If EUV is so important … and the United
States helped develop and commercialize it, then why aren't we making these
EUV machines in the United States? In 2001, ASML paid $1.6 billion to acquire
San Jose-based Silicon Valley Group or SVG. SVG was the last major
American lithography company. The decline and fall of the American lithography
industry holds a thoughtful lesson for policymakers. Today's video is a deep dive into
this significant moment in semiconductor history. ## Beginnings I am not going to give you a new update on EUV or lithography here. You can watch
a bunch of other videos about it. But I would like to give a special thanks to
Craig Addison. He did a great episode of his Chip Warriors podcast about America's lithography
industry. Great story and I recommend it so much. Our story picks up a little bit in the
second half of that. Once upon a time, two big American companies GCA and
PerkinElmer dominated the industry. GCA was once an American giant with
$300 million of revenue in 1984. But the machines suffered spotty
quality - frequently breaking down, which shut down expensive manufacturing
lines and angered plant managers. When customer Toshiba suggested a
few ways to improve, GCA waved them off - essentially saying "Buy
what we build and don't bother us". This arrogance cost GCA a great
deal when Toshiba turned to Nikon. Nikon and Canon purchased GCA's machines, modified them, and made them work better.
After Nikon and Canon broke through in 1984, GCA and PerkinElmer struggled. GCA's
revenue fell off a cliff, collapsing 50% in 1985. The company lost $100 million
over two years and cut 70% of their payroll. They had no legal recourse. Apparently
GCA's founder Milton Greenberg never even patented his device. And
even if they had, the Japanese lithography machines had sufficiently
changed enough from the originals. ## PerkinElmer PerkinElmer was the junior company of the
two with about 11% of the market in 1989. That same year, GCA had about 13%. Nikon -
the market leader had 30% and Canon, 15%. PerkinElmer's chipmaking unit struggled to
stay profitable. In the year ended July 1988, they produced just $3 million in
profit on $200 million of sales. The reality of it is that -
for all the sweet and precious words spoken by pundits and politicians -
lithography is not a very good business. You have to spend hundreds of millions of dollars
in upgrades to keep up with the competition. Lithography companies need to spend about 12-25%
of their revenue on R&D and 10-15% on capital investments. These are up to twice the amount
of other semiconductor equipment companies. The customers are very demanding and quite
particular about their machine’s specs. And these machines barely work. They break
down all the time, so servicing is critical. And the semiconductor industry -
particularly the memory industry - is extremely cyclical. You might be up
30% one year and down 30% the next. This really sucks because there are long
lead times between ordering and getting paid. Sometimes as much as two years. So a
customer might put in an order when times are good … and then suddenly cancel when times
get hard. Then you are in for a bad time. In such an industry, the customers really matter.
TSMC helped get ASML off the ground by buying their tools over the Japanese incumbents. And IBM
helped save PerkinElmer's lithography division. ## Saving PerkinElmer In late 1989, news came out
that Nikon was exploring an acquisition of PerkinElmer's lithography business. Nikon denied the talks. But PerkinElmer did
want out. Policymakers soon fretted over the idea of American-made lithography technology
potentially falling into Japanese hands. So did IBM - PerkinElmer's single
largest customer. They were concerned about being forced to use Japanese
lithography technology. Japanese companies were then some of
their biggest competitors. So someone had to be found in order to take
over from PerkinElmer. But the key point was more than just finding a buyer. It wouldn't be too
expensive. No it was more about finding someone who can afford the skyrocketing capital investment
costs of producing better lithography equipment. IBM had this new, more productive lithography
technology. It would eventually become the step-and-scan machine or the stepper. They
gave it to PerkinElmer to commercialize. But who can afford spending millions
more dollars to finish the stepper? ## SVG In comes Silicon Valley Group - SVG. SVG started off as a small equipment
maker. They first made edge grinders for wafers and then expanded
into wafer coating equipment. They then appointed a new CEO named Papken
der Torossian - a Syrian-American who first came to the states in the 1950s.
Torossian decided that the company had to expand to stay competitive as the
big guys shrank their supplier base. So he bought Thermco one of the leading makers
of vertical furnaces for thermal oxidation. Thermal oxidation is a way to apply an oxide onto something like a silicon
wafer using heat and oxygen. He then cross-sold SVG products through
Thermco's very established user network. This eventually helped grow the company
and bring it to public prominence. Though they were still very small - about
$120 million market cap at the time. Torossian wanted to take over PerkinElmer’s
lithography business and worked hard to overcome skeptical feelings from both IBM
and PerkinElmer - lobbying to PerkinElmer chairman Horace G. McDonell.
