Part IThe Story
The Man Who Did Not Take the Call
In the autumn of 2009, reporters from Forbes Asia telephoned MediaTek's headquarters in Hsinchu several times, hoping to speak with the company's founder and chairman. Tsai Ming-kai, then fifty-nine, did not come to the phone. On its own this was unremarkable. Taiwanese chairmen of his generation are not known for chattiness, and the chip-design business, which sells to manufacturers and never to the public, has little use for a talkative founder. What made the silence worth noting was the product. The silicon his engineers designed was expected to go into roughly 300 million mobile phones that year. In 2004 the figure had been 3 million. The man whose chips were about to sit inside a few hundred million handsets, in villages and bazaars and migrant dormitories across Asia, the Middle East, and South America, declined to take a call.
The magazine thought it knew the reason. Somewhere along the way Tsai had acquired a title he had not asked for: father of China's shanzhai handsets. Shanzhai means "mountain stronghold," and by extension "bandit." By the late 2000s it had come to mean the unbranded, brand-mimicking phones that poured out of Shenzhen by the tens of millions, with fake logos, three SIM slots, and speakers loud enough to rattle a bus window. Analysts estimated that 10 to 20 percent of MediaTek's chipsets still ended up in outright pirated devices. Tsai, Forbes noted, "doesn't embrace the moniker." His chief financial officer, Yu Mingto, took the call instead, and he was, in the magazine's phrase, "less abashed."
One year earlier, in November 2008, Nokia had run a newspaper advertisement in India that read like a warning from an older order. Its originals, the ad said, "last forever." Duplicates "come and go." The ad went on to argue that white-box phones without traceable identification codes posed a threat to national security, since they might be of more use to terrorists. In plain terms, the world's largest phone maker was telling a billion people that the cheap phone in their pocket was dangerous. Somewhere inside that cheap phone, more often than not, was a chip designed in Hsinchu.
Tsai's given name is written 明介. The first character means bright. The second, jie, carries an older sense of being situated between: it appears in the words for introduce, for intermediary, for medium. It would be too neat to make much of this, and nobody who knows Tsai has gone on record doing so. Still, the story of his company keeps arranging itself around in-betweens. MediaTek sits between the foundry and the brand, between the copy and the original, and between a Taiwan that makes chips and a China that buys them. It also sits between a founder who would not pick up the phone and the professional manager he eventually hired to speak for him.
The public record of what Tsai Ming-kai has actually said is thin, and it should be admitted early that this is a portrait assembled largely from what he built. He gives the odd speech, consults on the odd white paper, and accepts the odd medal. Mostly the chips speak. Typing his name into a search engine in English returns, almost as often as the chairman, two other men. One is the chemist Ming-Daw Tsai of Academia Sinica, who has spent a career studying how DNA polymerases "catalyze replication and repair with contrasting fidelity," which is to say how biological copying machines decide how many errors to tolerate. The other is the filmmaker Tsai Ming-liang, famous for static long takes in which almost nothing moves. The first namesake studies the productive uses of imperfect copying. The second has built an art out of standing still. Tsai Ming-kai's career amounts to a refusal of the second and a forty-year industrial application of the first.
By the Numbers
MediaTek, Measured
NT$530.59BMediaTek revenue in 2024, up 22.4% year over year
3M → 300MHandset chipset volume, 2004 versus the 2009 forecast
31%Share of global smartphone chipsets in Q3 2020, overtaking Qualcomm
2B+Devices powered by MediaTek silicon each year
$3.8BPrice of the MStar acquisition, announced June 22, 2012
22,546Employees as of February 28, 2025
A Satellite Over Pingtung
He was born in April 1950 in Nanzhou Township, Pingtung County, at the island's agricultural southern end. By the accounts that have been assembled of his childhood, it was a household that valued schooling more than land. His father taught primary school, and his siblings largely became teachers too. During school vacations the boy did farm work. There was no television, and telephones were rare. Newspapers and the radio carried the outside world into the house.
In October 1957, when he was seven, the radio brought news of Sputnik. The Soviet satellite was a beeping aluminum sphere that crossed the sky over every rice paddy on earth, Pingtung's included, and nobody in Nanzhou could have done anything about it. Tsai has recalled that the launch moved him enough that he resolved, while still in primary school, to become a scientist or an engineer. The IEEE, writing his biography decades later, was more specific: he "initially planned to be a rocket scientist." It is worth holding onto the image of a farm boy in the south of Taiwan, in a house without a television, learning from a radio that a machine was circling overhead, because the story comes back to it.
He entered National Taiwan University in 1967, at a time when the island's best students were sorted by test score into a handful of prestige departments. His first was chemical engineering. That was the safe choice, the dutiful son's choice. Chemical engineers could go to Formosa Plastics, the largest private enterprise in Taiwan, and a saying of the era held that once you got into Formosa Plastics you would never have to worry about eating and drinking for the rest of your life. The phrase captures a whole generation's understanding of security: a single employer, a lifetime meal ticket, a career shaped like a closed loop.
Two things pulled him out of that loop. The first was the integrated circuit, invented at Texas Instruments in 1958 and still, a decade later, a new and somewhat mystical object for a student in Taipei. The second was an older alumnus from his secondary school in Kaohsiung, who told him that electrical engineering's future was, in the Chinese idiom, hai kuo tian kong: wide as the sea, open as the sky. Tsai transferred to electrical engineering. Later he went to the University of Cincinnati for a master's degree in the same field, part of the great postwar outflow of Taiwanese engineers who studied in the American Midwest and South and then, unlike earlier émigrés, came home.
What he came home to was an industry in its larval stage. Taiwan's semiconductor founding myth is usually dated to 1974 and a breakfast at a soy milk shop called Xiao Xinxin, where officials and engineers sketched a plan to acquire integrated-circuit technology for the island. The plan produced the Industrial Technology Research Institute's electronics programs. Those programs produced United Microelectronics Corporation, Taiwan's first semiconductor company, and much later, after a long detour through Morris Chang's imagination, TSMC. Tsai's early career fits this lineage exactly. He worked as an IC design manager at ITRI, then joined UMC in 1983.
The arrangement deserves a moment, because it is the template for everything Tsai later did. Taiwan in the 1970s had no semiconductor industry, no domestic chip customers of consequence, no venture capital, and no Silicon Valley density of talent. What it had was a state willing to buy technology, a pool of engineers trained abroad, and a willingness to begin as somebody else's licensee. It started, in other words, as a copy. The question was whether the copy could become something else.
