Rise of TSMC–why and how to replicate?

Superior performance in intangible asset development and monetisation for improving chip performance and higher wafer yield

TSMC success as money printing machine has been due to superior performance in systematic learning and monetisation in chip design optimisation and wafer yield enhancement
TSMC success as money printing machine has been due to superior performance in systematic learning and monetisation in chip design optimisation and wafer yield enhancement

With the success of inventing semiconductor devices and creating the iconic success of Silicon Valley, why does America rely on TSMC to deliver high-end microchips? The rise of TSMC (Taiwan Semiconductor Manufacturing Company) as a global microchip icon in a relatedly backward industrial economy is intriguing. Unlike America’s Silicon Valley, Taiwan did not offer a rich Innovation ecosystem to give birth to TSMC and nurture it. Besides, TSMC is not a typical example of a startup led by an iconic character like Steve Jobs or Bill Gates.  Furthermore, unlike American semiconductor firms like Fairchild or Texas Instruments, TSMC did not get preferential treatment from military customers. It did not benefit from a sizeable domestic market, high-end universities, rich publication records, trade agreements, billions of dollars of public money, and superb infrastructure.  In conventional innovation indicators, Taiwan has been behind many countries, including the USA, Canada, Japan, Germany, and many more.

No company in the world can profit from a microchip-making edge like TSMC. Hence, TSMC produces as high as 90 percent of global high-end chips, leaving the rest to Samsung. More importantly, silicon chip inventing America depends on TSMC for its high-end products like iPhone and smart weapons. Besides, once global US icon Intel has been desperately trying to catch up with TSMC in process node maturity—but increasingly falling behind. While TSMC keeps moving from 5nm to 4nm to 3nm, Intel keeps missing deadlines from 10nm to 7nm—widening the gap. Besides, while Intel keeps reporting revenue decline and loss, TSMC keeps reporting record revenue and profit growth. Such reality raises a vital question: what is the underlying cause of the rise of TSMC as an unbeatable iconic success?  

Key Takeawys from the rise of TSMC

  • Rise of monopoly from the humble beginning through flywheel effect–the creation of the innovation flywheel effect through conversion of insights of foundry processes and customers’ requirements into growing yield and profitable process node migration.
  • Riding the fabless model–synchronization between TSMC foundry services and the rise of the fabless model created a resonance of growth.
  • Reliance of mobile handset makers on fabless companies’ processors–instead of using processors provided by integrated device makers (IDMs) like Intel or Texas Instruments, mobile handset makers relied on fabless companies like Qualcomm or MediaTech. As a result, TSMC’s foundry service volume kept rising along with the exponential growth of mobile phone penetration. Furthermore, smartphone growth accelerated the demand for upgrading process nodes to reduce energy wastage and increase processor density.
  • Avoidance of foundry service and smartphone wave by IDMs–as processor production for PC was highly profitable, IDMs like Intel avoided tapping into the smartphone processors.
  • Focus on R&D and commercialization of ideas as foundry service yield–instead of just operating the fab, TSMC focused on R&D for generating ideas and converting them into higher yield and lower process node migration lead time.
  • Forming partnerships with customers and equipment suppliers–partnership formation with customers like Apple and equipment makers like ASML has led to very high process optimization and chip performance.
  • Transfer of Dr. Chang’s acumen into TSMC’s institutional competence–TSMC’s success is partly rooted in the institutionalization of Dr. Chang’s experience in TI and his internal burning to outperform industry leaders.

Formation and historical timeline of TSMC

Taiwan Semiconductor Manufacturing Company (TSMC) is a Taiwanese microchip-producing foundry service provider that has a global monopolistic position in high-end process nodes.

  • 1976–seeding semiconductor competence in Taiwan: Signing of contract between Taiwan’s ITRI and RCA for transferring IC design and manufacturing capacity to Taiwn by American Radio Corporation (RCA). ITRI’s maiden success from this technology transfer was designing and manufacturing a timer chip–IC CIC001–in 1976. After two years, in 1978, ITRI succeeded in developing and delivering a microchip for a wristwatch. Through the process, Taiwan started developing human resources and market linkage with the semiconductor clients.
  • 1985–Dr. Morris Chang joined ITRI to scale up embryonic semiconductor capacity: Upon retiring from Texas Instruments as vice president of the semiconductor unit, Dr. Morris Chang joined ITRI. His mission at ITRI was to commercially scale up ITRI’s embryonic semiconductor design and manufacturing capacity.
  • 1987–founding of TSMC: Dr. Morris Chang ventured into scaling up ITRI’s semiconductor competence by founding Taiwan Semiconductor Manufacturing Company (TSMC). TSMC began the journey as a joint venture between the Government of Taiwan, Dutch Philips, and a few private investors. On the same year, TSMC’s Fab 1 started with 3 micrometer process node that was a few generation behind of Intel. From the very beginning, Dr. Morris Chang focused on delivering foundry services for printing microchip designs on wafers.
  • 1990–Fab 2, fully owned by TSMC
  • 1996–First US fab of TSMC
  • 2008–Open innovation platform is launched.
  • 2014–Symbiotic relationship between Apple and TSMC began when Apple first used TSMC to produce chips for the iPhone 6.
  • 2020–TSMC emerged as the monopoly in the latest process node, starting volume production at 5nm FinFET technology while Intel was struggling to graduate from 10nm node.
  • 2022–TSMC foundry produced Apple’s A17 bionic chip using 3nm process node.

