The Rise and Peak of Japanese Semiconductors
If you want to watch the video, it is below:
Japan’s semiconductor history is utterly fascinating to me. There are some other really good articles about this period of time. I recommend this great piece from Doug of Fabricated Knowledge
Thanks for reading The Asianometry Newsletter! Subscribe for free to receive new posts and support my work.
It is hard to understate just how big of an impact that the Japanese consumer electronics-makers made on the world. I just finished a piece about Swedish mechanical calculators - which still dominated the world as late as 1969. Then in a span of three years, they were gone - sunk under a rogue wave of cheap Japanese digital calculators.
Japan's semiconductor story is unique in modern technology and business.
Coming out of World War II, the country rapidly gained competence in an emerging technology and became a global leader.
In this video, we look at the 30-year rise and peak of the Japanese semiconductor industry starting from the 1950s into the 1980s.
How about we start at the beginning? Our story begins in America with the transistor's invention at Bell Labs - a division of AT&T, the American telephone monopoly.
The company kept the invention a secret for six months while it finished its patent application. Then publicly announced the breakthrough in the summer of 1948.
At this time, Japan was occupied by the Allies after World War II.
A member of the Japanese Ministry of International Trade and Industry or MITI - Watanabe - befriends a former Bell Labs engineer working at the Allied Occupation General Headquarters or GHQ.
Shortly after the Bell Labs announcement, the American engineer tells Watanabe that he has a report about a solid-state amplifier that had just been invented back in the United States.
The engineer says to Watanabe that the report is top secret and cannot be shared.
But then - right in front of Watanabe - he puts the report down on his desk and walks out of the room.
Watanabe gets the unspoken message and takes the report home to read and study.
He and the rest of his colleagues at MITI recognize the importance of this discovery and directs the country’s industry to come up with their own indigenous versions.
Learning the Transistor
Japanese transistor learning efforts were wide-ranging. The most important of which were coordinated by MITI through its Electrotechnical Laboratory or ETL. ETL would later spin off as Nippon Telephone and Telegraph or NTT.
Japan was still recovering from its ill-advised participation in World War II. So researchers had a hard time not only finding equipment but also enough food to eat. International travel was difficult and the country had little foreign capital with which to buy foreign technologies.
ETL began with a simple transistor study group led by Watanabe and other ETL researchers. The ETL director flew to the United States and brought back transistor samples. The GHQ had set up libraries in Japan and those libraries had subscriptions to Bell technical journals. The researchers read them.
And Japan's Physics or Physical Society held semiconductor symposiums where researchers presented their findings. The first of which took place in April 1950.
By 1951, Japanese scientists at NTT and Nippon Electric Company or NEC succeeded in creating proto-versions of the transistor. However, this remains far behind what the rest of the industry had to offer.
In 1953, a company called Tokyo Tsushin Kogyo sees in a news circular - obviously the Asianometry newsletter - that Western Electric - the manufacturing arm of the AT&T Bell System - is openly licensing their transistor manufacturing patents for a fee of $25,000 and a 2-3% royalty.
You might be wondering why AT&T would do this. After all, wasn't the transistor a precious technical breakthrough and a foundation of national security? Why didn't they protect it?
Yes, in the long run. But at the time, to quote a famous computing figure, you cannot connect the dots looking forward. Only when looking backwards.
The 1948 Bell Labs germanium transistor was crude. And its technical benefits over existing vacuum tube technology were not immediately apparent - especially prior to the shift to silicon. Germanium's finicky nature was a real manufacturing issue.
Furthermore, when Bell invented the transistor, concerns immediately sprouted about the American telephone monopoly having exclusive control of such a valuable technology. They were already in the process of fending off an anti-trust suit from the US government.
Licensing out transistor technology to other firms was a way to circumvent those public concerns. 35 electronics companies took up Ma Bell's license offer. This included several American manufacturers like GE, Raytheon, and Texas Instruments.
Anyway, this small company in Japan - just 120 employees at the time - petitions MITI for the funds to do this transaction. Without waiting for MITI's approval, they reach out to Western Electric and strike an agreement. MITI was annoyed that this little startup moved forward but gave approval nonetheless.
Ever heard of this company Tokyo Tsushin Kogyo? No? Perhaps, you might be more familiar with its later name.
Sony and the Radios
America invented the transistor. But Japan made it a must-have for any household consumer item with the popularization of transistor-based radios. The radio was the first portable killer app.
In 1954 Texas Instruments released the first commercial transistor radio: The Regency TR-1.