The Products Group, or How to Breed a Litter
At UMC, Tsai rose to run the products business group, the part of the company that designed chips rather than merely manufactured them. The IEEE citation lists the categories in which he "pioneered Taiwan's chip designs" there: CPUs, mass storage, communications, multimedia. These are wildly different product families, and the list reads less like a strategy than like a man trying everything to see what would take. He also brought in modern electronic design automation tools, the software that lets engineers lay out millions of transistors without drawing them by hand. The point of the tools was productivity, and the point of productivity was that a small, young, underfunded team in Hsinchu could design chips at something approaching the pace of teams in California.
He also trained people. The IEEE puts it carefully: Tsai "mentored dozens of design engineers who went on to become key leaders in today's chip design industry." In Taiwan he is called, without much irony, the godfather of IC design, IC sheji jiaofu. Forbes in 2009 used the gentler "father of Taiwan's chip design." Either way, the paternal metaphor is earned less by MediaTek's market capitalization than by what happened when UMC broke itself apart.
UMC's leadership decided in the mid-1990s to separate manufacturing from design. Foundries make chips for other companies, and those customers are reluctant to share designs with a foundry that also competes with them. TSMC had demonstrated that a pure-play foundry could win the trust of the world's fabless companies, and UMC, competing for the same customers, needed to look clean. So the products division was carved into independent fabless companies. Faraday was one. Novatek was another. The unit that designed chipsets for home entertainment products, CD-ROM and optical-storage controllers above all, was incorporated on May 28, 1997, as MediaTek. Tsai became its chairman. According to the IEEE, many of the spinoffs, "together with numerous rising fabless design startup companies," were chaired or directed by him at one time or another.
The founding of MediaTek, then, was not a garage story. It was a divestiture and a structural necessity, a branch cut so the tree could grow. Tsai was forty-seven, middle-aged and mid-career, and his company was born with a parent, a customer base, a product line, and a building in the Hsinchu Science Park. It went public on the Taiwan Stock Exchange on July 23, 2001, under the code 2454. That same year it established the MediaTek Foundation to promote technology education, which suggests a man who already understood that his real constraint would be the supply of engineers, not the supply of capital.
The product that carried the company through its first years was unglamorous. Personal computers in Asia were discovering multimedia, and every household seemed to be buying a CD-ROM drive. MediaTek's chipsets for optical drives were cheaper, more integrated, and easier for manufacturers to build into a product than the alternatives. The company dominated. It then moved into DVD players, using the same logic: take a function scattered across several chips, collapse it into one, and hand the manufacturer a design that worked. Photographs of disassembled DVD players from the era show a MediaTek MT1389 at the center of the board like a small black stamp, the whole machine organized around it.
A year after the IPO, in 2002, Taiwan launched a National System-on-Chip Project, a government program to push the island's designers up the value chain. In a 2023 interview with the trade paper DigiTimes, Tsai credited it with helping Taiwanese IC design become a trillion-New-Taiwan-dollar industry. It is a telling thing for him to emphasize. He is not, on the record, a libertarian founder myth. He is a man who believes that small islands build industries on purpose.
Leaving the Belt on the Canvas
There is a theory in Taiwan's chip industry, attributed to Tsai and repeated often enough that it has become something like scripture at MediaTek, that the chip-design business is like boxing. You can knock out every opponent and hold the championship belt over your head, and the next challenger can still put you flat on the canvas. Winning one generation means nothing. Survival requires winning three. A 2014 Nikkei profile of him by Kazunari Yamashita carried the headline "Tsai: the boxer that won't go down."
The metaphor is more interesting than it first appears, because it is a theory about decay, not about combat. In boxing, the champion is the person everyone is training to beat. In semiconductors, the product that won you the title is depreciating from the day it ships. The optical-drive business was generating MediaTek's best numbers when, by the account in Taiwan's Storm Media, Tsai began redirecting engineering resources away from it. Solid-state storage would eventually make optical drives obsolete, and several competitors who stayed in the ring too long were wiped out. MediaTek was already elsewhere.
Elsewhere was the telephone. The company launched a division to design products for mobile devices in 2004 and shipped its first wireless chipset to China that year, for 3 million units. Forbes would later identify this as the company's pivot point. In 2004 the global handset market belonged to vertically integrated giants such as Nokia, Motorola, and Samsung, along with a few others. They had their own engineering armies and treated chips as components to be specified, sourced, and integrated through years of internal development. Small manufacturers could not enter, because building a phone required precisely those armies.
MediaTek's MT62xx chips, the MT6225 and MT6235 and their siblings, changed the arithmetic. The company did not just sell a baseband processor. It sold the processor with a reference design, software, and what Wikipedia's historians call "extensive system engineering assistance." Inside the company and later across the industry, this was called the turnkey solution. A small Shenzhen workshop with a few dozen employees and an industrial designer could buy MediaTek's platform, wrap a plastic shell around it, and have a working phone on the market in weeks. The engineering army came in the box.
Seven years after entering the business, MediaTek was taking orders for more than 500 million mobile system-on-chip units per year.
The boxing theory and the turnkey model belong together. Both rest on the conviction that the advantage is never in the thing itself but in the rate at which you can make the next thing. A champion who stands still is a target. A chip that stands still is a commodity. Tsai's answer to both problems was velocity, his own and, more radically, his customers'.
Bandits
By 2009, MediaTek's Chinese customers were releasing around 3,000 phone models a year, which Yu Mingto told Forbes was ten times as many as the world's top five established brands combined. Yu compared them to a fast-fashion house: the Chinese manufacturers, he said, could capture a highly fragmented market "as nimble as Zara in the fashion industry." The image is apt and slightly sinister. Zara does not invent the runway. It watches the runway and ships something close to it to stores within weeks, at a price the runway's customers would not recognize.
MediaTek had grabbed half the Chinese chipset market for feature phones. Forty percent of the Chinese-made handsets using its chips were re-exported to the Middle East, India, and South America. The ten largest Chinese-branded phone makers, among them Huawei, ZTE, and Beijing Tianyu, had become clients. MediaTek had customers on both sides of the line between legitimate and bandit, and the line was moving.
— Yu Mingto, MediaTek chief financial officer, to Forbes Asia, November 2009The majority of our 100-odd clients are no longer copycats; instead, they've created their own style. It's not right to remain a copycat. For any company to be able to sustain its business, innovation is a must, even if it started out by replicating others' work.
