Jeff Hirsh Read: February 2023
Inspiration: Wanted to understand what a “semiconductor” is and why the industry is so critical/heavily scrutinized
Summary
Written with the help of ChatGPT, below is a brief summary to understand what is covered in the book.
“Chip War”, published in 2022 by author and professor Chris Miller, is a comprehensive analysis of the ongoing competition between the United States and China to dominate the semiconductor industry, a vital component of modern technology. Miller explores the historical, economic, and political factors that have contributed to the current state of affairs and examines the implications of this rivalry for both nations and the world. He argues that the semiconductor industry is critical to national security and economic power, and the stakes are high in this geopolitical competition. The book also delves into the strategies and tactics employed by both countries to gain an advantage in this high-stakes game, such as investing in research and development, creating domestic supply chains, and imposing trade restrictions. Overall, “Chip War” provides a fascinating and in-depth analysis of one of the most critical issues facing the global technology industry today.
Unedited Notes
Direct from my original book log, below are my unedited notes (abbreviations and misspellings included) to show how I take notes as I read.
A chip is a grid of millions/billions of transistors which are tiny electrical switches that flip on/off to process and remember 1s and 0s via electrical currents, TSMC is the Taiwanese company that makes the chips behind all iphones—11.8bn tiny transistors in A14 processor made of silicon, first chip in 1961 had 4 transistors—Fairchild Semiconductor in SF founded by Gordon Moore of Moore’s Law (computing power grow exponentially), Moore founded Intel as well, what goes into all devices touches many countries globally but tiny number of firms, key firms in Taiwan and Japan at high vulnerability given conflict with china and earthquakes, however cheap labor and deliberate decisions made Asian base not accidental, semiconductors unique as enable control of current—first transistor (i.e. contains switch) invented 1948 by Bell Labs—William Shockley was british scientist pioneer 1947 who spark innovation Bell used, Jack Kilby 1958 invented first integrated circuit (“chip”) with multiple transistors on single slab of silicon or germanium, Fairchild founders used to be under Shockley then left to found Fairchild (“traitorous eight”—Moore, Kleiner, Noyce, others), USSR work to copy work of Texas Instruments (Noyce, Moore, etc) but given Moore’s Law essentially mean always behind, Akio Morita was Japanese sake empire heir who founded Sony licensing US transistor tech and made radios—Japan post ww2 just focused on increasing income of citizens, Taiwan and Texas Instr rel’ship begin post Vietnam defeat where Taiwan weary of communism and newly nuclear China—see integration with US as good and TI Morris Chang and others had Taiwan ties—set up plant 1969, spread to Singapore, HK, Malaysia—see semis as good for econ and politics with US, Noyce and Moore leave Fairchild 1968 to found Intel—Dyanmic Random Access Memory chip was first product (core of computer memory today, leverage capacitors and transistors to dynamically/repeatedly recharge capacitors via transistor reliant on silicon carving not magnetism rings and wires, then Intel launch first microprocessor which was micro-programmable and mass producable as standard design with manipulatable software by use case, by early 80s Japan and Sony surge ahead in consumer electronics (Walkman), US military was biggest consumer of semis in 60s/70s then turn focus on consumer, Hitachi and Toshiba as well spend billions in 80s—advantaged vs US as rates in Japan lower and banks flush with deposits vs US 20%+ rates, litigation frequent as well amongst firms, by 1986 Japan surpass US in chips produced and lithography—US ceo’s to gvt to lobby for more support “new oil”, by mid-80s Japan declared self ahead of US as US firms flounder—GCA lithography fall as Nikon dominate, Japan see self on offensive, Jack Simplot one of people who revived US chips—background as potato salesman to half mcdonalds—backed Micron which made DRAMs with $1m when outlook worst, focus on cost cutting ruthlessly and outspoken about Japan practices, based in Idaho and eventually buy TI’s DRAM business, Intel struggling in DRAM and pivot to small biz of PCs with IBM who was working with Bill Gates 1980–enter microprocessor market and abandon DRAM to transform Intel, big risk that succeeded, Samsung founded 1938 S Korea by Lee Byung-Chul selling dried fish and grain then to sugar, fertilizer, construction, banking in 60s/70s, 1982 visited HP and decide to copy with $100m investment in ’83–Samsung pivot to semis and partner with SK gvt and US to fend off Japan as well “enemy’s enemy is my friend”, in the US gvt-funded DARPA program help fund massive innovations in 70/80s—key to microprocessor success, spark Qualcomm founding too, TI built circuitry in Paveway laser-guided bombs used in Vietnam then in Gulf War in Iraq 1991 va Saddam Hussein—decisive in quick war for US, show future of precision warfare built on silicon, Japan 1990 financial crisis/yen volatility show overinvestment vulnerability, Sony slump, stocks collapse in Japan and missed PC boom so by ’93 US retook first place, Soviets/KGB also see