The TR-1 was smaller and more portable than anything before it. But its product features - audio quality and what not - lagged that of its vacuum tube competitors due to cost cutting measures.
Despite the issues, anyone who looked at this product knew that it was the future. A radio receiver in your pocket! Imagine that! Analysts predicted sales of "20 million units in 3 years". Every Japanese company worth their salt worked towards making their own.
Kobe Kogyo, an established company producing electron tubes, struck a deal with American giant RCA to receive their transistor technology. By 1954, they had announced their first transistor radio. This radio worked but product and production were not good enough to sell to the public.
It left the door open for a company to go for broke. And Sony wanted in. After some convincing, the small startup acquired the Bell Labs patent package - which consisted of a few sample transistors, a piece of high-grade germanium crystal, and a pretty unhelpful textbook - Bell Labs' "Transistor Technology".
More importantly, Sony engineers had a chance to spend three months at a Western Electric plant. They sketched and sent back detailed notes of what they learned there.
In order to produce a workable radio, Sony needed to produce a new type of transistor - the grown junction transistor - at an acceptably high yield. By 1955, Sony's yields were at 5%, less than half of that in America.
Undeterred, the company started producing and selling transistor radios, which was internally controversial. It was immensely risky to believe that their team could raise yields high enough to turn a profit. With Sony borrowing to fund its expensive R&D work, it was a bet-the-company decision.
That radio - the Sony TR-55 - was the first transistor radio commercially available in Japan. Its audio quality was still not that good, but it was small, portable, and had great battery life.
The Sony radio sold very well, especially abroad. In 1955, less than 5% of the value of Japan's exports were electronics. By 1960, that number had become 16%. That year, Japan exported 10 million transistor radios to the United States alone.
Within a few years, old-guard electron tube firms like NEC, Toshiba, and Hitachi released their own radios. The aforementioned Kobe Kogyo managed to get their yields up and entered the market in 1957. They eventually merged with Fujitsu in 1964.
The transistor radio market crashed as products from eleven Japanese electronics firms hit the market. Sony survived the crash by introducing other products like the FM transistor radio and eventually the Television.
By 1960, Japanese transistor production reached parity with that of the United States. But by then the next big thing was already on the rise.
Rise of the Integrated Circuit
In 1959, Robert Noyce of Fairchild independently invented the first monolithic integrated circuit. This IC was far more commercially viable than Jack Kilby's hybrid integrated circuit, which had been invented the previous year.
This is because the monolithic integrated circuit had all of its transistors on a single piece of silicon - making it far easier to commercially produce.
After patenting this technology, Fairchild licensed it to NEC. Just three years after its invention.
This time however Japanese IC companies found it much more difficult to catch up with their American competitors for largely two reasons.
First, the American IC industry was moving very fast. The commercial market was taking over from the government as the main market. The mainframe markets were exploding, fueling demand for more sophisticated integrated circuits.
From 1963 to 1965, the average IC selling price fell from $50 to under $9. Production grew twenty times over.
Second, much of the necessary IC fabrication equipment - in areas like lithography, sputtering equipment, and ion implanters - came from American companies. The Japanese companies had to scramble to import equipment over from the US, and this was not easy.
Throughout the 1960s, MITI's policy focus encouraged technology imports. The way they subsidized the licensing and acquisition of early transistor technology, for instance. Or by blocking foreign-made electronics from entering the lucrative Japanese market without a technology transfer agreement.
This technology import policy led to some competency catch-up at the start and even a few commercial successes. For instance, the development of a small instrument manufacturer cluster in the South Tokyo area.
It soon became clear however that the Japanese integrated chip makers were not catching up to the Americans. This was especially the case in the newest, hottest thing in the industry: Memory circuits. More specifically - Dynamic Random Access Memory, or DRAM.
Roughly speaking, DRAM is used in computers to store information while the power is on. It provides high capacity - measured in units called kilobit, megabit, and gigabit - at a low cost.
DRAM - like other types of semiconductor memories - is a commodity. Customers like desktop, notebook, and server makers can easily switch suppliers depending on price or features.
From 1974 to 2000, the average DRAM product life cycle was about 2-3 years. Price fluctuations are wild and brutal. For instance, 2007 saw 1Gb DRAM memory quoted at about $10. A year later, that collapsed to under a dollar due to oversupply.
Because of this, product leadership in memory is even more critical than it usually is in the semiconductor industry. You want to be first into the market with a product so that you can price high and make back invested capital before the rest of the market catches up.
This gets harder over time as the capital costs of new fabs and chip-making equipment continue to rise. It means you have to sell more memory units and so on.