Tsai's own version was shorter. His customers, he had said, might be shanzhai today, but they would be "mainstream tomorrow." In Taiwan's tech circles the line was later compressed into a motto: today's copycat, tomorrow's mainstream. The aphorism is not a defense of piracy, though it was frequently heard as one. It is a theory of industrial development. Emerging markets absorb technology from the bottom up. Volume creates capability, and capability eventually creates quality. The firm that serves the copyist in his first year will be his supplier in his tenth, when he has a brand, a design studio, and a listing on the Hong Kong exchange.
Anyone who knew Taiwan's history would have recognized the argument immediately. It was Taiwan's own story, retold with a Shenzhen accent. Fifty years earlier, the island's factories had been the cheap copyists, assembling transistor radios and calculators under foreign brands, absorbing know-how one licensing deal at a time. Its semiconductor industry began with imported technology after a breakfast at a soy milk shop. The bandit phones were, in a sense, Taiwan's younger siblings, and Tsai was explaining to the world what had happened to the elder.
He seemed to believe the macroeconomics of it with real conviction. In 2013, when World Finance named him to its annual list of business leaders, it recorded him as saying that the financial crisis had "confirmed one thing… that emerging countries were almost not affected. Their economic development will continue and the opportunities there will be bigger." The sentence was spoken from experience. MediaTek's net sales were NT$68.0 billion in 2008, the year Lehman collapsed, and NT$77.3 billion in 2009. Where the rich world was shrinking, MediaTek's customers were still buying their first phones.
Tsai never fully shed the bandit label. Profiles in Chinese still call him the king of the bandit phones, a phrase that manages to sound both pejorative and admiring. Nor did he ever publicly fight it. He simply did not answer the phone.
Buying the Pieces
Velocity has a limit, which is the set of things you already know how to build. Here the record shows a second Tsai, less an aphorist than a purchaser.
In 2005 MediaTek acquired Inprocomm, a wireless design house producing 802.11a, b, and a/g Wi-Fi chips. On September 10, 2007, it announced that it would buy the cellular radio and baseband chipset divisions of Analog Devices for $350 million. The deal closed on January 11, 2008, and it gave a company that had grown up on low-cost 2G handsets an engineering foothold in the 3G technologies that the incumbents guarded jealously. On May 5, 2011, it acquired Ralink Technology, a Taiwanese rival in Wi-Fi, DSL, and Ethernet connectivity. On April 11, 2012, it bought Coresonic, a digital signal processing company in Linköping, Sweden, a university town on the Stångån river about as far from Pingtung as an engineer could get.
Then came the large one. On June 22, 2012, MediaTek announced it would acquire MStar Semiconductor, a Taiwanese rival that held a strong position in digital television chips, for $3.8 billion. The structure was cautious. MediaTek took an initial 48 percent stake, with an option to buy the rest. Antitrust regulators in China and South Korea delayed the full merger, which was not finalized until February 1, 2014. In 2015 MediaTek announced the purchase of Richtek, a Hsinchu maker of analog and power-management chips, and the deal closed in the second quarter of 2016.
The acquisitions read as a list, but they track a single anxiety. A phone or a television is a bundle of functions: compute, cellular radio, Wi-Fi, Bluetooth, positioning, display, power. A company that sells turnkey solutions has to own as many of those functions as possible, because every function it lacks is a seam where a competitor can wedge in. The ADI purchase bought the cellular radio. Ralink bought Wi-Fi. Coresonic bought signal-processing intellectual property. MStar bought the living room. Richtek bought the power supply. One by one, Tsai was buying the missing parts of a single chip.
The timing of the MStar deal is revealing, because it was struck from a position of weakness. MediaTek's net sales fell from NT$71.99 billion in 2010 to NT$53.84 billion in 2011, a drop of roughly a quarter, and operating income collapsed from NT$17.27 billion to NT$4.84 billion. The feature-phone business that had carried the company through the financial crisis was being eaten by smartphones, and MediaTek was late to smartphones. The boxer had been knocked down. Within a year of the trough he had signed a $3.8 billion acquisition of a domestic competitor. The Financial Times had given MediaTek its award for boldness in business in 2011, the trough year itself. Whatever the judges had in mind, the timing turned out to be uncannily correct.
Octa-Core and the Next Five Billion
The recovery was fast and loud. Net sales were NT$99.26 billion in 2012, NT$136.06 billion in 2013, and NT$213.06 billion in 2014. Part of the 2014 jump was MStar's revenue arriving on the books, but most of the earlier growth came from smartphones and tablets. MediaTek was doing to Android what it had done to the feature phone: it gave the Shenzhen workshops a box with a smartphone inside.
The products of this era were technically aggressive and sometimes theatrical. In July 2013 the company announced the MT8135 tablet chip, the industry's first implementation of ARM's big.LITTLE architecture, which pairs high-performance cores with efficient ones. A variant went into Amazon's Kindle Fire HD, a reminder that the bandit's supplier was now also the supplier to Seattle. On November 20, 2013, it launched the MT6592, the first system-on-chip whose eight CPU cores could all run at once. Competitors' eight-core chips could only use a subset at a time. MediaTek registered a trademark, "True Octa-Core," to make the distinction impossible to miss. In May 2015 came the Helio X20, the industry's first tri-cluster CPU and the first with ten cores. HiSilicon, Qualcomm, and Samsung would all adopt the tri-cluster idea within four years.
At Mobile World Congress in 2014, the company unveiled a brand, "Everyday Genius," and coined a market category, the "super-mid market." The phrase was ungainly but the idea was not. Between the $600 flagship and the $40 bandit lay a vast middle where most of the world's next buyers would live, and MediaTek intended to own it. By November of that year, more than 1,500 phone models had shipped with MediaTek chips, accounting for 700 million units, and first-half revenue of $5.3 billion nearly matched all of 2013.
That February, at the IEEE International Solid-State Circuits Conference in San Francisco, the semiconductor industry's most rigorous technical venue, MediaTek had eight papers accepted, a high for the industry. Tsai was invited to give a keynote. He said something that would become the closest thing he has to a public creed. Only about 2 billion people in the world, he observed, had the means to enjoy modern technology. Five billion had not yet used it. The gap between those numbers, he argued, was the direction fabless companies should work toward. The IEEE later rendered his ambition in a phrase: he wished to "connect with the next 5 billion people."
The phrase is less grandiose than it sounds, because it is an inventory of customers. Five billion is a market size. But it is also a statement about who counts. Nokia's newspaper ad in India, five years earlier, had implied that the owner of a cheap duplicate was a security risk. Tsai's keynote implied the opposite: that owner was the whole point.
