First of all, I have never received formal academic training in history, so I am unsure if my writing will be up to the standard of this sub, but I will give it a try as I have related industry experience and have always been interested in the history of the semiconductor industry.

So first let's lay down some crucial definitions before I can approach this problem. In the industry, loosely speaking, there are four types of companies:

1. IDM (Integrated Device Manufacturers): To put it simply, they design their own semiconductor chips to suit their own needs and manufacture them in their own facilities. They usually do not manufacture chips for other companies. Representative companies include Intel and Texas Instruments.
2. Fabless (Fabrication less): They only design their chips and outsource the actual making to other companies. Kinda like a painter letting a printer shop do the printing instead of printing the posters on one's own. Representative companies include Nvidia and Qualcomm.
3. Pure-Play: They only manufacture chips based on designs sent by their customers, and they offer some auxiliary tools to facilitate the designing process. They never design the chips themselves. Representative companies include TSMC (Taiwan Semiconductor Manufacturing Company, a very straightforward name isn't it)
4. OAST (Outsourced Semiconductor Assembly and Test Services): They mainly do assembly and testing after a chip's core parts are manufactured. The line between pure-play and OAST is sometimes blurry as many pure-play companies also offer packaging and testing services.

​

When we talk about semiconductor manufacturing or fabrication, another important definition is process node or simply node. It describes a set of manufacturing processes or techniques. Each node corresponds to one generation of technology. You can think of 3G, 4G, and 5G in telecommunication. Nowadays widely adopted nodes are 28/22nm and 14nm, while the most cutting edge is 3nm.

Here nm refers to the unit nanometer. Originally this number is supposed to measure the transistor gate length that this process can manufacture. But that really hasn't been the case since 1994. Nowadays this number is more or less a marketing term that pure-play companies update whenever they feel confident that their newest technology can bring a significant performance boost over the last generation. It's hard to find what's actually 5nm in TSMC's 5nm nodes

To build a semiconductor fabrication plant (fab) is very, very, very expensive. Nowadays, to build a cutting edge fab, its been estimated it'd cost 20 billion dollars to build just one such fab. The cost to build a fab will be a recurring theme in this response.

So when Taiwan first decided to go into semiconductor manufacturing back in the 70s, it was nowhere as expensive back then to build a fab but still far too expensive for a Taiwanese company to build one.

So, what it took was top-down Taiwanese government direction and financial support to help jump start the process. ( I think the Taiwanese are far too eager to credit the Japanese for Taiwan's economic development, and not the enough to the KMT which was the government in charge during the Taiwan Miracle).

With the economic oil crisis of 1973 and the disruption it caused to the Taiwanese economy, the Taiwanese government decided it needed to develop new industries such as the electronics industry. So, Taiwan set up the Industrial Technology Research Institute (ITRI) in 1973 to do applied research and help develop new industries for Taiwan.

Then, in 1974, over breakfast at a soy milk stall, Dr. Pan Wen-Yuan and 6 Taiwanese government officials decided that integrated circuits had the biggest growth potential in the electronics industry and that's what Taiwan should focus on.

So, in 1976, Taiwan got RCA to agree to do a technology transfer of its semiconductor technology to Taiwan. Using that knowledge, ITRI built a wafer fab that becomes Taiwan's first semiconductor company, United Microelectronics Company (UMC).

(It should be noted that technology transfer and government support is what China is trying to do today to build its semiconductor industry, but Taiwan's development was far less controversial because it is a geopolitical ally of the United States and so Taiwan was heralded as "Free China" even though Taiwan was under the control of the authoritarian control of the KMT.)

But, we're not talking about Taiwan semiconductors because of UMC which had a limited ceiling because it focused on the Taiwanese domestic market and its chips were for simpler devices.

Instead, we're talking about Taiwan semiconductors because of a different Taiwanese semiconductor company, TSMC, and the decisions Morris Chang, its founder, made.

Morris Chang worked at Texas Instruments (TI) for 25 years, rising through the ranks and becoming the VP of TI's semiconductor division. However, he left TI after he hit the bamboo ceiling because of his Chinese ethnicity. (Others have argued that his career stalled at TI because he had been transferred to the struggling Consumers Business and could not turn it around).

Chang comes to the realization that he'll never be the CEO of a major US Company. So, the Taiwanese government is able to convince Chang, a Chinese-American with no previous ties to Taiwan, to come to Taiwan and head ITRI in 1985.

Then, America hit Japan, a major producers of semiconductors, with sanctions on Japanese semiconductors that drives up prices on Japanese semiconductors. (Japan is an even more important geopolitical ally than Taiwan, but this was also a time of fear of rising yellow peril in the United States).

The Taiwan government realizes this is an opportunity and so it tells Chang it wants him to start a semiconductor business and so TSMC is spun off from ITRI.

The Taiwanese government owns half of TSMC, but insists that the other half be privately financed. Part of that private money comes from Phillips, a Dutch Company. The rest of the private money comes from Taiwanese businessmen, but only after the Taiwanese government twists their arms to invest in TSMC. (Some of those Taiwanese businessmen were these geriatric men with no understanding or appreciation of what semiconductors were).

At TI, Chang had noticed there were people itching to start their own companies but couldn't because they couldn't raise enough money to build an expensive fab. Or, if you had a design but no fab, you'd have to go to a major semiconductor company like IBM who'd mistreat their customers. IBM would only manufacture your chips only after IBM was done producing their own chips. Even worse, IBM would insist that you transfer over your design to them, and then come out with a competing chip under the IBM label.

So, Chang has this Field of Dreams epiphany- that if he builds it (fab), they will come.

TSMC differentiates itself by Chang pioneering this 'pure play' model where TSMC will just concentrate on the manufacture of semiconductors when the motto in the industry was 'real men have fabs', ie companies would both design and manufacture the chips.

And, companies could bring their designs to TSMC and trust that it won't compete and come out with its own chips under a TSMC label. This is why most people don't know the name of TSMC even though it was the world's 10th most vaulable company last month.

For a company to design and build a semiconductor, the design part is much more profitable because building a fab is so expensive. So, companies concentrate on the profitable design aspect and then outsource the manufacturing to TSMC. New companies like NVadia and Qualcom arise, and they see explosive growth by focusing on design.

But, the inflection point to TSMC's success comes in 2009 with the Iphone.

Apple had come out with the Iphone in 2007, and wanted to use Intel chips but Intel declined that offer because Apple's price was too low for Intel. (This was the beginning of Intel's decline- although Intel chips still dominate PCs, it missed the boat on mobile phones).

So, Apple used Samsung chips for Apple's first Iphones, but Apple HATES Samsung and sees Samsung as a rival as Samsung comes out with their own mobile phones.

Thus, TSMC's policy of being this trust worthy non-competitor pays off as Apple then turned to TSMC to build the chips for the Iphones. Chang pushed in all his chips, and made a big bet on Apple- TSMC built a 9 billion dollar fab and hired additional 6000 workers.

This bet paid off and as TSMC generated more and more money and profits, it poured that money back into building newer, better fabs which helps to separate it more and more from its competitors.