At the start of the 1970s, the United States took the early lead in the DRAM market. Intel and Advanced Memory Systems delivered a 1K product for the mainframe market. Intel took the market lead due to its cost-effectiveness.
The Japanese attempted to catch up to the Americans. But whenever they succeeded in replicating a memory product, the Americans would jump ahead to the next generation. This happened as the industry transitioned from 1K to 4K, and then again to 16K.
Without better chips, the Japanese mainframe computer industry can never be competitive against the American giants. Namely, IBM. At the time they were barely holding on against the juggernaut.
In 1971, MITI reorganized the Japanese computer industry from six separate companies into three paired-groups. But the pairs continued to compete so MITI directly sought to organize and coordinate R&D across the entire Japanese semiconductor industry to enable the next generation of Japanese computers.
This became the VLSI Project in 1975 - which stands for Very Large Scale Integrated Circuits.
The VLSI Project gathered together five of Japan’s biggest semiconductor companies - Fujitsu, Hitachi, Mitsubishi Electric, NEC, and Toshiba - normally fierce rivals. Consumer electronics were excluded, which is why Sony didn't make the cut.
The project assigned various parts of the microfabrication process to each of the participants' research staff - backed by ample government funds.
The project kicked off in 1976 and ran four years until 1980. It resulted in over a thousand patents on all parts of the microfabrication process.
Critically, the project helped spur Japanese leadership in two key industries: Lithography and silicon crystal technology. Japan took technology leadership in these areas and its semiconductor industry shifted almost entirely from American suppliers to domestic ones.
I also discussed the VLSI Project in another video about the Japanese photolithography industry. For additional detail, you might want to check that out.
The next step after 16K was 64K. At this point they were behind by one generation
The Japanese DRAM industry decided that they needed to develop the manufacturing processes for both 16K and 64K simultaneously. The learnings from the VLSI project helped make it possible.
Thus, in April 1977 NTT announced the successful development of 64K DRAM, first in the world to do so.
Two years later, NEC, Hitachi and Fujitsu all brought 64K to the market, at the same time as the Americans. Erasing the American lead.
Furthermore, the Japanese product was better. Though the Americans invented and pioneered automation techniques on the semiconductor production line, the Japanese mastered it.
Of key importance was investment in computer-controlled bonding processes - an assembly step where you attach finished ICs onto a board.
Americans opted to use cheaper offshore labor to do this finishing step.
The Japanese on the other hand automated it completely and ended up ahead on their investment.
A 1980 Hewlett-Packard study comparing the quality of the Americans and the Japanese found that products from the best American firm still had 6 times more errors than the worst Japanese firm.
Japanese yields were hitting 70-80%, while the Americans topped out at 50-60%. So not only were Japanese products better, they cost less too.
The DRAM market cratered into a nuclear winter. After 1985, just two American companies stayed in the business: Micron and Texas Instruments. The rest were Japanese firms and one Korean, Samsung.
Why It Worked
There was supposed to be a second VLSI Project. But the US protested government funding for such R&D projects as an unfair trade practice. Nevertheless, the project emphasized the policy importance of semiconductors and a frenzy of VLSI research in Japan soon followed.
Researchers have studied why the 1976 VLSI project succeeded. The concept came from the United Kingdom where Research Associations would spring up in response to government grants. But similar public-private partnerships abroad haven't been as fruitful.
The best answer would probably be that the VLSI Project came at the right time. The semiconductor industry back then wasn't as diverse or complicated as it is today. Personal computing, the iPhone, the cloud, and AI. None of that existed then. All you had were mainframes. This allowed for focus.
Furthermore, the 1970s were a time when the cost of chip-making factories were starting to skyrocket into the billions of dollars. Yet it was still possible to achieve large technical gains simply by driving process manufacturing improvements.
Looking back at it, the Japanese rise to the top of the semiconductor mountain was an economic miracle. Like many other things in Japan, it made no sense - full of close calls and paradoxical policies.
But every action has an equal reaction. Japan's rise triggered intense reflection within the American semiconductor manufacturing industry. How did America fall so badly? And unlike with other manufacturing industries, they bounced back with ferocity.
Then a number of other countries started copying the Japanese playbook. Most notably, an ambitious, fast-following semiconductor giant over in South Korea.
To modify a Japanese saying: America pounded the rice, Japan knead the dough, and South Korea ate the DRAM cake. Japan's reign as king would not last for very long.
Thanks for reading The Asianometry Newsletter! Subscribe for free to receive new posts and support my work.