He eventually convinced them. So in the end, they paid just $20 million
for a big stake to create SVG Lithography. The new SVG also accepted minor equity
investments from IBM and PerkinElmer. SVG Lithography took the nascent IBM
lithography technology and with help from Sematech eventually produced the
Micrascan II, a DUV stepper machine. ## General Signal Meanwhile in 1988, GCA was sold to General
Signal Corporation for about $80 million. General Signal was a grab
bag technology conglomerate of random stuff. Some of GCA's siblings
included telecom gear and signal boxes. Thanks to ample amounts of financial support
from the US Government, GCA improved their tools' reliability and produced a machine more
technologically advanced than the Japanese. It is estimated that Sematech infused
some $60-80 million into GCA in the five years after 1988. Support
came in the form of investment, direct purchases of GCA equipment for testing
and feedback, and government arm-twisting. In 1990, Sematech started pushing its
members to try GCA's upgraded equipment. The companies resisted, but Sematech
insisted - saying that lithography was 50% of Moore's Law and that Japan would
never give America its best equipment. As it turns out, the new equipment
was good. Customers acknowledged the vastly improved quality. GCA's XLS
series of steppers were even said to be "world-class". The tools were central to
achieving the upcoming 350 nanometer node. In 1992, Digital Equipment Corporation chose
GCA tools over those from Canon and Nikon for their next line of chips. Larry Thompson,
head of lithography at Bell Labs, said: > From our viewpoint, the performance of
GCA's latest DUV stepper the XLS 7800/31 indicates that GCA is at least 12
months ahead of the competition Brian Clark of Motorola added: > GCA has made great strides
in the last few years ... it will be a travesty if it does not find a buyer Yet the XLS was too little too late. And an
18-month delay in 1990 in producing i-line lithography equipment nixed their last
chance to take market share. By 1992, the once mighty GCA had just 2% of the market. And they still lost money - about $4
million a quarter in the early 1990s. ## The Fall of GCA General Signal eventually tired of these losses. Their CEO wanted to focus on businesses
with returns on equity over 20%. So in early 1993 he announced that
General Signal would attempt to sell several of GCA's semiconductor equipment units. Yet again there was another political huff. More grandstanding by people about
the national security value of an American lithography industry. But no
one wanted to back that talk with money. Sematech scrambled to find another suitor and
arrange another shotgun marriage. Foreigners feared political blowback - though one of
the Korean chaebol did kick the tires a bit. Domestic players did not really take to
the idea. A merger with Applied Materials might have been interesting, but for
whatever reason they turned it down. And in terms of customers, IBM had arranged
the marriage between PerkinElmer and SVG, but there was no similar player for GCA. And as for government, Sematech said that they
would not involve themselves in business rescues. They had already spent the majority of their $100
million annual budgets on lithography development. In the end, General Signal sold off bits
and pieces of GCA like the lens and service businesses. But there was no saving
the large wafer stepper manufacturing facilities in Massachusetts. It closed
down in May 1993, with 120 jobs lost. ## The Last American Lithography Company With this, SVG became the last
American lithography company. Over the next few years, SVG
continued developing its DUV machinery. The political environment
surrounding Japan loosened up as America's semiconductor comeback in
the early 1990s played itself out. Throughout the decade, SVG was deeply involved
in next generation lithography development. As early as 1994, SVG was investigating new
forms of ultraviolet light lithography - taking some $30 million from Sematech in the
form of a direct investment to do so. In 1996, Intel, AMD, Motorola and the
other American semiconductor companies signed agreements with the Department of Energy
to found EUV LLC. This strange joint venture would carry out EUV's development
over the next three to four years. In late 1997, SVG signed an
agreement with EUV LLC. They were the first American lithography
maker to enter the partnership. A year later, ASML and a subsidiary of the American semiconductor manufacturer
Ultratech Stepper joined as well. Ultratech had been one of GCA's siblings at
General Signal. They produced a different type of lithography machine - 1x lithography
- for specific, unconventional purposes. ## 193 nanometers Then SVG took their eye off the ball. In the late 1990s, the semiconductor
industry transitioned from 248 nanometer krypton-fluoride DUV steppers to
193-nanometer argon-fluoride DUV. In September 1998, ASML announced the
move to 193 with the PAS 5500/900. The transition required great amounts of
materials engineering by Carl Zeiss to produce a unique lens system made
from Calcium Fluoride and fused Silica. In February 1999, SVG also announced a move
to 193 nanometers with the Micrascan 193. They also announced a 157-nanometer project
thanks to additional equity funding from Intel. But the Micrascan 193 suffered massive problems
and repeated delays. In August 1999, SVG reported that customers had canceled $53 million worth of
193-nanometer orders. It was a devastating blow. Intel took steps to shore up confidence
