Review

This book attempts to provide a complete history of the semiconductor industry and the global competition for chip manufacturing and innovation. It covers the early history of Silicon Valley, the rise of Japan's chip industry, the emergence of Taiwan and South Korea as dominant players, the challenges faced by Intel, and the growing influence of China and the importance of the Taiwan Strait.

My interest in semiconductor history was sparked by reading a blog post by Steve Blank, which may have been sufficient. The book's 450 pages although well written, wasn’t insight dense enough to justify the investment. However, it did provide me with a better understanding of the history of semiconductors and some genuine insights about competition and innovation. Although, if you want to learn about competition and innovation, those concepts are better presented in alternative works like 'The Innovator's Dilemma'.

You Might Also Like:

Book Review and Summary: Inmates Running the Asylum by Alan Cooper

Get a comprehensive review and summary of the best-selling book "Inmates Running the Asylum" by Alan Cooper, the software development expert and author. Learn about the key principles and practices for creating user-friendly software and avoiding common pitfalls in the development process.

andrewclark.co.uk

Book Review and Summary: The Dream Machine by M. Mitchell Waldrop

Get a comprehensive review and summary of the best-selling book "The Dream Machine" by M. Mitchell Waldrop, the computer science and technology history expert. Learn about the key principles and practices of the pioneers of computing and the development of the internet.

andrewclark.co.uk

Book Review and Summary: Where Good Ideas Come From by Steven Johnson

Get a comprehensive review and summary of the best-selling book "Where Good Ideas Come From" by Steven Johnson, the innovation expert. Learn about the key principles and practices of creativity and innovation, based on Johnson's own research and insights.

andrewclark.co.uk

Key Takeaways

The 20% that gave me 80% of the value.

• By production numbers, the transistor is the most successful invention of all time.
• A tiny number of companies control the worlds chip production
• Fabricating and miniaturising semi-conductors has been the greatest engineering challenge of our time
• Many manufacturing innovations have been required to maintain Moores law, with a focus on increasing both the density and production yield of processors
• TSMC is now etching shapes half the size of a corona virus
• The chip supply chain is international and complex. It has a there are many choke points. No other part of the economy is so dependent on so few companies. Both ASML and TSMC are critical to the production of the most advanced chips.
• The Cold War accelerated investment into semi-conductors. USSR had a ‘copy it’ strategy of theft and replication, but that kept them years behind the US.
• Japan won chip dominance in the 1980s by investing heavily. Eventually they would over-invest and create too much supply. They also alienated the US who made partnerships elsewhere in Asia.
• Taiwan was threatened by China’s development of nuclear weapons. They made a conscious decision to align and integrate with the US and used semi-conductors to do it
• They put up 48% of the captital for TSMC and gave it tax breaks
• TSMC promised never to design chips, only to build them. So there was never a conflict of interest
• Fabless firms started to emerge (Nvidia, Qualcomm)
• Grove and Clayton Christianson knew each other. Grove was aware of the disruptive innovation theory.
• Grove was ruthless and decisive
• He got out of the memory market and bet big on micro-processors (expecting personal computers to be a big thing)
• The thought experiment that pushed him over the edge was to ask… What would a new CEO do?
• The US showed the power of next generation weaponry for the first time in the first Gulf War (laser guided bombs, cruise missiles)
• Most of the money is made by the companies designing and producing the most advanced chips
• The cost of manufacturing equipment and fabrication rises with each generation of chips
• Chip fabrication is only profitable at huge scale, the more you produce the stronger your monopoly becomes
• Chip makers had to spend more, on more advanced manufacturing processes… every 2 years to keep up.
• Eventually only Samsung and TSMC were left in the fabrication race
• Samsung has it’s own line of hardware, which makes it a competitor to many
• Lithography is incredibly difficult, and would become dominated by ASML (a Phillips spin off)
• ASML sources components from other companies, they rely a lot on suppliers and historically have bought those that weren’t run well.
• EUV lithography (Extreme Ultraviolet) uses shorter wavelengths of light.
• It took decades to develop and required many innovations (focusing optics, ultra pure diamonds, new lasers)
• A state of the art ASLM machine has hundreds of thousands of parts and costs $100m
• The investments required in EUV were so big, that if it didn’t work there was no plan B.
• Only TSMC, Samsung and Intel invested in the EUV machines
• Intel spent heavily on innovation ($10b a year through the 2010’s, made 1/4 trillion dollars of profit) and should have won the EUV era. But TSMCs scale won out (produced 3x more chips)
• ARM came from nowhere to dominate the mobile chip industry (33% of chips)
• It took advantage of RISC architecture allowed more efficient chips
• Intel didn’t make the transition because it had too many customers on x86
• The original iPhone used components available to others but Apple quickly started to design it’s own silicon.
• China launched an initiative ‘Made in China 2025’ to import less chips by creating their own. China has spent billions subsidising investment in chip fabrication.
• The US is starting to block Chinese investment and blacklist them from procuring key bits of technology
• The future of war will be defined by computing power
• Globalisation was starting to look concentrated in Taiwan. TSMC has become indispensable.
• Peace in the Taiwan straight is incredibly important
• It would take years for the West to replicate the TSMC fabs
• There are many choke points in the chip process. Components or software that can only be built by one or two firms
• Chip shortages were common after the pandemic. Supply issues were blamed but more semi-conductors were produced in 2021 than ever before. Demand growth was actually the problem.
• Many are declaring that Moore’s law is dead. Although, its survived longer than anyone thought and there’s a conceivable path to increasing transistor density by another 50x on chip