The same year exposed a wound. MediaTek began shipping chips with an integrated 4G LTE modem in volume only in the second half of 2014, later than Qualcomm. Until then, a MediaTek-based LTE phone needed a separate baseband chip, which added cost and broke the turnkey promise. Alcatel One Touch and ZTE, two of its larger customers, moved to Qualcomm's Snapdragon 400 and 410 platforms. Sales went flat. Net revenue was NT$213.26 billion in 2015, essentially unchanged from 2014, and operating income nearly halved, from NT$47.24 billion to NT$25.91 billion. Revenue reached NT$275.51 billion in 2016 and then fell back to NT$238.22 billion in 2017, when operating income dropped to NT$9.82 billion, the second-worst result of the decade. In the language of the theory, MediaTek had won the feature-phone belt and the 3G smartphone belt and was losing the third generation. Analysts called it a fast follower, and the label was not a compliment.
The Other Tsai
On the morning of June 11, 2009, the board of TSMC met to consider a proposal put forward by the company's founder, Morris Chang. It took less than ten minutes. The chief executive, Rick Tsai, would step down. The shareholders' meeting had concluded the day before. Tsai was informed, and he was gone the next day.
Rick Tsai is no relation to Tsai Ming-kai, although the coincidence of surnames would later generate an industry nickname and at least one headline that confused the two. He was born in April 1951, a year after Ming-kai, studied physics at National Taiwan University, and earned a doctorate in materials science from Cornell. He then spent eight years at Hewlett-Packard in Fort Collins, Colorado, working on 1-micron CMOS process integration, yield improvement, and reliability. These are the least romantic and most essential problems in semiconductor manufacturing. He joined TSMC in 1989 as deputy plant manager of what was then its only fab. He ran Fab 2, Module A; designed, built, and ramped Fab 3; ran Fabs 3 and 4; and served two terms as executive vice president of operations, overseeing six fabs at once. Between those terms there was a short stint as president of Vanguard International Semiconductor. He became Chang's successor as CEO in 2005.
The industry called him the iron-fisted CEO. During the 2008 financial crisis he pushed an aggressive layoff plan, and that became his undoing. The father of a veteran employee wrote to Morris Chang pleading on his son's behalf. An investigation found that Rick Tsai's account of the matter did not match the facts. Trust was the foundation of Chang's management philosophy, and once it cracked there was no repairing it. In the ten-minute meeting, Chang took the company back.
What happened next shows a certain kind of Taiwanese institutional patience. Rick Tsai was not fired from the TSMC system. He was moved sideways, first to head a new business organization and then to run the solid-state lighting and solar subsidiaries TSMC had spun off. He did not formally leave until the end of 2013. In 2014 he became chairman of Chunghwa Telecom, Taiwan's largest telecommunications carrier, where he drove the island's early rollout of 4G LTE, the technology that had wounded MediaTek. He stepped down at the end of 2016.
On March 22, 2017, MediaTek announced that it had recruited him as co-chief executive and board member, effective that July. Nikkei Asia reported that the chip designer hoped the appointment would help it become a global player. The industry was skeptical. An executive known for running fabs with an iron fist, removed from the top of TSMC after a breach of trust, was coming to lead a company whose identity rested on its engineers' culture. When he arrived, MediaTek's stock hovered around NT$260. Looking back from January 2022, Nikkei's chief tech correspondent Cheng Ting-fang wrote that industry experts and investors had all thought Rick Tsai "had been handed a poison chalice."
The press called the arrangement the "dual Tsai" leadership. It is the most interesting decision in Tsai Ming-kai's career, partly because it is so uncharacteristic of founders. He was sixty-seven. The company he had founded was stalled. The obvious founder moves were to reassert personal control, fire the management, and give speeches about returning to the company's roots. Tsai instead hired a man who was in almost every respect his opposite: a manufacturing physicist, not a design engineer; a TSMC lifer, not a UMC spinoff; a public figure with a public fall. He then gradually stepped aside. Rick Tsai became vice chairman and sole CEO. Joe Chen became president and chief operating officer. Tsai Ming-kai remained chairman of a nine-member board with five independent directors.
There is a story in the Chinese-language press about the chalice, and the two men sometimes tell versions of it, but the plainest evidence is in the awards. Years apart, both Tsais received the Global Semiconductor Alliance's highest honor, the Dr. Morris Chang Exemplary Leadership Award. Rick Tsai's came on December 7, 2023. In accepting it, he used a phrase that could have come from his chairman's ISSCC keynote nearly a decade earlier.
— Rick Tsai, on receiving the GSA's Dr. Morris Chang Exemplary Leadership Award, October 2023My vision has always been to make great technology more accessible, and it's been remarkable to see this come to fruition over the course of my career.
The award was named for the man who had removed Rick Tsai from TSMC in ten minutes. He accepted it at MediaTek, thanking, among others, "TSMC, as well as Taiwan's fabless industry." The island is small, and grudges, it seems, are expensive.
Not Repeating Four-G
The dual-Tsai leadership made one decision that mattered more than any other: they would not repeat the hesitation of the 4G years. For 5G, they bet on two tracks at once. They pursued the Sub-6 GHz band to capture the Chinese market, while also building millimeter-wave capability aimed at the United States and Japan. On November 26, 2019, MediaTek announced the Dimensity 1000, its 5G system-on-chip and the world's first mobile SoC to support the AV1 video codec. The day before, it had announced a partnership with Intel to bring 5G to personal computers. This was the boxer coming off the ropes.
Then came the year everything happened at once. In the third quarter of 2020, according to Counterpoint Research, MediaTek overtook Qualcomm as the world's largest vendor of smartphone chipsets, with 31 percent of the market. The growth came from India and Latin America, the countries of the Nokia advertisement and the re-exported Shenzhen handsets of a decade earlier. For the full year it shipped 351.8 million smartphone chipsets. In July 2020, three years after Rick Tsai's arrival, MediaTek's market value crossed NT$1 trillion for the first time, making it only the third company on the Taiwan exchange to reach that scale. The stock that had hovered around NT$260 when the poison chalice was handed over had multiplied several times. Net sales in 2021 were NT$493.42 billion, and operating income was NT$108.04 billion, more than eleven times the 2017 figure.