in SVG. In December 1999, the CPU giant announced that it would purchase $100 million
worth of 193-nanometer DUV equipment from SVG. But the delays dragged on, forcing Intel to
delay the rollout of their 130 nanometer node. The 193 nanometer debacle sparked rumors
of trouble at SVG. In terms of 1999 sales, SVG was a distant fourth place
- generating just $270 million in sales compared with ASML's $1.12 billion. SG Cowen estimated that after
the August 1999 cancellations, Intel had 90% of SVG's 193-nanometer DUV
machine orders. SVG's advanced lithography program was basically Intel and nothing else. It was becoming increasingly clear that
SVG could not compete. And looking ahead to 157 nanometers or even EUV, SVG simply did not
have the money to bring the device to fruition. Torossian told people that it was going to take
a billion dollars to bring EUV to the fab floor. He had no idea how SVG was going to get this
money. In actuality EUV would take far more. These concerns only intensified as the
dotcom bubble burst in the 1999-2000 period, taking down the semiconductor and semiconductor
manufacturing equipment markets with them. ## The Merger In October 2nd 2000, ASML
and SVG announced a merger. ASML would acquire SVG for $1.6
billion in an all-stock transaction. The deal would make ASML the largest
lithography company in the industry. Behind the scenes, SVG's struggles with 193
nanometers alarmed Intel, their biggest customer. Intel sourced from two lithography makers - Nikon
and SVG - and they wanted to keep it that way. If SVG were to go down, it not
only constricted the supplier base, but also risked disaster when it
came to next generation lithography. ASML had successfully shipped 193 nanometers
and was keen to have Intel as a customer. But if Intel left, then SVG would surely collapse.
So the best way to handle things would be a merger. The two companies can combine
their R&D spend and develop EUV together. ## Controversy The big problem however was the politics. SVG was America's last lithography
maker. Sematech and the US government put millions of dollars of investment into
the company. ASML was a Dutch company. The US and the Netherlands are friendly,
but this would be a big export issue. Furthermore, one of SVG's big subsidiaries was
Tinsley Labs - the optics system manufacturer that SVG acquired in 1997. They were leading the
effort to produce the 157 and EUV optics systems. The issue was that Tinsley had once been a defense
contractor producing optics for spy satellites in the 1990s, but SVG claimed that currently there
was no business being done with the Pentagon. Acquisitions by foreign companies
need to be approved by CFIUS, the Committee on Foreign
Investment in the United States. At first, approval seemed a slam
dunk. The panel at the time was seen as a rubber stamp. But near
the end of the 30 day review period, a group of Congressmen asked federal
officials to do a second look. It is likely that this interruption came at
the behest of Ultratech - the other American lithography system integrator
within the EUV LLC consortium. They had circulated around Congress 650 copies of a video tape titled “Why The Sale Of
SVG Co. Is Bad For The United States”. In the tape, they talked mostly about
Tinsley Labs and national security. But they also voiced concerns about
leaving ASML with a monopoly on next generation lithography machines, saying
that they would raise prices on vendors. They also speculated that ASML would give
preference to European semiconductor makers over American ones. And that ASML might send EUV
technology to "enemies" like China and Russia. But Intel refused to be dissuaded.
They steadfastly insisted upon the acquisition. Their CEO Craig
Barrett met with President Bush and Vice President Cheney to
lobby for the deal. Barrett said: > Without the merger, the development path to
the new tools in the U.S. will be delayed and, in time, other global competitors
of SVG will seize the advantage This is a reference to Japan’s
own EUV efforts - then ramping up. Which ASML and America were able to fend off. ## Close Finally in May 2001, ASML promised that they
would search for an American buyer for Tinsley Labs. They replaced the vendor with Carl Zeiss,
and things seemed to have gone along fine. ASML also agreed to strictly comply
with US export control requirements. And they agreed to keep and invest in
certain operations in the United States. A few months after closing the deal, ASML dropped
development on SVG's 193-nanometer Micrascan tool. The tool's ultra-complicated optics system
was too slow and apparently was beyond saving. Intel took up the Twinscan as they
probably wanted to in the first place. In 2003, ASML sold SVG's two other
businesses - the thermal furnace and track businesses. Within two
years, ASML basically shut down SVG. They did retain SVG's R&D centers including ASML
Wilton - today their largest R&D hub in the US. ## Conclusion Robert Noyce himself said in a Congressional
hearing that Sematech's legacy would be in large part remembered for how successful it
is in saving America's lithography machines. Sematech did a great many things. It is
in no little part thanks to them that American companies like Applied Materials, Lam
Research, and KLA remain leaders in their fields. But Sematech could not save the American
lithography industry. Intel at the very peak of their powers could not save
the American lithography industry. The reasons behind America's lithography failure remain up for debate. But it should be
discussed as we enter this new era of government-led industrial policy for
semiconductors. What do you think?
