Chip peace

By Daniel Heyler

Chip peace

The next wave of global electronics growth driven by Artificial Intelligence (AI) and Machine Learning (ML) upgrades is contingent on maintaining the current status quo of Taiwan-China-US relations. There can be squabbling and muscle-flexing, but a hot war would be catastrophic for the technology sector and, for that matter, the global economy. Furthermore, if either side decided to destroy the semiconductor fabrication plants, or fabs of Taiwan Semiconductor Manufacturing Company Limited’s (NYSE: TSM) (assuming a scorched earth policy if they expected to lose a war in the Taiwan Strait), the world would face two decades of economic contraction and a level of social chaos and hardship not seen since World War II.  

Decoupling over the next decade is also unlikely as neither China nor the US can even marginally diminish their dependency on TSMC. The United States’ Chip Act will fail miserably and China’s two-decade effort to scale its own foundries has failed. With tech controls, the future is even more challenging.  

Unsuccessful onshoring would prove to be a stabilizing geopolitical force since neither side could supply their respective tech industries, even at a subsistence level.  Of course, this analysis assumes rationality on the part of decision-makers (rather than military hawks) in-so-far as both sides shall act out of the self-interest of their own economies.

China can gradually close the gap with TSMC in key high-volume areas such as mobile phone and memory.  Huawei’s Mate 60 uses processors designed internally and fabricated by SMIC on 7 nanometers (nm) in China. While 7 nm is several years behind TSMC’s 3 nm (in high volume), it is “good-enough” technology to power smartphones.  

Released in September 2023, the new Huawei phone relies on imported mobile-specific memory from Korea.  Technology controls from the US will not be effective at pushing China geopolitically.  China’s economy, which is increasingly technology dependent, can remain strong and competitive globally using mainstream technology. The US cannot cut off mainstream technology sales to China without decimating its own tech companies.  The optimal scenario for both sides has been when China has had access to leading edge chips which are used in high-end electronics that are sold in China and exported globally, such as iPhones and servers. The entire supply chain and global consumer win in the end.

Media drama and populist politics

The media and some recent books about the concentration of chip production in Taiwan as a significant risk to national security have exacerbated political tensions.  Perhaps, these have served a purpose to educate the general population. But some have done a disservice by dramatizing the situation which has emboldened populist political narratives. 

The reality is that Taiwan’s importance to tech is a five-decade development driven by a multitude of factors of which industrial policy or lack-of, thereof, is the least important.  Both China and the US have made attempts at government-funded semiconductor manufacturing with inconsequential impact relative to TSMC’s success. 

The only factor that has changed is the intensity with which both sides want their own fabs.  Billions have gone to Globalfoundries (AMD) and now Intel (also in decline) is looking for subsidies. 

How to measure geopolitical risk

Geopolitical risk is a complicated equation, incorporating soft and hard factors, such as populations’ willingness to fight, the real necessity of the war, and decision-makers.  Global investors have historically been accurate assessors of geopolitical risk through the valuation (risk-premium) assigned to regions and companies.  Taiwan-domiciled companies have historically traded at lower P/E multiples due to “country risk”.  That discount has not changed even in recent years.  The valuation of TSMC and even its key customers, including Nvidia, AMD, Apple, Tesla, Meta and others who depend on it to build nearly all their advanced chips have been unaffected by the negative news flow. 

If one reviews the past seven decades, China’s military build-up, drills and threats to Taiwan have consistently been focused responses to what China calls “separatist” forces.  When the US recognized the People’s Republic of China (PRC) in 1979, it agreed to a one-China policy with peaceful re-unification.  Since then, every PRC regime has considered a formal declaration of independence by Taiwan as equivalent to an act of war.  Taiwanese have believed China’s invasion as highly unlikely as long as the One-China policy is observed.  In recent surveys, Taiwanese have grown concerned about confrontation between China and the US in the region.  Surveys show 50% of Taiwanese now distrust China and the US equally. This could be due to effort to move TSMC to the US, the controversial visit to Taiwan by US House Speaker Nancy Pelosi in August 2022, the Ukraine war and isolationist policies to China.