The same year also produced the company's least flattering episode. On April 8, 2020, the technical site AnandTech, through its mobile editor Andrei Frumusanu, published an investigation into what MediaTek's firmware called "Sports Mode." The feature detected popular benchmarking applications from a whitelist and pushed the chip into maximum performance whenever one was running, including benchmarks meant to measure everyday user experience. MediaTek posted a response the same day, titled "Why MediaTek Stands Behind Our Benchmarking Practices." It argued that the practice was standard in the industry and that device makers could choose whether to enable it. On April 14, Qualcomm said publicly that it did not use whitelisting because it considered whitelisting cheating. On April 16, Oppo said it had tried to remove Sports Mode but had not realized the feature was still cached, and it pushed a firmware update. UL, which publishes the 3DMark and PCMark benchmarks, delisted several MediaTek Helio chips from its rankings.
It is fair to note that AnandTech had made similar criticisms of Samsung's Exynos and Huawei's HiSilicon chips. It is also fair to note what the episode reveals about a company whose identity was formed by the gap between the copy and the original. A company that once trademarked "True Octa-Core" to separate its real eight cores from its rivals' nominal ones had been caught making its own numbers look better than they were. The Sports Mode episode was a small thing next to 351.8 million chipsets. It was also exactly the kind of small thing that keeps the word shanzhai attached to a company long after the bandits have gone mainstream.
The flagship climb continued regardless. The Dimensity 9000 gave way to the 9200, the 9300, announced November 6, 2023, and billed as ready for generative AI, then the 9400, and on September 22, 2025, the Dimensity 9500, the first chip built on Arm's C1 cores. The chips moved from 4-nanometer to TSMC's third-generation 3-nanometer process. In 2024 MediaTek's smartphone revenue passed $10 billion for the first time. That May, Nvidia had announced a collaboration to use MediaTek's Dimensity chips in automotive infotainment systems. Reuters reported in June 2024 that MediaTek was designing an Arm-based PC chip for Microsoft's Windows. At Computex 2026, Nvidia launched its RTX Spark line, co-developed with MediaTek.
Then, in April 2026, at Google's Cloud Next conference, Google announced that its eighth-generation tensor processing unit, the chip behind some of the most advanced AI systems in the world, would for the first time be split into two dedicated chips, one for training and one for inference. Broadcom designed the training chip. MediaTek designed the inference chip.
Two months earlier, Rick Tsai had delivered a plenary address at ISSCC, the conference where Tsai Ming-kai had talked about the next five billion in 2014. It was compiled under a title that would have baffled the optical-drive engineers of 1997: what semiconductors should sell is no longer chips. The argument was that the unit of value had moved from the die to the rack, to entire systems of compute. This was the turnkey solution again, at data-center scale. In 2004 MediaTek sold Shenzhen workshops everything they needed to build a phone. In 2026 it was selling the largest technology companies in the world part of the machinery to think.
An Ecosystem Without a Shield
On March 28, 2023, in Hsinchu, Tsai Ming-kai did something he rarely does. He stood in front of reporters and made an argument. The occasion was the release of a white paper on Taiwan's chip-design industry by the Taiwan Semiconductor Industry Association and DigiTimes, for which he had served as chief consulting officer. Beside him sat representatives of Himax, Phison, Novatek, and Realtek. Novatek was one of the companies born, like MediaTek, from the breakup of UMC's products group a quarter-century earlier.
His argument was sober and specific. In October 2022 the United States had imposed export controls on China's advanced chip sector. As a result, Tsai said, Chinese government funds were flowing into mature chip technologies, and Taiwan's small and mid-size design houses would "possibly be the first to be affected." The United States, Europe, Japan, and South Korea were all rolling out hefty subsidies. "Even the European Union," he noted, "is now emphasizing the importance of advanced chip design capabilities." Taiwan's own chip legislation, amendments to the Act for Industrial Innovation that offered a 25 percent research tax break to a handful of companies, mostly subsidized manufacturing. He proposed a budget of more than NT$10 billion and something more strategic, covering design houses, system integrators, and software developers.
The white paper's numbers framed the stakes. Taiwan's chip-design sector held 18 percent of the global market in 2022, second only to the United States at 63 percent and just ahead of China at 15 percent. It generated around $40 billion in revenue, about 2.4 percent of Taiwan's GDP, and employed roughly 52,000 people. Without government support, the paper projected, Taiwan's share would slip to 17 percent by 2026, behind China's 18. Tsai added a domestic worry: "talent poaching and a low birthrate." The man who had founded the MediaTek Foundation in 2001 because engineers were the island's scarcest resource was now pointing out that the island was making fewer people.
K.S. Pua, the Malaysian-born chief executive of Phison, the world's leading NAND flash controller maker, spoke on the panel afterward and put the island's position more bleakly than Tsai would.
— K.S. Pua, CEO of Phison Electronics, Hsinchu, March 28, 2023China also has the powers of government, mass market, and capital. Taiwan, however, doesn't have these preexisting merits.
The Taipei Times covered the event the next morning. Its story led with a sentence attributing the remarks to "MediaTek Inc chairman Rick Tsai (蔡明介)." The English name was wrong. The Chinese characters were right. In a single caption, the two men had become one.
Tsai's views on China have never been simple. In January 2016, speaking at an event that promoted technology careers, he told Nikkei Asia that Taiwan should not exclude China when recruiting talent and accepting overseas capital. This was an unpopular position in a politically charged year. Seven years later he was warning that Chinese capital would flood the mature-node markets on which Taiwan's smaller designers depended. These are not contradictory positions so much as two readings of the same map at different moments. A company that sells the majority of its smartphone chips to Chinese brands and fabricates them at TSMC in Hsinchu lives in the strait. It cannot afford to be ideological about either shore.
In June 2025 a documentary about Taiwan's chip industry, five years in production and built on more than eighty interviews, opened in Taiwanese cinemas. Its original title was Silicon Shield, the phrase for the theory that Taiwan's semiconductor indispensability protects it from invasion. Before release the title was changed, on the advice that Taiwanese chip companies with factories in China did not want to be seen across the strait as a shield. The film became A Chip Odyssey. On the evening of June 26, at a cinema in Taipei's Xinyi District, a masked elderly couple bought their own tickets and sat in the last row. When the lights came up, the man walked slowly down the aisle with both hands raised. It was Morris Chang, ninety-four, who had declined to be interviewed for the film because he was finishing the second volume of his autobiography, nearly 200,000 characters, published by CommonWealth on November 29, 2024. The audience applauded. He stood in front of the first row and acknowledged them.
The founders of Taiwan's semiconductor industry belong to a generation that prefers the last row. Chang had to be filmed from the side at public events because he would not sit for the camera. Tsai would not come to the phone for Forbes. They are engineers who built the most strategically important industry on earth and seem faintly embarrassed whenever anyone points it out.