Deep Summary

Longer form notes, typically condensed, reworded and de-duplicated.

• By production numbers, the transistor is the most successful invention of all time. We produce more transistors each year than all the other goods combined
• Apple sells more than 100m iPhones of each generation (each with 11.8 billion transistors)
• A tiny number of companies control the worlds chip production
• iPhone’s chips can only be produced by a single company in a single building. The most expensive factory in human industry. AT TSMC.
• Fabricating and miniaturising semi-conductors has been the greatest engineering challenge of our time
• Moore predicted every year for the next decade Fairchild would double the number of components that could fit on a silicon chip
• More compute + lower prices = more customers
• Moores Law required increasing complex machinery, designs, materials and processes
• Many manufacturing innovations have been required to maintain Moores law.
• Both in reducing the size of processors, density of processors and production yield
• TSMC is etching shapes half the size of a corona virus
• Photolithography (imagine turning a microscope upside down) is the process of printing with light, it has helped make transistors smaller
• The Supply Chain is international and complex…
• Designed by ARM (UK based, Japanese owned)
• Using special software from 3 firms (in the USA)
• Fabricated in a single facility (in Taiwan)
• Using pure silicon wafers and special gasses (from Japan)
• Etched using machinery from a few companies (from Netherlands, Japan and US)
• Packaged and tested (in South East Asia)
• Final assembly (in China)
• The supply chain has a number of choke points, where only a handful of firms produce key software or components.
• No other part of the economy is so dependent on so few companies
• Chips from Taiwan provide 37% of the worlds computing power each year
• 2 Korean company’s produce 44% of the worlds memory chips
• Dutch company ASML builds 100% of the machines that make the worlds best chips
• Both China and US are fixated on controlling the future of computing
• Beijing won’t rule out invading Taiwan to re-unify it with the mainland

• When the transistor was invented, nobody cared. They were announced on page 46 of the New York Times in 1948
• Initially they were used for signal amplification. They were an alternative to the large unreliable vacuum tubes that were used in early computing
• The Cold War accelerated investment into semi-conductors.
• The military valued miniaturisation and ruggedness
• Fairchild R&D focused primarily on civilian use cases. As costs fell, the number of users would grow
• Fairchild slashed prices for consumers to start demand
• Moore: Predicted every year for the next decade Fairchild would double the number of components that could fit on a silicon chip
• By 1968 - the computer industry was buying as many chips as the military
• America was only a couple of years ahead of the Soviet’s in terms of quality and quantity of chips produced.
• Soviets invested heavily in spying, but having a chip in your hand doesn’t tell you how it was made.
• Soviet engineers were ordered to reverse engineer American chips.
• Copying worked for Nuclear Bombs - but no so for chips. Ability to mass produce at high quality was the real advantage
• USSR lacked the smaller companies that were needed to support production
• The US was innovating faster than the USSR could catch up (No other technology moved so quickly) . Copying last year’s design was a hopeless strategy
• Petrov - KGB employed 1000 people to steal technology and chips
• USSR had a ‘copy it’ strategy. Theft and replication. Trying to steal all of the machines involved in manufacturing too
• Lithography machines
• That strategy helped the US, they were always 5 years behind
• Precision weapons were a problem for USSR
• USSR predicted the US could destroy 98% of their Nukes if they struck firs. Hence why they continued to grow the number of nukes they possessed
• Gorbachov visited SV in the US, stopped the Cold War, in exchange for more chips and technology

