their shared knowledge. (p. 275)

To draw this back to the Mokyr work and the Industrial Revolution, we’re back to the idea that the “Tacit artisinal savoir-faire, experience-driven insights, trial and error, and serendipity” were as important as any scientific insights in driving innovation. But again, this was a thin sliver of highly skilled engineers circulating in a small geographic area from firm to firm, not a body of knowledge spread out across the entire workforce.

None of this changes the underlying idea in growth economics that it is innovations and invention that drive growth in the long run. But the similarity of the Silicon Valley experience to how Mokyr describes the Industrial Revolution experience does suggest that this innovation and invention is less a function of economy-wide aggregate features (e.g. demographics, trade) and more on niche groups of innovators embedded in specific places and cultures. Put it this way, my Bayesian posterior on the origins or economic growth took a step in Mokyr’s direction after reading Lecuyer.

You could also file this as an example of the idea that it is better to concentrate your investment and R&D (and education?) rather than making it broad-based. For a couple of other posts I did thinking along those lines, see here and here.

I’m also wondering if this should make me revise opinion on how robust growth is in the long run? If growth depends on some contingent factors leading to niches of innovation, then wouldn’t we expect that to fail every now and then? Or is this where the non-rivalry of the the production processes - once they are developed - kick in a la Romer and save us?