The Medal and the Modem
The IEEE Robert N. Noyce Medal, sponsored by Intel, is sometimes called the Nobel Prize of semiconductors. Its first recipient, in 2000, was Morris Chang. On November 29, 2023, MediaTek announced that the 2024 medal would go to Tsai Ming-kai, only the second recipient in the award's history from a Taiwanese company. The citation read: "For vision and leadership in the global semiconductor industry, democratizing technology access for billions of people." He received it in person at the IEEE Honors Ceremony in Boston on May 3, 2024, and spoke about embracing a new era of semiconductor innovation. He was seventy-four, and by one estimate MediaTek was enabling more than two billion new devices a year. In terms of the arithmetic of the 2014 keynote, the next five billion had largely been connected.
Honors had been accumulating for some time: the ITRI Laureate award in 2012, a place among Harvard Business Review's best-performing chief executives, honorary doctorates from National Taiwan University, National Tsing Hua University, and National Chiao Tung University. In 2025, Asian Scientist magazine named him to its Asian Scientist 100. In 2014 Forbes had estimated his fortune at $1.8 billion, twentieth in Taiwan, and the number has risen many times since. None of it seems to have made him chattier.
What remains, if you go looking for Tsai Ming-kai, is the product line. MediaTek's own description of itself says it supports productivity in 20 percent of homes and nearly one of every three mobile phones on earth. Its 2013 GNSS receiver was the first five-in-one multi-constellation chip, supporting GPS, GLONASS, Galileo, the Japanese QZSS, and China's Beidou, with every major positioning satellite system in the sky on one die. Its Wi-Fi 7 chips sit in routers. Its inference chip sits in Google's data centers.
At Mobile World Congress in 2025, MediaTek introduced the M90, a 5G-Advanced modem rated at up to 12 gigabits per second and built, among other things, for non-terrestrial networks. The specification says what that means in practice: satellite connectivity. A phone carrying it, in a place with no cell tower, in a field at the agricultural end of some island or continent, can look for a satellite overhead and, finding one, connect.
In 1957 a boy in Nanzhou heard over the radio that a machine was crossing the sky. Sixty-eight years later, the company he founded built a chip that lets a cheap phone in a field without a tower talk to one.
Part IIThe Playbook
What follows is an attempt to extract from Tsai Ming-kai's forty-year record a set of operating principles. These are not the ones he has stated, because he has stated very few, but the ones visible in what he built, bought, abandoned, and delegated. Some are uncomfortable. A founder who served bandits, survived a lost generation, and handed his company to a man who had been publicly removed from TSMC does not produce tidy lessons. He produces useful ones.
The principles
- 1.Sell the engineering department, not the component.
- 2.Treat today's copycat as tomorrow's anchor customer.
- 3.Leave the belt while it is still shining.
- 4.Count generations, not victories.
- 5.Make your customers fast, not just yourself.
- 6.Buy the missing function, especially from the trough.
- 7.Find the seam in your bundle before a competitor does.
- 8.Import your opposite temperament.
- 9.Let silence be a division of labor.
- 10.Breed companies, not just products.
- 11.Lobby for the ecosystem, not the champion.
- 12.Differentiation that cannot survive an audit is a liability.
Principle 1
Sell the engineering department, not the component
The incumbents of 2004 treated a phone chip as a part. You specified it, sourced it, and spent eighteen months integrating it with your own software and radio engineering. MediaTek's turnkey model treated the chip as the visible end of a much larger thing: reference designs, software stacks, and the "extensive system engineering assistance" that let a forty-person Shenzhen workshop behave like Nokia. The customer was not buying silicon. It was renting an engineering department it could never afford to build.
The economic consequence was that MediaTek redefined who could be a customer at all. When the barrier to building a phone is a thousand engineers, the market holds a dozen manufacturers. When it falls to a box, the market holds thousands, and by 2009 they were releasing 3,000 models a year. Tsai did not take share from the incumbents' suppliers. He manufactured new demand by enlarging the population of buyers.
The model has repeated at every scale. Rick Tsai's 2026 argument at ISSCC, that semiconductor companies should sell racks rather than chips, is the same idea applied to a data center. The unit of sale grows until it matches the customer's actual problem.
Tactic: Identify the most expensive capability your smallest potential customer lacks, and ship it inside your product so that capability stops being the price of admission.
Principle 2
Treat today's copycat as tomorrow's anchor customer
"Today's copycat, tomorrow's mainstream" was heard by critics as a rationalization for serving pirates. It is better read as a forecast about customer lifecycles. The shanzhai makers of 2009 included firms that would become, within a decade, some of the largest phone brands on earth. Huawei and ZTE were already MediaTek clients when Forbes came calling. Suppliers that refused the bandits on principle refused, in effect, to be present at the birth of their future largest customers.
Yu Mingto's distinction matters here. "It's not right to remain a copycat," he said, and innovation is a must "even if it started out by replicating others' work." The bet was not on imitation. It was on the trajectory out of imitation, which Taiwan itself had followed, from licensed technology at ITRI to the world's most advanced foundry. Tsai recognized the pattern because he had lived it.
The cost was reputational and lasting. Ten to twenty percent of MediaTek's chips went into outright pirated phones at one point, and the "king of bandit phones" label outlived the bandits. Any operator adopting this principle should price that cost honestly.
Tactic: Segment your low-end customers by trajectory rather than current status, and invest disproportionately in those whose capabilities are compounding, even when their present brand is an embarrassment.
Principle 3
Leave the belt while it is still shining
The optical-drive business was producing MediaTek's best numbers when Tsai began pulling engineers off it. That is the hardest move in management, because the evidence is always against it. The current product is profitable, its customers are loyal, and its engineers are the company's heroes. The next product is a cost center with no revenue and an uncertain market.
The boxing theory explains the move. A champion's belt marks a position everyone else is training to attack, and in semiconductors the attack is not even a rival. It is obsolescence itself. Solid-state storage did not need to beat MediaTek at optical-drive chips. It made the category irrelevant. The competitors who stayed in that ring were not outfought. They were left behind.
The 2011 trough shows the principle's limit. MediaTek moved into feature phones in time, but it was slower to move from feature phones into smartphones, and revenue fell by roughly a quarter in one year. Leaving the belt is not a single act. It has to happen every generation, and the reflex weakens when the current belt is very heavy.
Tactic: When a product line posts its best year, assign a fixed share of its best engineers to whatever is most likely to make that product irrelevant, and treat the reassignment as non-negotiable.
Principle 4
Count generations, not victories
Tsai's formulation that you must win three generations, not one, changes what a company measures. A single-generation win is measured in market share. A three-generation win is measured in the capacity to keep producing winners: talent pipelines, acquired capabilities, financial reserves, the institutional memory of having pivoted before.