• Sony comes from the latin word Sonos which means sound
• Sony knew consumer electronics would be the future → had a conscious plan to capture more value by innovating and owning the relationship with the customer
• Sony new they had to innovate. Produced the Walkman, and revolutionised the music industry (350m units worldwide)
• Throughout the 80s Japan’s chips were more reliable, higher quality and lower price.
• Japan government protected their industry and made it hard for the US to compete
• Japan chip makers made 50% of the worlds investment in chip manufacturing facilities
• By 1986 Japan was producing more chips, and producing 70% of the worlds lithography equipment
• Japans subsidies started to look like they were causing the US a national security issue
• Japan became the worlds second largest economy
• Sony thought they could challenge the US. Japan was creating engineers, the US lawyers. US was short-term thinking, Japan could think long term. Saw America as faltering, and wanted Japan to become the biggest economy in the world
• Asia began to dominate chip manufacturing. A lot of work was done by hand. Assembly workers in Japan and Hong Kong were twice as fast as those in the US. Wages were 10% of those in the US.
• The US innovated around Japan, offshoring to Taiwan and South Korea. Banned their control back
• Pentagon’s bets on chips paid off, helping them secure military superiority

• Taiwan was threatened by China’s development of nuclear weapons. They made a conscious decision to align and integrate with the US and used semi-conductors to do it

• Noyce and Moore left Fairchild and founded Intel. Their vision was to make transistors the worlds cheapest product -so the world could produce trillions and trillions of them
• Intel was originally a RAM business.
• There’s a tradeoff in computing between customised software and customised hardware. Intel bet on standardising the logic chip, and generalising software
• Intel 4004 was the first mass general logic chip

• Grove and Clayton Christianson knew each other. Grove was aware of the disruptive innovation theory.
• Grove was ruthless and decisive
• He got out of the memory market and bet big on micro-processors (expecting personal computers to be a big thing)
• The thought experiment that pushed him over the edge was to ask… What would a new CEO do?
• Intel sent managers to Japan. They were told to copy the best. Not to think, but to copy to replicate

• Samsung would become a chip superpower. The US seeing Japan as a threat supported South Korea. US firms gave up IP for manufacturing might. The logic was simple, my enemies enemy is my friend
• DARPA could place big bets on next generation technology
• Jacobs realised chips could help send more signals into the same radio bandwidth - founded Qualcomm

• The US showed the power of next generation weaponry for the first time in the first Gulf War ( (laser guided bombs, cruise missiles)
• Precision munitions were even more accurate than pilots claimed
• Proved decisive in the war
• Silicon, information and processing was the future of war

• Japan’s dominance was built on government loans, and over-investment. Banks kept lending, CEOs kept spending. They swamped supply and ended up competing with each other
• The PC revolution meant that the US and Korea won and Japan lost out

• Most of the money is made by the companies designing and producing the most advanced chips
• Cost of manufacturing equipment and fabrication rises with each generation of chips
• The Taiwan government theorised that chip design could be separated from chip production… creating a Gutenberg moment for chips.
• They put up 48% of the captital for TSMC and gave it tax breaks
• TSMC promised never to design chips, only to build them. So there was never a conflict of interest
• Fabless firms started to emerge in the US
• Chip fabrication is only profitable at huge scale, the more you produce the stronger your monopoly becomes
• Lithography is incredibly difficult
• ASML (a Phillips spin off) would dominate lithography
• ASML sources components from other companies, they rely a lot on suppliers and historically have bought those that weren’t run well.
• New technologies were needed to make smaller transistors.
• EUV lithography (Extreme Ultraviolet) uses shorter wavelengths of light.
• It took decades to develop and required many innovations (focusing optics, ultra pure diamonds, new lasers)
• A state of the art ASLM machine has hundreds of thousands of parts
• Each EUV machine costs $100m, downtime is expensive so extreme reliability is key
• The investments required in EUV were so big, that if it didn’t work there was no plan B.
• TSMC spent huge amounts of money on R&D and investing in fabrications
• It makes no sense for companies to spend billions on chip fabrication when they could build high margin software for the same investment
• Only TSMC, Samsung and Intel invested in the EUV machines
• Intel spent heavily on innovation ($10b a year through the 2010’s)
• They should have won the EUV era as they outspent TSMC
• TSMCs scale is it’s weapon, building 3x more chips allows it to evolve production processes faster
• Intel missed AI, chips optimised for AI are more power efficient
• Nvidia heard about PHDs at Stanford using GPUs for AI. They used that as one of the early signals to invest heavily in developing the AI GPU ecosystem.
• Nvidia has become Americas most valuable chip company.
• As the 2010s ended, only TSMC and Samsung could produce the latest generation of chips. The entire production of the world was taking place in Taiwan and South Korea