Irreversible dominance of TSMC

TSMC is a remarkable, centralized organization that should be deemed the most important company in the world because it provides the critical chip companies that drive the digital information economy. Its vision, adaptability and execution over three decades have been second to none in the tech sector. But, equally important to understand, is how TSMC has been the beneficiary of market shifts and ruthless chip economics.  

Fastest time-to-volume for customers 

TSMC’s founder Morris Chang’s idea of a dedicated foundry model scaled up during the PC-clone era as chip designers left their inflexible IDMs (integrated device manufacturers) to start “fabless” chip companies catering to Intel and Microsoft-compatible logic chips and peripherals.  The efficient and responsive business model repeated itself for the next three decades. Customer-centric TSMC was always there to take ever more market share with each cycle and each shift in end-markets. 

Capturing and driving each growth wave

New growth waves in electronics layered on top of prior markets creating a massive increase in penetration of semiconductors into the global economy.  The initial Asian semiconductor success grew out of consumer electronics, personal computers (PCs) and peripherals, then local area networks (LAN), feature phones from Nokia, Motorola and billions of white-box phones from China, the global internet build-out, and, the biggest wave of all, smartphones that experienced a decade of upgrades in displays, cameras, internet connectivity, and high quality video up to 4G. Now that TSMC and its customers have surpassed Intel, massive share gains will continue to be made, rendering Intel’s plea for Chip’s Act money in consequential. 

Technology leadership

With a clear and focused strategy to be everyone’s foundry, including those who still had fabs, TSMC grew consistently for three decades for three key reasons – its advanced technology, diversification into new target growth areas once PCs waned, and manufacturing excellence.  With each economic cycle, the semiconductor industry lost marginal players and drove a need to cut overhead and outsource. With each semiconductor cycle, TSMC gained a generation in technology. It started as a low-end or legacy technology provider in pure digital and would keep investing during downturns. Hence, in each up-turn it gained a generation, and, with that, came more end-markets and customers.

Comprehensive ecosystem 

In terms of capabilities, TSMC has been able to expand and take market share in three directions – to lower-end nodes, higher-nodes and into deeper services and support.  TSMC should be thought of as a comprehensive ecosystem of specialists and teams with hundreds of external suppliers and customers plugged into it. TSMC has needed to cluster its fabs close to one-another, first in the north of Taiwan. Then, due to earthquake risk and the completion of the north-south high-speed rail line, it built a cluster in the south of Taiwan.  Both China and the US urged TSMC to build stand-alone fabs on their soil and neither fab was successful. 

Leadership in manufacturing scale and quality

TSMC continuously reduced costs in its manufacturing facilities with complicated transistor structure transitions, process changes, and equipment upgrades. The big wins were scaling its 8-inch chip early in that generation and then again when the industry moved to 12-inch wafer sizes. Many IDMs were unable to stay competitive with TSMC and moved to an outsourcing model or lost share to those who used TSMC.  TSMC’s fabless customers destroyed the IDM incumbents in mobile phones.  In data centers, AMD and Nvidia are grabbing share from Intel which used to be a monopoly in the data center industry.  We think Intel will see share of overall compute drop to 20-25% over the next five years from around 80% currently.

Dominance spans from legacy to high-performance technologies

TSMC has also continuously developed specialty analog and mixed signal technologies that can utilize its fully-depreciated legacy equipment.  In short, TSMC’s customers sell to almost every end-market – high performance data centers, mobile phones, communications infrastructure, automotive, consumer electronics and the US military.  As software and applications are constantly upgraded, semiconductors also get upgraded for high-performance, lower-lower consumption, or cost-down.  One missing chip in a car, whether it’s two dollars or $25,000, renders it useless. The same applies to all electronics, from rice-cookers to super-computers. 