Lay MediaTek's history out as generations and the pattern is clear. It won optical storage, won feature phones, nearly lost 4G smartphones, and won 5G. It is now contesting AI inference. Each victory funded and staffed the next attempt, and each defeat (2011, 2015–2017) was survived because the balance sheet and the engineering bench were built for losing a round.
B
The Belts
MediaTek's generational bets, read through Tsai's boxing theory.
| Generation | Signature product | Outcome | Status |
|---|---|---|---|
| Optical storage, 1997–early 2000s | CD-ROM and DVD chipsets (MT1389) | Dominance; exited before obsolescence | Won |
| Feature phones, 2004–2010 | MT62xx turnkey platform | 3M to 300M units in five years | Won |
| 3G smartphones, 2012–2014 | MT6592 "True Octa-Core," MT8135 | Revenue from NT$53.8B (2011) to NT$213.1B (2014) | Won |
| 4G LTE, 2014–2017 | Late integrated LTE modem | Customer defections; 2017 operating income NT$9.8B | Knocked down |
| 5G, 2019–present | Dimensity 1000 through 9500 | World's top smartphone chip vendor, Q3 2020 | Won |
| AI and data center, 2024– | Google TPU inference chip | Seat beside Broadcom at Google | In the ring |
Tactic: Evaluate leadership not by the current product's share but by whether the organization could survive losing that product entirely within the next two years.
Principle 5
Make your customers fast, not just yourself
Yu Mingto's comparison to Zara is the clearest statement of MediaTek's competitive mechanism. The company's speed did not come mainly from its own development cycles. It came from what its platform let customers do: 3,000 models a year from MediaTek's Chinese clients, ten times the output of the five largest brands combined. MediaTek's velocity was the sum of its customers' velocities.
This is a different strategic posture from most supplier relationships, in which the supplier competes on the merits of its component. Here the supplier competes on the throughput of its customer base. A customer who can ship a new model every few weeks will crush a customer who ships one a year in any fragmented market, and both will remember who made the difference.
The principle carries a dependency risk. Fast customers are also fickle customers. When Qualcomm offered integrated LTE and MediaTek could not, Alcatel One Touch and ZTE left quickly. The speed MediaTek had given them was speed they could take elsewhere.
Tactic: Measure your success partly by your customers' time-to-market, and design your product so that their fastest path to launch runs through you.
Principle 6
Buy the missing function, especially from the trough
Tsai's acquisition record reads as a methodical assembly of the modern single-chip device: cellular radio from Analog Devices ($350 million, closed January 2008), Wi-Fi and wired connectivity from Ralink (2011), signal-processing IP from Coresonic (2012), television from MStar ($3.8 billion, announced 2012), power management from Richtek (2016). Each deal closed a gap in the turnkey bundle.
The MStar timing is the instructive part. It was announced in June 2012, about a year after MediaTek's revenue had fallen to NT$53.8 billion and operating income to NT$4.8 billion. Most companies retrench after a year like that. Tsai committed billions to a domestic rival, and he structured the deal cautiously, with an initial 48 percent stake and an option on the rest. That caution proved useful when Chinese and South Korean regulators delayed the merger until February 2014. The company was buying while sellers were cheap and competitors were scared.
Tactic: Keep a standing map of the functions missing from your core offering, with acquisition targets for each, so that when a downturn arrives you can buy rather than merely cut.
Principle 7
Find the seam in your bundle before a competitor does
The 4G years are the clearest failure in Tsai's record. They are worth studying because the failure was structural, not technical. MediaTek's whole proposition was integration: one platform, one box, one supplier. When it arrived late with an integrated LTE modem, every MediaTek-based 4G phone needed a separate baseband chip. That added cost and broke the company's own promise. The seam was visible to customers before MediaTek closed it, and Qualcomm walked straight through it.
The lesson is that a bundling strategy is only as strong as its weakest missing piece. A company selling individual components can lag in one category and survive. A company selling complete solutions that lags in one category has, for that customer, no solution at all. Rick Tsai's resolve not to "repeat the indecision of the 4G period," and the two-track bet on Sub-6 and millimeter-wave 5G, was a direct response to this.
Tactic: Every quarter, list the components a customer must still buy from someone else to use your product, rank them by how likely a competitor is to bundle them, and fund the top item before you need it.
Principle 8
Import your opposite temperament
In 2017 Tsai Ming-kai, a design engineer, aphorist, and reluctant public figure, hired as co-CEO a manufacturing physicist known as the iron-fisted CEO, who had been removed from TSMC in a ten-minute board meeting after a breach of trust. The market called it a poison chalice. It turned out to be the best personnel decision of his career.
The value lay in the contrast. MediaTek's culture of speed, opportunism, and good-enough chips for the super-mid market had won three generations and was losing the fourth. What it lacked was the discipline of a fab: yield, precision, and the willingness to commit capital to a technology before the market had proved it. Rick Tsai had spent eight years at HP and two decades at TSMC learning exactly that. The founder did not try to become that person. He hired him.
T
The Two Tsais
No relation. Complementary by design.
| Dimension | Tsai Ming-kai | Rick Tsai |
|---|---|---|
| Born | April 6, 1950, Pingtung | April 19, 1951 |
| Training | Electrical engineering (NTU; MS, Cincinnati) | Physics (NTU); PhD materials science (Cornell) |
| Formative institution | ITRI, then UMC products group | HP Fort Collins, then TSMC fabs |
| Reputation | "Godfather of IC design"; reticent | "Iron-fisted CEO"; public |
| Defining setback | The 4G LTE delay, 2014–2017 | Removed as TSMC CEO, June 11, 2009 |
| Role since 2017 | Chairman | Co-CEO, then vice chairman and CEO |
Tactic: When your company stalls, list the qualities you personally lack that the next stage requires, and recruit for those even if the candidate's record carries a visible scar.
Principle 9
Let silence be a division of labor
Tsai ducked Forbes in 2009 and let his CFO make the argument. A decade later he let his CEO accept the industry awards, give the ISSCC plenary, and carry the company's public story. The few times he spoke in public, at the 2014 keynote, the 2023 white paper, and the 2024 Noyce ceremony, he spoke about the industry and the island rather than himself.
This is not just shyness. It is a structural choice that has kept the founder out of the news cycle, the product out of a cult of personality, and the company's messaging in the hands of people suited to deliver it. When the bandit-phone label was most dangerous, the founder's silence denied it a face. When MediaTek needed to sell itself to Google and Nvidia as a premium partner, the face it showed belonged to TSMC's former CEO.