Rise of TSMC has caused global trade frictions

The rise of TSMC has caused a chip war—pouring billions in subsidies and imposing sanctions

Due to the enviable success of TSMC, dependence on its ability to sharpen civilian and military edges, and the impending threat of loss of Taiwan to China, the White House has got nervous—triggering Chip War. The recently enacted Chip Act has allocated a staggering $76 billion in subsidies and tax incentives for replicating the success of TSMC on American soil.  On the other hand, by poaching people from TSMC and buying equipment from the sources where TSMC does, China’s billions of dollars of public money have been ending up in a futile exercise. Besides, Japan, Europe and South Korea have been losing market share of TSMC. Furthermore, IMDs like Intel, AMD and many more depends on TSMC’s foundry services for their high-end chips.  Such a reality raises the question—what is the underlying recipe for the rise of TSMC, and is it replicable?

Some hypotheses explaining the underlying cause have been subsidies, low-cost workforce, and undervalued Taiwanese currencies. Yes, partly, there are valid. But are they sufficient conditions? By offering all of them, will a country succeed in replicating the rise of TSMC? If yes, why could India not emerge as a global semiconductor powerhouse by giving 100% subsidies in capital expenditure to its Semiconductor Complex ltd? On the other hand, achieving Intel’s investment in the early 1970s, followed by joining an additional 300+ MNCs, why could Malaysia not attain a similar height?   

Rise of TSMC by nurturing American inventions and outperforming US companies 

TSMC has created this success by nurturing American inventions and making iconic American companies like Intel, AMD, Apple, and Qualcomm its customers. Instead of leveraging what existed, it has created a rich ecosystem. Intriguingly, it has upgraded Taiwan from the idea importing a middle-income economy to the idea exporting a high-income miracle. In addition to tech giants, economists and innovation experts are bewildered in interpreting the cause. What is the secret sauce behind the rise of TSMC? Does the underlying cause offer a new recipe for attaining an innovation edge and creating an idea economy?

Rise of TSMC from the commercialization of R&D outputs

TSMC appears to be a wonder to profit from research outputs in a vital and highly competitive industry. Ironically, TSMC’s R&D capacity in the 1980s was far inferior to that of the USA, Europe and Japan. Interestingly, it profits from age-old research agenda—design for manufacturing. It sources all the equipment and chemicals from 3rd parties to fabricate microchips designed by others. Its core competence has been to align the designs with the production process and update the process to reduce defects and improve chip performance. Hence, TSMC focuses on learning and transferring that learning into fine-tuned chip designs and production processes.

Besides, it captures that learning into institutional knowledge and idea and keeps replicating it at scale. Does it mean all other semiconductor firms like Intel or Global Foundry have not been focusing on it? Of course, they have been working on it. This is a core manufacturing competence. The difference is that TSMC has been doing this essential work better than anybody else in the industry, resulting in winning the global race. 

Learning from doing and repeating at scale—attaining higher yield faster than competitors  

Since the very beginning, under the leadership of Dr. Morris Chang, TSMC has been focusing on learning chip design optimization for manufacturing, yield enhancement and process improvement. Instead of just processing wafers in printing the designs given by clients, TSMC engineers kept studying them and providing feedback to designers to optimize them for maximum process yield and chip performance. Furthermore, its engineers kept capturing that learning as patents and trade secrets. And they kept using and adding other ideas, deriving growing benefits from the same foundry equipment and operating cost. They kept repeating this process at an immense scale. Furthermore, they succeeded doing it better than their competitors. Consequently, due to the superior cumulative effect of this learning practice, TSMC began outperforming competitors.  

Throughput and diversity contributed to learning productivity

Unlike IDMs like Intel, TSMC was in the business of processing wafers for printing the designs of diverse clients. Hence, in addition to facing challenges, TSMC engineers also got the chance to learn from diversity. Furthermore, TSMC’s growing wafer processing performance kept increasing work orders, expanding the learning scope. Hence, a self-reinstating or virtuous cycle started to take place, making TSMC increasingly more capable than other chip makers. Therefore, although everyone else can access same machines and chemicals such as TSMC, how to push the apparent limit is not written in the manual. TSMC succeeded in doing far better than competitors due to superior performance in on job learning—critical underlying criteria of TSMC’s superior performance.

