Now that the US is increasing its pressure on China, what can we learn from the semiconductor war with Japan in the early eighties?
At the end of the seventies, American semiconductor companies got a heavy beating. Their customers told them to their faces: you’re losers. In his book “The microprocessor, a biography,” Michael Malone describes the turmoil in the chip market with an anecdote from a 1979 conference in Boston. There, an executive from Hewlett-Packard presented a thorny issue to his suppliers. He explained that HP had bought in volume several hundred thousand components from outside manufacturers for use in its computers, calculators and instruments. Most of these devices had been purchased nationally but a few had been sourced from big Japanese electronics firms like NEC, Hitachi, Fujitsu, Toshiba and Mitsubishi.
Testing the US chip deliveries versus those from Japan, HP found out that the latter weren’t just better, but shockingly so. Whereas the computer maker typically had to return one in every twenty US chips, the Japanese devices rarely were faulty. What’s more, initial deliveries from Japan were always on time, and any replacements were made immediately. US chips were usually late and replacements were sometimes made with equally faulty chips. “The complaining customer, HP, was typically treated as if it should feel lucky to be getting any replacements at all,” writes Malone.
Same story for semiconductor production equipment. It started out with American superiority, with Perkin-Elmer and GCA as the superior suppliers of chip lithography machines – Perkin-Elmer from 1974 on with its mask aligners, GCA from 1978 on with its wafer steppers. Then Canon began to copy Perkin-Elmer and GCA found out that Nikon’s machines resembled its steppers in miraculous detail.
While chip manufacturers were swamping GCA with wafer stepper orders in the late seventies, Canon and Nikon were putting on their running shoes. The Ministry of International Trade and Industry (MITI) lead the Japanese semiconductor industry’s charge to quash American dominance with a VLSI project. In keiretsu – groups of affiliated firms – the companies worked closely together with chipmakers to perfect their machines.
Compared to Canon and Nikon, GCA was at a significant disadvantage. It didn’t have an intimate relationship with customers. American chipmakers distrusted their machinery suppliers and didn’t want to share the details of their production technology – afraid as they were that the information would make it to competitors. They’d rather service the machines in their fabs themselves than get suppliers involved. As a result, US machinery manufacturers had a harder time refining their technology and they were on their own in charting an R&D roadmap.
By contrast, machinery developers in Japan worked closely with process operators and chip fabs. They learned from each other. Almost three decades later, ASML and TSMC proved that this is crucial as they finetuned and matured the mass production of chips with EUV technology.
When Japan started preparing to mass-produce 64-kilobit DRAMs at the start of the eighties, Canon and Nikon saw formidable opportunities opening up. They served their compatriots well and had a guaranteed outlet for their wares – Japanese chipmakers had been burned by relying on American machinery manufacturers. In the late seventies, GCA couldn’t keep up with demand and prioritized US semiconductor companies, leaving its Japanese customers out in the cold. The Japanese were determined not to let that happen again.
At the beginning of the eighties, the computer industry needed increasing quantities of memory chips. The Japanese were producing them in abundance. Analysts pointed to the growing influence of Canon and Nikon in DRAM manufacturing. In 1984, the Japanese already had 40 percent of the global market. Two years later, their 256-kilobit DRAMs would help them conquer a staggering 90 percent stake.
The Japanese victory left the American semiconductor makers in shock. They considered a hegemony in chip technology their birthright. Bell Labs had invented the transistor, Intel and Texas Instruments had created the integrated circuit and in the early seventies, Intel had developed the microprocessor. DRAMs and non-volatile memories were American inventions as well. In short, to the Americans, the chip industry was America.
A tiny company in The Netherlands profited from these developments. Born in 1984, ASML wouldn’t normally have stood a chance in a mature market. But the mistakes made by GCA were numerous. Its management lacked focus. Things went bad when a lengthy chip crisis hit the company hard in 1984 and 1985. Carl Zeiss, GCA’s main supplier of crucial lens technology, was willing to deliver its optics to the Dutch as well. In 1986, the former market leader in wafer steppers was in such a bad condition that it was offered for sale to newcomer ASML.
In the ASML context, it’s worth noting that one American company did manage to innovate in memories. In the midst of the explosive growth of Japan’s market share in DRAMs, Idaho-based Micron Technology started producing competitive 64-kilobit DRAM memories in 1981. Because its designs required fewer lithographic steps, it could produce its chips more cheaply. Micron became a loyal ASML customer. In the early 1990s, it was in fact ASML’s lifeline, together with TSMC. Both helped the Dutch survive some very lean years.