The trade-off is that it leaves the founder's thinking largely invisible and his legacy written by others. The IEEE wrote his biography. The Taipei Times attached his Chinese name to the wrong English one.
Tactic: Decide deliberately which public conversations your company needs its founder in, and staff the rest with the executive whose credibility best fits each audience.
Principle 10
Breed companies, not just products
Before MediaTek, Tsai's most lasting work was organizational. At UMC he brought in modern EDA tools, mentored dozens of engineers who went on to lead the industry, and presided over a products group that was later split into MediaTek, Novatek, Faraday, and others. Many of those companies he chaired or directed. When he stood in Hsinchu in March 2023 to release the white paper, Novatek's representative was on the stage beside him.
The MediaTek Foundation, established in 2001, and the hundreds of university research collaborations the company has sponsored extend the same logic outward. Taiwan's constraint has never been capital or ambition. It has been engineers, a constraint Tsai named directly when he warned about talent poaching and a low birthrate. A founder who trains people and spins off companies builds an industry. One who hoards them builds a firm.
Tactic: Treat every strong team that outgrows its parent as a potential company rather than a retention problem, and keep a board seat rather than a grudge.
Principle 11
Lobby for the ecosystem, not the champion
Tsai's 2023 policy argument was conspicuously not about MediaTek. MediaTek, the largest chip designer in Taiwan and one of the largest in the world, could survive Chinese competition in mature nodes. His concern was the small and mid-size design houses, the system integrators, and the software developers. "Taiwan's version of the Chips Act," he said, "can't focus only on top chipmaking companies."
The self-interest is subtle but real. A fabless champion depends on a dense local supply of engineers, startups, IP vendors, and acquisition targets. Richtek, Ralink, and MStar were all Taiwanese. If the ecosystem thins, the champion's future acquisitions, hires, and partnerships thin with it. Lobbying for the whole field is lobbying for the soil the champion grows in.
Tactic: When you engage with policymakers, argue for the conditions that produce your future suppliers, hires, and acquisition targets rather than for benefits that accrue only to you.
Principle 12
Differentiation that cannot survive an audit is a liability
MediaTek has made two kinds of performance claims, and they ended very differently. The "True Octa-Core" trademark of 2013 rested on a real architectural difference: all eight cores of the MT6592 could run simultaneously, and competitors' could not. It stood up to scrutiny and became a marketing asset. "Sports Mode," exposed in April 2020, detected benchmarks from a whitelist and inflated their results. It did not stand up. It drew a public rebuke from Qualcomm, forced Oppo to issue a firmware fix, and got MediaTek chips delisted from UL's rankings.
For a company that has spent two decades trying to escape the word shanzhai, the asymmetry is brutal. A real distinction compounds, because every technical review confirms it. A manufactured distinction decays, because every technical review is a fresh chance for exposure, and the exposure confirms the oldest prejudice about the company. MediaTek's response, that the practice was industry-standard and device makers could choose, was defensible as procedure and disastrous as positioning.
M
Decisions That Shaped MediaTek
A timeline of the moves that defined Tsai Ming-kai's company.
1957
Sputnik launches; a seven-year-old in Pingtung resolves to become a scientist or engineer.
1983
Tsai joins UMC, Taiwan's first semiconductor company, after a stint at ITRI.
1997
UMC's multimedia unit is incorporated as MediaTek on May 28.
2001
IPO on July 23 under code 2454; MediaTek Foundation established.
2004
Mobile division launched; 3 million handset chipsets shipped to China.
2008
Purchase of Analog Devices' cellular divisions closes on January 11.
2011
Revenue falls to NT$53.8B; Ralink acquired May 5.
2012
$3.8B MStar deal announced June 22, from the trough.
2014
ISSCC keynote on "the next 5 billion"; integrated LTE ships late.
2017
Rick Tsai hired as co-CEO; operating income bottoms at NT$9.8B.
2019
Dimensity 1000 announced November 26.
2020
Overtakes Qualcomm in Q3; market value passes NT$1 trillion.
2023
Tsai urges an ecosystem-wide Taiwanese chips act on March 28.
2024
IEEE Robert N. Noyce Medal presented in Boston, May 3.
2026
MediaTek designs the inference chip for Google's eighth-generation TPU.
Tactic: Before publishing any performance claim, ask whether it would survive a hostile expert reproducing it independently, and if the answer is uncertain, cut the claim rather than the expert.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Tsai Ming-kai, as quoted by the IEEE on his 2014 ambitionConnect with the next 5 billion people.
— Tsai Ming-kai, World Finance 100, 2013The financial crisis has confirmed one thing… that emerging countries were almost not affected. Their economic development will continue and the opportunities there will be bigger.
— Tsai Ming-kai, release of the Taiwan IC design white paper, Hsinchu, March 28, 2023The U.S. export controls against China's advanced chip sector in October last year has prompted Chinese government funds to flow into the mature chip technology sector, and we believe Taiwan's small and mid-size chip design houses will possibly be the first to be affected.
— Tsai Ming-kai, Hsinchu, March 28, 2023Taiwan's version of the Chips Act can't focus only on top chipmaking companies. It should also include chip design, system integrators and the whole ecosystem.
— Tsai Ming-kai, Hsinchu, March 28, 2023Apart from global competition, Taiwan's chip design industry faces workforce challenges from talent poaching and a low birthrate.
Maxims
- Today's copycat, tomorrow's mainstream. Serve the imitator on the way up, and you will be his supplier when he becomes the brand.
- Win three generations, not one. A single championship is a target, so build an organization that can lose a round and still be standing for the next.
- Pivot at the peak. The best year of a product line is the right moment to move its best engineers toward whatever will replace it.
- Ship the engineering department. When customers can't afford the expertise, put it in the box and the market of possible buyers multiplies.
- Your customers' speed is your moat. A platform that lets thousands of small makers ship in weeks outruns any single giant that ships in years.
- A bundle breaks at its missing piece. For a customer, an integrated solution with one gap is no solution at all.
- Buy from the trough. Acquisitions struck in a downturn, when sellers are cheap and rivals are frightened, compound the longest.
- Hire the temperament you lack. The founder who builds the culture is rarely the operator who can discipline it, so import that operator, scar and all.
- Lobby for the soil, not the tree. A champion's future hires, partners, and acquisitions all grow from the ecosystem around it.
- Real distinctions compound; manufactured ones decay. Every audit rewards a true advantage and punishes a staged one.