TSMC has penetrated nearly every end-market 

Today, TSMC’s mix of technology mirrors the global industry overall. Any disruption to TSMC means shortages in autos, super-computers, data centers, mobile phones, televisions, civil aviation, military and aerospace.  The incredible scale, expertise, intellectual property, and culture of excellence was built over several decades. It’s evolved into so much more than a foundry which simply takes designs (software) and turns them into working logic integrated circuits.  This cannot be replicated anywhere because it requires a concentrated workforce, culture and coordination.  

TSMC’s culture of excellence and responsiveness 

Employees and suppliers to TSMC in Taiwan respond in real time day-or-night to resolve problems.  Missing shipments to key customers such as Apple would cost that company billions in sales. Time-to-volume with high-yielding factories is where the rubber hits the road. TSMC’s unique culture of service and execution is now scaled in an army of 6,000 top engineers, chemists, and physicists who are supported by an even larger army 65,000 support, sales, marketing and accounting. 

High Return on Invested Capital (ROIC) from Taiwan is critical 

In short, TSMC’s formula will continue to be the same: (1) incredible execution  at ramping technology and bringing its customers’ designs to market (from its home base), (2) time-to-volume, (3) cost-down which lowers electronics costs for global consumers, (4) increasing functional and performance to drive lead-edge applications, such as computer/AI, communications, and automation, and (5) high ROIC from Taiwan’s operations. All these are  necessary to continue to bring up new technology, scale volume and drive down costs which will drive adoption of next generation products.

High cost in Arizona will cap orders from key customers

Now that TSMC has a few US customers that are large enough to fill one or two $10 billion wafer fabs, it could give the US another try. They continue to face delays on construction and are unable to find enough qualified personal within the US. Even with political tensions high across the Taiwan Strait, few employees are interested in living and working in the United States.  As a result, TSMC’s manufacturing cost are estimated to be 30-50% higher in Arizona than Taiwan.  Some customers have agreed to share the higher costs in theory.  Once operational, why would customers buy high-priced chips when they can get cheaper ones from TSMC’s Taiwan factories? 

What this means for policies towards Taiwan

If our analysis is correct (and 95% of semiconductor veterans will agree), then the implications for geopolitics are simple.  Whatever your politics, neither China nor US economies can function without a fully operational TSMC.  

The most dangerous scenario would be a partial success of scaling in the US or China whereby leaders see a manageable decoupling scenario with “sufficient chip” capacity to run the domestic economy.  This is a highly unlikely scenario over the next few decades. Conversely, if either China or the US were to seize control of Taiwan, then the balance of power would shift irrevocably. 

Semiconductor and electronics executives are not voicing what is common knowledge within their companies and the industry.  Historically that has not been the case. In the 1990s to 2000s, even as the US chip industry was under attack from first from Japan and then Korea, executives from the US computer industry lobbied for Korean DRAM makers to compete aggressively with US makers, even as US chip makers went out of business.  

War is not an option, but we all fear miscalculations and over-reactions. Another concern is that the anti-Chinese sentiment in the US is only getting worse concurrent with its economic woes. And the reverse is true but not as severe. As a politician, it’s just too easy to blame China for your own economic and social woes. At the end of the day, the private sector needs to step up and voice for engagement, free and fair trade. The Chips Act will be a colossal waste of taxpayer money just as China’s government-led efforts have wasted even more of their tax-payers money.

Daniel Heyler is an award-winning technology analyst. Currently he is the portfolio manager of Isola Capital’s Global Technology Apha Fund. Heyler was ranked #1 for five consecutive years by Institutional Investor Magazine as a research analyst at Bank of America Merrill Lynch where he worked 16 years. Between 2016 and 2022, Heyler’s portfolio at Isola Capital generated a five-fold return and was up 18% in the first six months of 2023. Prior to Isola, Daniel was a financial consultant for TSMC for nine months, Aegis Custody, CloudAlpha, and eFusion capital, and a Professor of Practice at Hong Kong PolyTechnic University. He holds an MBA from the Kellogg School of Business as well as a BA from Bowdoin College and is fluent in Mandarin.