• Intel’s x86 architecture dominated the PC industry for years. It also dominated early cloud hardware.
• Intel and AMD had a duopoly on x86 chips
• Arm was founded to take advantage of RISC architecture that had potential to produce more efficient chips. Intel didn’t make the transition because it had too many customers on x86
• ARM failed to win PC or data centre market share but it did win mobile where efficiency was more important
• Apple asked Intel to manufacture a processor for the iPhone, they declined and they went to ARM. Viewed mobile as niche.
• Mobile is now 33% of microprocessor market.
• Intel has made more than 1/4 trillion dollars of profit

• The US lost its competitiveness in batteries, memory, and microprocessor manufacturing

• There was a time where if you didn’t have your own fab, you weren’t considered a real semi-conductor company
• Nvidia become the leader of graphics hardware which positioned them well for serving Machine Learning needs. Nvidia are fabless.
• Qualcomm’s innovations are about using processing power to transmit more data through the same spectrum, and its chip designs and patents are required to make a modern smartphone. Also fabless.

• Chip makers had to spend more, on more advanced manufacturing processes… every 2 years to keep up.
• Chip generations are measured in nano-meters. The smaller they get, the less power they draw
• 2010 transistors had to go 3D to get smaller
• Samsung and TSMC were the only 2 fabs. But Samsung having it’s own line of phones made it a competitor to many/ TSMC was a neutral player → not competing with anyone that they served. TSMC became the centre of the chip universe
• The original iPhone used components available to others. But Apple quickly started to design it’s own silicon. iPhones are now…
• Designed in California, assembled in China, the chips are made in Taiwan

• China felt nervous about it’s reliance on imported chips from western nations. It had only a small stake in the complicated chip supply chain (6%)
• China launched ‘Made in China 2025’ an initiative to import less chips by creating their own. China has spent billions subsidising investment in chip fabrication.
• China launched a giant $250bn semi-conductor fund
• Similar investments in the past crushed all competition in the solar panel industry
• China has partnered with firms like IBM and AMD
• China has sponsored intellectual property theft
• ‘No experience, no technology and no talent’ → billions have been wasted in China on unrealistic projects
• EUV machines took 30 years to develop. It’s hard to catch up. Just replicating the laser requires replicating 400k parts. Even with the design specifications, you still need PHDs to understand them and to build a whole industry to create the component parts
• China might lean into the new open source architecture Risc5
• The US in return has rejected many bids from China to buy crucial firms and technology on national security grounds. China has also tried to obscure it’s investment attempts.
• US has made some attempts at times to cut off Chinese firms from buying key bits of machinery developed in the US
• US blacklists have had huge impacts on China. So far China haven’t retaliated much.
• China’s state has ploughed $75b into Huawei, who build phones and crucial networking components for cell towers
• Huawei looked likely to dominate the 5G infrastructure of the future until governments around the world banned their involvement for fear of national security

• The future of war will be defined by computing power
• DARPA are now working on chips designs that are tamper free
• Taiwan is both where the majority of the worlds chips are built, but the most likely future battle ground of a war between China and the US
• Globalisation was starting to look dangerous. As it wasn’t global, production was all in Taiwan
• Weaponised interdependence. When countries need each other, and they can turn that into a weapon
• There are many choke points in the chip process. Components or software that can only be built by one or two firms

• Chip shortages were common after the pandemic. Supply issues were blamed but more semi-conductors were produced in 2021 than ever before. Demand growth was actually the problem.
• The car industry ordered badly and ‘just in time’ manufacturing compounded issues
• At one point ASML couldn’t source enough processors to build the machines that could build processors (that’s crazy)
• TSMC has become indispensable
• Peace in the Taiwan straight is incredibly important
• Both Washington and Beijing want more control over the TSMCs fabs
• Taiwans military strategy is to stay in the fight long enough for the US or Japan to come and protect them
• It would take years for the West to replicate the TSMC fabs

• Defence dollars propelled the semi-conductor industry forward in the early years
• Each day, the industry produces more semi-conductor chips than there are cells in the human body
• The pentagon has a $700b budget, but it can’t build it’s own chips
• Designing a single cutting edge chip costs more than building an aircraft carrier
• The chip made the modern world
• Many are declaring that Moore’s law is dead. Although, its survived longer than anyone thought
• There’s a path to 50x more density on chips, but there need to be a number of breakthroughs to enable that
• There might be a slow lane and fast lane divergence
• Fast lane: powerful customised chips
• Slow lane: general purpose chips
• Anyone can access fast lane chips (TPUs on Google Cloud)