Through this self-reinforcing capability, TSMC kept attaining higher yields faster than the competition and profitably exploiting the current process node earlier than others. Consequentially, TSMC kept moving to the next node faster than others, including Intel. Due to the economics of cost and performance, leading customers like Apple kept giving high-margin, high-volume work orders to TSMC’s next process node. For example, TSMC’s record profit and revenue growth in 2022 has been due to Apple’s orders of processing A16 chip using 4nm process node.  As a result, in process node maturity, volume, and profitability, TSMC kept growing faster than competitors, leading to an unbeatable monopoly in high-end microchip production.   

Alluring TSMC through subsidies to set up fabs—how will it help acquire semiconductor edge?

Due to the enviable rise and the capability of producing high-end chips, countries like the USA and India have been alluring TSMC. They have been offering billions of dollars in subsidies, covering as high as 50% of capital expenditure, to entice TSMC. In some cases, they have been succeeding. For example, TSMC plans a $40 billion investment in Phoenix, Arizona. But what will host countries receive upon succeeding with TSMC’s plants on their soil?

TSMC sources all equipment and chemicals from 3rd party vendors, mainly from the USA, Japan, and Dutch ASML. TSMC’s value addition has been the use of knowledge and ideas which they have been accumulating. Hence, the parent company in Taiwan will be getting more significant competitive edge, caused by subsidies, to monetize already developed intangibles. They will also be developing additional intellectual assets through R&D operations in Taiwan. Hence, host countries will get minimal economic benefits and semiconductor capability by enticing TSMC to set up plants through subsidies. Such a hosting success will not sharpen those countries’ domestic semiconductor edge.

Why is it a big deal to replicate TSMC’s learning practice?

As explained, the underlying force for the rise of TSMC has been learning, capturing, and turning the captured knowledge and ideas into chip designs and improving wafer processing. Such a learning curve looks like S-curve, tending to ramp up and reach saturation. The migration to a new process node opens a new S-curve of learning, and the scope of moving to the next nodes has been shrinking. The technological feasibility and economic viability of upgrading process nodes to keep Moore’s law alive have been getting thin. Hence, even for TSMC, keeping the innovation engine alive for higher yield and better-performing chips has been flattening.

The cost of migration to successive nodes and the complexity of learning for deriving economic benefits has been exponentially growing. Not all applications or product innovations do offer an adequate willingness to pay for the processing of wafers using the following process node. For example, in addition to weak performance in moving to the next process node (e.g., from 10 nm to 7nm), Intel has been facing the barrier of monetization due to the saturation of the PC wave. On the other hand, TSMC got a far greater scope of monetization of learning in moving to 7nm, 5nm, and 4nm process nodes due to the iPhone’s ability to turn it into a profitable business opportunity. 

How to replicate the rise of TSMC?

First, we need to pursue innovation waves, often generated from Reinvention, to exploit the profitable opportunity for growing microchip complexity. For producing those chips with increasing yield, the focus should be on learning to optimize chip designs for manufacturing. Besides, learning should focus on reducing defects for higher yield through improving production processes. Often such knowledge would surface as Tacit capability for codifying that, underlying science should be discovered and repeatable technique should be spelled so institutional competence base keeps growing. The next challenge is to monetize the investment made in learning to a sufficient scale so that profitability from R&D investment keeps growing.  Furthermore, instead of making isolated learning activities, it should be well orchestrated to create a virtuous effect for moving to the next more profitable state by leveraging the past success of learning and monetization.

The reason behind the rise of TSMC is overlooked in strategy and policy discourse

Although governments worldwide have been willing to pump billions of dollars into expanding or establishing new facilities for producing high-end chips on their soil, there has been a soft focus on building TSMC’s self-reinforcing learning success. It seems the agenda of developing the semiconductor industry has been in setting up and operating plants. But despite the growing maturity, still, the semiconductor industry is amenable to generating profitable revenue out of learning. Hence, the focus should be on learning and monetization for higher chip performance and wafer processing yield.

Unfortunately, this core underlying reason for the rise of TSMC is missing in the strategy and policy discourse. Instead, the focus has been on giving subsidies and creating barriers for competitors to access technology. We have faced this reality due to the circulation of the wrong thesis: Asia dominates in microchip making due to subsidies and low-cost labor advantage. There is no denying that they have played a role in the rise of TSMC, but they are not sufficient to win the race.

As a result, through subsidies and trade restrictions, TSMC’s fabs in Arizona, Japan, Germany, and wherever will remain dependent on the company’s deep expertise back in Taiwan. Consequently, host countries like the USA will likely fail to do what TSMC does upon getting leading-edge fabs.

Enlighten others

Subscribe

Subscribe now to our newsletter

You May Also Like

Leave a Reply

Your email address will not be published. Required fields are marked *