Semiwiki on LinkedIn Semiwiki on Facebook Semiwiki on Twitter Register / Log in 
Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/looking-for-stats-going-back-to-the-60s.17772/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Looking for stats going back to the 60s

M

mozartct

Active member
I am in charge of upcoming ASMC panel discussion. We are bringing together experts to talk about "unintended consequences of CHIPS on Moore's Law". I need to put together a few strong graphics but I am having difficulty locating long-series data. Here is specifically what I am looking for:

1- Chart showing the growth in WW IC shipments going back to early 60s. I have data from 1978 to 2018. I would like to go to the present and all the way back to invention of IC by TI if possible. Main point: semiconductor business has increased by orders of magnitude.
2- Chart showing growth of semi Korea and Taiwan, in absolute terms (by wafer, by ICs, etc.). We often see charts showing market share but I would prefer to show in absolute terms. I think that market share does not convey the full story since one would have had to invest massively, just to keep the same market share (see 1-).
3- Chart showing percentage of chips used by military as a function of total market. I suspect it went from 90% in early 60s to under 1% today.
4- Some way of showing "cost of adding a unit of production" over time by country. What I am getting here is that by the cost of building and operate factories has increased in the USA more than it has in other countries. Not sure who compiles that information.

As always, thanks!
 
I suspect the 1960s would be problematic since the product category would not occur to anyone, on top of the general lack of statistics for any product categories. You might be more realistic in looking for single data points for 1965 and 1970 by looking for the major participants and seeing if there were any claims by them about sales. But it may be tricky, for example TI in early days probably had confidential contracts with a few. You would also need to think about captive production, like how much TTL and ECL did IBM produce in-house?

Japan was first of the "tigers" to figure large, by the early 80s they had convinced Intel to bow out of memory. You might find share data on that. Taiwan and Korea were pretty small until the 90s by comparison, though they had some niches. Again, until the category gets large enough no-one gathers the stats, so you might need to go back and find early revenue figures for TSMC and UMC and Samsung, HK, and Hynix in their early years.

Probably not as much as you think. IBM used SSI and MSI in mainframes in the 60s, there were 7 competitors making up about 30% of the commercial market, and there was logic in some of the peripherals. Early calculators, machine tools, and other products had integrated circuits. The military were then, as now, notorious for not using the latest tech, outside of a few low volume specialities. So I would guess you will find the opposite, commercial markets were 90%.

on (4), best of luck. I would guess that USA fabs were quite competitive at least until the 90s because the ecosystem was best, only Japan and Europe would have had local suppliers and service for the most advanced machines until the late 90s, and they would have taken time to build the workforces too. Silicon Valley became a crowded place with expensive land, causing expansion to the rest of the world. Geopolitics made setting up R&D or production in other countries a no-brainer for the larger companies. Good for sales, access to talent. TSMC has been really slow to figure that out.
 
@Tanj Thanks for that. Good insights. Much of the rhetoric in the USA is that invented the transistor, we invented the IC, DRAM, etc. That is true but is tangential to the requirements of successful manufacturing in today's world: large factories, plenty of customers to full all the time (so foundry model) and ability to align thousands of hard headed engineers and PhD in the same direction. This is where Taiwan and Samsung innovated. Access to cheap capital and favorable government policies helped for sure.
 
mozartct said:
plenty of customers to full all the time (so foundry model)
Click to expand...
That is the oligopoly model, not foundry. Lots of other industries have foundries and are competitive at less than full production. Stay aware of why we have this, don't assume it is a given.
mozartct said:
and ability to align thousands of hard headed engineers and PhD in the same direction. This is where Taiwan and Samsung innovated. Access to cheap capital and favorable government policies helped for sure.
Click to expand...
Well, Taiwan and Korea both have worries whether the new generations are available to keep that going. And there are other factors that could improve productivity, promote retraining from related professions not just expecting university to churn out all the recruits, changes in production lines requiring less humans, etc.
 
@Tanj My full sentence should have read "plenty of customers so that the factory(ies) is full all the time". In my mind, with the new mega factories, this implies being a foundry unless Apple is your customer and they buy 100% of your output. Intel's model relies on building large factories but the actual utilization will depend on market conditions for PC and servers. Tsmc does not have the risk. Mega factories in our sector need to be nearly full to be profitable because of the capitalization. Car factories are much cheaper and can be run profitably at 16 hours per day...

I also am not sure if Taiwan engineers can keep up the cadence (without major wage increases). At least South Korea's salaries are higher.
 
My response to your questions will begin with a bit of sarcasm, because when I was young, we had to go to a building called a "library". In this building, we could research many print publications with the assistance of somebody called a "research librarian", and if we were lucky the library had subscriptions to the print publications with the information we needed. For engineering related topics, we usually had to go to a university with a good engineering program and hope the thinly circulated engineering journal we wanted to read existed in their library.

For the most accurate information about integrated circuits prior to 1978, I would try to find a library that still has publications that date back to the early 1960's. This kind of information won't be easy to find on the internet because most of it wasn't digitized. Anectdotally, I recall the two commercial products that used IC's in the late 60's/early 70's were portable radios and calculators.
 