well behind (lack manuf capacity and copying US is too slow), 1985 Morris Chang take role to lead Taiwan semis industry beyond just having plants for firms like TI, Chang had idea for Foundry concept—manuf chips designed by customers instead of chips designed in-house by producer, had blank check from Taiwan gvt, TSMC begin 1987 (gvt had 48% formal stake but essentially control), TSMC was a fabless chip design firm—work with customers not compete, not design chips only build, 1987 Ren Zhengfei found Huawei in Taiwan—telecom equipment trading, China behind like Soviets given Mao’s rule and craziness, Deng Xiaoping in 80s promise to modernize, fabs=fabrication facilities, Chang open SMIC in China in 2000 to mimic what he did with TSMC—really wanted to open in China, vast gvt support, Singapore also create own foundry Chartered Semi, into 2000s more precise lithography required to shoot light at photoresist chemicals and carves shapes into silicon to make smaller chips/transistors—turn to either beams of electroncs or extreme UV light or x-rays (lithography wars to figure out next type of beam to shoot at silicon wafers), ASML in Netherlands lead lithography tech in mid 90s (US partner with them, ties to TSMC as helped fund TSMC—political concerns not really at forefront with ASML having access to all US EUV lithography tools), Intel missed wave on mobile phones b/c so focused on PC profits and margin—Apple asked Intel to make chips for phones but turned down for fear of margin hit, Samsung made early iphone chips, Intel rely on x86 processor PC moat, mid 00s Andy Grove concerned about US positioning, politicans not want to stifle globalization—say US doing fine and okay other countries have roles, “run faster” principle for US to stay ahead but many neglect US actually slowing and others catching up, 3 types of chips: logic (processors running smartphones and comps), memory (DRAM and NAND, micron one of 3 controlling firms in DRAM and NAND but fabrication is mainly in Asia despite US firm), Jerry Sanders found AMD which focus on logic chips while others outsource fab to TSMC type foundries, Intel also do logic chips still, fabless chip firms now taking off (unlike AMD and Intel)—rely on TSMC, design semis in house and outsource manuf, graphics processors dominated by Nvidia founded 1993–design chips and software around them, Nvidia chips manuf by TSMC largely, Nvidia is fabless, Qualcomm similarly designs integral chips for telecom/cell phones but fabless—rely on samsung/TSMC, fabs are super expensive—why Qualcomm and Nvidia not do so focus elsewhere, 2012 AMD spin off GlobalFoundries, for smartphones really only TSMC (apple) and samsung but samsung has conflict, then GlobalFoundries catching up but behind, apple rely on taiwan, Chang stepped down TSMC pre-GFC then back in as dislike path 2012, intel miss pivot to AI as focus on PC processors (slower upgrade cycle), intel do CPUs (central proccessing unit—brain of comp or data center, general purpose, do calcs one after another), CPUs slow for AI and costly, Nvidia GPUs (graphics) run multiple iterations of same calc at once (parallel processing) and best for AI, Nvidia bet on AI and worked—chips easy to program and can train models cheaply as process data simultaneously (data centers now running on GPUs more so intel lose position), intel miss AI, flopped on foundry attempt, integrated model of intel hurt—too stubborn, TSMC and Samsung now lead processors (taiwan amd korea), Made In China 2025 plan for semis significant bc spend $260bn/yr to foreigners such as neighbors and US, in 2016 AMD helped tech transfer to China in semis as strapped for cash and sell stake in facilities—handed crown jewels though AMD try to minimize, Arm is british and designs chip architecture—tech transfer to china as well via subsidiary, big semi companies (IBM included) willing to share non-core tech for $/access to China but on whole creates broad risk with China as tech transfer), Tsinghua Unigroip run by Chinese chip billionaire Zhao Weiguo (close to china gvt) consolidate globally—try to acquire Micron even, Taiwan firms, etc (frame as business/market oriented but not logical), Huawei not spawn from chinese gvt—true business story that aligned with China interests (R&D machine, low costs, telecom infra giant with Nokia and Ericcson), did have stealing instances as grow, best in class 5G equipment though which in ’17 as contracts handed out Huawei beating Nokia and Ericcson—design chips in house and use TSMC as fab (big in mobile market too), china still depend on US design (nvidia, intel, amd) and Taiwan fabrication, military power is battle over spectrum space (driven by data, algos, processing power), see radar jamming more which challenges AI/autonomous weapons and drones, trump admin pivot to seeing semi supply chain as way to punish china—not just run faster principle anymore, cut off choke points that crush Huawei for example as US made tools integral, China has 10s of billions in gvt funds to insert self into supply chain—not profit focused, remain technologically behind but finding slots to enter, 2021 Pat Gelsinger (Grove mentor) take over Intel—plan to build foundry, deal with ASML for first gen EUV tech (US best hope), peace in Taiwan influences trillions of dollars but proximity for china poses threat, big tech now design own chips but flow is all connected to taiwan with massive investment and implications as go forward
