TSMC is Now Asia’s Biggest Listed Company. Here’s How It Got There.

TSMC has cornered nearly the entire market for advanced chipmaking.

Published in

EquitiesTracker

7 min readAug 14, 2022

In 2021, Taiwan Semiconductor Manufacturing Company (NYSE:TSM, TWSE: 2330) became Asia’s largest company by market capitalisation. TSMC has been thrust into the spotlight thanks to the world’s growing demand for chips. But just how did this little-known company manage to corner nearly the entire market for advanced chipmaking?

There are two important reasons for this, in our view.

A pioneer in the pure-play foundry business

The first reason has to do with TSMC’s business model. TSMC is a pure-play contract chip manufacturer, or “dedicated semiconductor foundry.” This means it only makes chips for other companies. Unlike Intel Corporation (NASDAQ:INTC) and Samsung, TSMC does not design or manufacture chips for its own brand — which means it does not compete with its own clients.

This game plan was unheard of when TSMC was founded in 1987. Intel, which held a literal monopoly on the computer chip industry, designed and manufactured its famous ‘Intel Inside’ chips.

But with TSMC, chip designers like AMD (NASDAQ:AMD) and Qualcomm (NASDAQ:QCOM) — Intel’s rivals — could focus on doing what they did best: designing the most advanced chips. They also didn’t have to invest in new foundries to make them.

This became increasingly important over time, due to the rising costs of chip manufacturing. As chips became smaller, yet increasingly powerful, they were becoming increasingly expensive and difficult to produce. TSMC’s foundries meant the cost of equipment and manpower could be spread across different customers, creating economies of scale.

Learning by doing

As TSMC gained more and more customers from all over the world, the company amassed the cash it needed to invest in improving its ability to make the world’s most advanced chips.

“The power of (TSMC’s) model didn’t become evident until they reached a very large scale. Once that calculation changed, it changed the name of the game,” said David Yoffie, a Harvard Business School professor and former Intel director.

The cost of the most advanced fabs surged from USD1 billion in the 2000s to nearly USD20 billion in 2020. A single chip equipment machine, such as those made by ASML, now costs more than a few hundred million dollars. Beyond that, there are high costs involved in research and development. In 2020, TSMC invested USD3.72 billion in R&D and USD17 billion in capital expenditure (capex).

This year, TSMC expects to spend as much as USD44 billion on capex. For perspective, there are only about 20 listed semiconductor companies with a market value of more than USD44 billion.

Besides TSMC, Samsung and Intel, few other companies can invest the massive amounts needed to stay on top of the R&D game. Building new chip fabs takes years and lots of careful planning. TSMC’s new 5nm fab that is being built in Arizona, for instance, is a copy of the one it has in Taiwan.

Besides Samsung, any other company trying to do what TSMC is doing will have to reinvent the wheel.

TSMC has excelled not only at making the most advanced chips, but also churning them out by the millions. “There is no one else who has the capacity as a foundry to deliver the volumes and capabilities” that TSMC’s customers want, says Risto Puhakka, president of semiconductor analytics firm VLSI Research.

TSMC’s expertise in making chips has made it indispensable to its clients’ success, too.

When AMD, an Intel rival, began outsourcing its chip manufacturing to TSMC, it began winning market share from its rivals. Since AMD began using TSMC’s foundries, Intel has lost market share in the CPU market to AMD. In 2021, AMD’s CPU market share hit its highest in 14 years.

It is worth noting that merely throwing money at chip production is no guarantee of success.

One of Intel’s biggest mistakes was rejecting a chance to manufacture chips for Apple’s iPhone in 2007. According to Paul Otellini, who led Intel from 2005 to 2013, Intel thought Apple would not be able to sell enough iPhones to pay off the costs of making those chips. So Apple made the iPhone with chips manufactured by other companies — first with Samsung, then with TSMC.

The iPhone would go on to sell “100x” more than anyone had expected, Otellini said.

As Intel’s manufacturing output fell below TSMC and Samsung, so did its expertise in making the most advanced chips, said Mark Lipacis, an analyst at Jefferies, an investment bank. Lipacis estimates that between 2015 and 2020, TSMC processed 3x the number of wafers Intel processed.

Practice makes perfect. TSMC has gotten better at making chips because it has been making them for almost all the biggest and best technology companies. Intel, on the other hand, is now two generations behind TSMC. TSMC can make chips using 5-nanometer and 3 nano-meter technology; Intel only expects to make 7-nanometer chips in 2023.

“It doesn’t have to do with not being as smart — the Taiwan Semi engineers are really smart, and the Intel engineers are really smart,” said Lipacis. “It has to do with your experience.”

The chip industry is complex and concentrated

This brings us to our second point about TSMC, which is a reflection of the semiconductor industry in general.

Each part of the semiconductor supply chain could be seen as a mini-industry of its own. AMD needs companies like TSMC and Siliconware Precision Industries — a unit of ASE Technology Holding (NYSE:ASX) — to make and test its chips. TSMC and Silicon Precision Industries, in turn, depend on chip equipment makers to make machines for their plants.

Further down the supply chain, equipment makers rely on thousands of
global suppliers for parts and components to make its lithography
machines. Each new chip represents the cumulative effort of thousands of
engineers across the world, spread across different companies.

A handful of firms dominate each part of the supply chain. It is expensive
and risky for customers to switch to new vendors. Smaller players do not
have pricing power, which makes it harder for them to be profitable. The
high switching cost, need for economies of scale, learning curve (making
chips is hard!) , and established supplier relationships all build up over
time. This grows the “invisible moat” around the leading players.

As a result, the top one or two players in any given niche of the chip
industry “earn all the economic profits in that niche due to scale,” says
Christopher Thomas, a director at Velodyne Lidar (NASDAQ:VLDR) and
former General Manager of Intel China.

Understandably, this makes it hard for newcomers to steal away market share
from the biggest players.

As examples, Thomas cites NVIDIA Corporation (NASDAQ:NVDA), which created the GPU segment in 1999 and has never given up its leadership since.

TSMC created the first pure-play chip foundry in 1987, and has been dominant in this space for decades.

Similarly, the chip equipment industry has become increasingly concentrated.

In 2004 the top ten players earned 58% of all chip equipment sales, according to a U.S. government report. In 2018, the top ten were responsible for 69% of sales. The top five players — Applied Materials (NASDAQ:AMAT), Tokyo Electron (TYO: 8035), Lam Research (NASDAQ:LCRX), ASML (NASDAQ:ASML) and KLA-Tencor (NASDAQ:KLAC) — were responsible for 63% of all revenue.

It gets more interesting when you delve deeper within certain categories of this sector, such as EUV lithography. In EUV lithography, for instance, a single company — ASML — holds a 100% market share.

THE TAKEAWAYJapan and the United States, two of the world's most technologically advanced countries, are subsidizing the construction of local TSMC plants. This is the clearest evidence yet that TSMC has a sizeable technical edge above its rivals, something which is unlikely to change any time soon. TSMC's spectacular rise reflects the nature of the semiconductor industry, where the winners tend to keep winning.