Why answer at all if you can only be sarcastic? I have worked for months putting this panel together, reading nearly 1000 pages of articles and books.
 
mozartct said:
We are bringing together experts to talk about "unintended consequences of CHIPS on Moore's Law".
Click to expand...
on your original theme, what is "unintended"? If you read the 1965 original it was mostly about economics. It did not require single chip integration - multichip packages were common in 1965 and Moore mentions them. It did not require scaling - Carver Mead's technology projection to future devices around 200nm was not yet published, and Dennard scaling of the essence of MOS constant power per area was years away, and would not be important until 15 years after Moore published. In Moore's day the increased density was driven by ad-hoc innovations.

So essentially Moore said we will package more transistors together, more cheaply, by whatever means we find. What current consequences are unintended?
 
CHIPS refers to the US subsidy program. Nothing to do with ICs! You will need to attend ASMC to find out but I suspect the panelists will say that it may not be a good idea to build factories in high cost areas (e.g. the US) as it could end up weakening the underlying companies and limiting further technical progress. In other words, the money that TSMC spends on cost overuns in AZ does nothing to advance the state of the art.
 
@Tanj I had the wrong official title! This is the real title: Unintended Consequences of Government Subsidies on Moore’s Law and the Future of Semiconductors... Sorry for the error. See you in Saratoga I hope.
 
mozartct said:
Why answer at all if you can only be sarcastic? I have worked for months putting this panel together, reading nearly 1000 pages of articles and books.
Click to expand...
I will be more direct this time; the information you seek will NOT be found by doing an online search. Go to a university library, preferably one that has been in existence for more than 100 years and has degree programs in engineering, science, and most importantly, economics. A research librarian is an expert at doing this type of research; ask them for help, they live for this. The further back in time you want to research, the less likely the data you are interested in has been digitized and is readily available online.

Second, you are not even asking the right questions or the right people. Go talk to a professor of economics about comparative advantage, the origins of the industrial revolution, and the history of the global economy. There is a reason the semiconductor equipment suppliers are located where they are - they grew out of the photography business. Companies like Canon and Nikon had expertise in a complementary field. ASMI (now ASML) is located close to all the German lense manufacturers (Zeiss, Hasselblad, Leica, etc) and semiconductor design house (Philips) so there was plenty of talent available to create lithography equipment and a market for that equipment.

Lastly, I agree with your statements about the CHIPS act; government subsidies are nothing more than a vote buying scheme by politicians that want to maintain their grip on power. Comparative advantage is why TSMC builds their foundries where they do - the talent they need is available at a price that allows them to be profitable and produce goods that consumers want.

Good luck with your research and ASMC.
 
Thanks - I will make sure I read it. I think that going forward, the US military is concerned about maintaining its dominance indefinitely, particularly in the area of AI (for pilotless fighters for example - there are many others). AI is done with chips so you need to restrict chip sales (the 10-7 sanctions) and you need to be capable, at least on paper, of making cutting edge ICs (CHIPS act). NA euv happens to be the way in which cutting edge chips are made and so must be restricted. If it was e-beam lithography, then those OEMs would face restrictions. In that sense, it is different from the 60s where basic functions were adapted to make use of then-new chips, but without changing the balance in the battlefield. Viewed in this way, AI (used in a military context) is more similar to nuclear weapons. Should PRC be first to have AI dominance in military things, the resulting asymmetry might create a first strike mind set. "Use it while you can" thinking. DoD would place a non-zero probability on that outcome hence CHIPS.
 
mozartct said:
Chart showing percentage of chips used by military as a function of total market. I suspect it went from 90% in early 60s to under 1% today
Click to expand...
Suggest looking for "defense" rather than "military" in searches; I found a few, but this one in particular seems to be along the lines of what you're looking for: (published in 1977)


1682432854629.png


1682432996248.png


1682433066600.png
 
@jms_embedded Thanks a lot. Very enlightening. This is what I had written but without data to back it up:

The military was an early adopter of solid-state technology, first in guidance systems and radio, and gradually into everything else. In early years, chips sold for military applications represented the bulk of what was being produced but as transistor costs fell and new end-markets were created, military consumption as a proportion of total production volumes dropped under 2%.
 
Because CHIPS is driven by National Security considerations, issues of the past (when defense purchases were larger as a %) get conflated with issues of today (AI for military purposes could become a strategic weapons systems according to some analysts). In other words:

1- Defense was a big buyer when the overall IC production levels were low (i.e. the 60s)
2- Large civilian markets dwarfed whatever DoD was doing (the 70s)
3- Factories were built overseas and lo and behold, they developed their own comparative advantages over time. (the 80s)
4- Rise of foundries, 300 mm left most US IC making companies in the dust but the decline had started before (the 90s)
5- PRC accession to WTO spurred creation of fabs in China (mostly foreign owned) (the 2000s)
6- Modern military weapons systems require advanced ICs that are no longer made in the USA (the 2010s)
7- AI could be a game changer and that capability must be defined to China (the 10-7 rules)
8- These sanctions make life complicated for companies like Samsung, SK, TSMC, most OEMs so a bone is forged in the shape of CHIPS, Rapidus, European chips act.

In that sense, CHIPS is about the future (AI and other emerging technologies) but justified by the past (look, we once were great!).
 
@jms_embedded The article by Williams and Khan is very good. You can see the consensus building. The critical comments about the science policy approach of the 90s (the so-called road map that we all followed) was most interesting. The establishment of NSTC could lead to the same outcome and not in fact revive domestic manufacturing. I also loved the jab at intel wrt to supply chain. agree wholeheartedly that on the whole, intel was not a good force for the US industry writ large.
 
You must log in or register to reply here.
Back
Top