In 1987, the US semiconductor identity crisis resulted in a consortium of American semiconductor makers supported by the Reagan administration with an investment of 100 million dollars. The goal of this consortium, called Sematech, was to revive the entire semiconductor industry. As Michael Malone wrote in his biography of the microprocessor, US companies successfully innovated in microprocessors and digital signal processors, both much more lucrative segments of the semicon market than memories.
But chip litho proved to be a completely different ball game. Sematech also started funding the technology to print chips, after warnings from officials at the Pentagon and analysts at the CIA about threats to national security. They pointed to the fact that GCA was among the last American companies to make the equipment that was the centerpiece of every semiconductor production line. “It’s very, very critical,” an official from the Pentagon commented in the New York Times, January 1987. “This type of equipment is the key to producing finer and finer resolution semiconductor devices. It’s simply something we can’t lose, or we will find ourselves completely dependent on overseas manufacturers to make our most sensitive stuff.” Andy Grove warned that the US was in danger of becoming a tech colony of Japan.
Despite Sematech’s research funding, litho specialists struggled to master the development of complex lithography systems. In 1991 the US General Accounting Office concluded that the development of GCA’s new generation of wafer steppers was running behind schedule and risked missing a ‘window of opportunity’ for selling the leading-edge generation of equipment.
By the time Reagan left office, efforts to boost government spending started to become controversial. Sematech opened up a deep ideological divide between Democrats and Republicans about the role of the American government in boosting industries and particularly high tech. Republicans said by and large that the market works best in innovative industries. Every dollar spent there was madness. Intervening in the commercialization of inventions would be a waste of taxpayer money. Democrats pointed to the funding of the basic research that had created the space program and had not only landed men on the moon but had also spawned the military and commercial satellite business.
This bitter partisan warfare continued in the US Senate for decades, only to be ended a few weeks ago. Driven by perceived American weaknesses in semiconductor manufacturing technology, members of Congress with a 68 to 32 majority passed a historic spending bill that’s set to inject 250 billion dollars in high tech. The funds are earmarked for everything the American industry worries about, from AI, autonomous vehicles and robotics to synthetic biology. Semiconductors, the steel industry of the modern age, are the biggest part of it.
A US Senate that couldn’t agree on anything is now united in spending 250 billion dollars over five years. Even by Washington’s standards, that’s a lot of money. More so, when you take into account that the Americans are still the biggest force in the semiconductor arena, with a 47 percent market share in 2019, according to the Semiconductor Industry Association. The reason is, of course, China, which is pouring vast amounts of money in its goal to become technologically independent.
The semiconductor world is looking quite different now. How do Japan’s ambitions in the 70s and 80s compare to those of China today? Apart from an economic force, the big difference is that China has also become a military superpower.
The US controls almost every aspect of the semiconductor market, including chip production equipment. Not only domestic companies like Applied Materials and KLA are under Washington’s reign, but ASML looks to be under there as well. From a geopolitical perspective, with the Dutch company having crucial EUV source production in the States, the US has virtually total control over Veldhoven.
The US semiconductor ecosystem has a weak spot, though – a rather big one, unfortunately. Because of economies of scale, lower capital investments, shareholder pressure to reach higher margins and to reduce risk, big semicon manufacturers like IBM and AMD started to outsource their production to TSMC after witnessing the success of fabless players like Broadcomm, Qualcomm and Nvidia. This strategy has rendered the biggest part of US semicon dependent on Taiwan, an island that China considers its territory. Certainly after Apple switched from Intel to TSMC, the majority of America’s most advanced chips are made there. Most of the value US companies create is coming from TSMC’s factories.
The American people are paying the bill for this business strategy – short-sightedness if you will. Joe Biden calling semiconductor production “infrastructure” says it all: this is how Washington makes it politically acceptable to repair a hole of hundreds of billions of dollars that the US semicon industry created itself.
Despite the differences between Japan then and China now, there are some lessons to be learned. In any case, copying and stealing isn’t a recipe to conquer the world. Also: technology is nobody’s birthright. China has become a technological power to be reckoned with, but America has to trust its own innovative power. In the nineties, the US innovated itself out of its self-induced crisis by developing superior microprocessors, DSPs and later FPGAs and graphic processors. Micron outperformed the Japanese in memory designs. Later, Rambus did the same. With a whopping 47 percent of the market, and total control over ASML, the US has nothing to fear in the semiconductor market.
But the American leaders are playing a dangerous game. Largely driven by fear, they’re cutting off China from chip production technology. This way, they’re strongly motivating their rival to develop the crucial production technology on its own – or just take it by invading Taiwan.