David Mowery, “Federal policy and the development of semiconductors, computer hardware, and computer software: A policy model for climate-change R&D?”

Nanovision, Part III: Gray Goo
Mowery identifies five components of the innovation system (considering the sectors of semiconductors, computer hardware, and computer software). The first four rest with the federal government:

1. R&D funding
2. procurement
3. IPR weakening
4. antitrust

On top of these rest:

5. large commercial markets down the road

Mowery’s overall story is that large-scale and intrusive government intervention (not his vocabulary) created the conditions for these blockbuster industries. Federal agencies – and DOD especially – provided large-scale, stable funding. They required and funded the dissemination of information (e.g. the Sept 1951 Bell Labs conference that led to the printing of 3000 copies of what became the semiconductor “Bible”). Federal agencies then procured large amounts of crucial new products (e.g. Texas Instrument’s Si junction transitior 1954, and the integrated circuit 1958). The govt initiated law suits, consent decrees, and findings that led to licensing and cross-licensing at pivotal moments (e.g. the Army ruling that the ideas behind the EDVAC stored-program computer were not patentable; the 1956 AT&T consent degree that led to licensing of Bell Labs semincondutor patents).

In contrast to Western European defense ministries, which directed the bulk of their R&D funding and procurement funding to established defense suppliers (Flamm, 1983, p. 134), the U.S. military was willing to award substantial procurement contracts to firms, such as Texas Instruments, that had recently entered the semiconductor industry but had little or no history of supplying the military. The U.S. military’s willingness to purchase from untried suppliers was accompanied by conditions that mandated substantial technology transfer among U.S. semiconductor firms. To reduce the risk that a system designed around a particular integrated circuit would be delayed by production problems or by the exit of a supplier, the military required its suppliers to develop a “second source” for the product—that is, a domestic producer that could manufacture an electronically and functionally identical product. To comply with second-source requirements, firms exchanged designs and shared sufficient process knowledge to ensure that the component produced by a second source was identical to the original product. p 6

The military for various reasons wrote a kind of machine language for a national research community. It wasn’t open source, and Mowery stresses how competitive and selectionist it was. But competition wasn’t the point or the main outcome – exchange, sharing, and multiple trials in parallel were the outcomes, leading to the rapid determination of the best manufacturing methods under general pressure to improve continuously. Firms were able to “move rapidly down learning curves”- crucial practical know-how was created in an accelerated way.

It is tempting to see the federal role here as something like a generalizable recipe for developing major industries with general-purpose potential:

  • The government funds research well enough so that money is not a major limit (and pursuing it not the main research activity when 42% of “research time” is already spent on “administration” – Alan Leshner).
  • A big, stable demand is created by government procurement long before a commercial market is possible. Stability creates trust which allows significant private investment.
  • Manadatory pre-commerical information exchange and mandated weak IP lead to various forms of rapid learning and platform development.
  • An environment favorable to forming SMEs without strong IPR. This may mean: cross-licensing plus superior technical performance plus something like protection-through-publication – the community knows you made this – plus an ‘honest broker’ function in the agencies.
  • Inclusion of SMEs in major contracting creates a classically diverse development and production ecosystem.

Some important limits:

  1. Pooling of manufacturing and other kinds of know-how among direct competitors never happened (7).
  2. Some sectors may have been spoiled rotten by unreal market shares: “80 percent in most classes of packaged software.” This may never be repeated.
  3. User-based “co-invention” and other open-source benefits were supported by the adoption of the Internet, meaning gov computer policy may have lucked out.

Mowery notes that energy has never been like this, where programs “have tended to combine instability in R&D funding with little systematic effort to support demand for early versions of new technologies” (17).

But could energy be made more like this? Mowery’s claims for the relevance of the computer industry to energy is very restrained. But could the computer industry paradigm improve renewable energy innovation if it were adapted to energy as deliberate policy?


<a href=”http://innovate.ucsb.edu/wp-content/uploads/2010/02/Computer-Innov-Mowery-NBER.pdf”>Link to Article</a>

Leave a Reply

Comments are closed.