TSMC's Global-Not-Global Strategy Must End
TSMC's Global-Not-Global Strategy Must End
Explore within a 100-mile radius of central Taiwan and you’ll stumble across some of world’s most majestic mountains, breathtaking lakes and awe-inspiring valleys. You’ll also find 95% of the world’s most-advanced chipmaking.
While lacking the same postcard views as Jade Mountain or Sun-Moon Lake, Taiwan Semiconductor Manufacturing Co. is still a treasure. The company went from being the upstart of a government industrial think tank to the most crucial chip supplier in the world. But even as it’s grown into a $540 billion company, management has stubbornly kept all state-of-the-art manufacturing capacity at just three locations.
That needs to change. The time is over when it made sense to put all of TSMC’s fabs within an easy two-hour high-speed rail journey, allowing engineers to visit multiple sites in a day. The world and the electronics industry have evolved, and this chip giant is failing to keep up. Western allies are increasingly wary of having so much of their supply chains outside their control. Even temporary disruptions, such as a recent shortage of automotive chips, have world leaders talking about national security.
TSMC is the only company that makes the leading-edge circuits required by the likes of Apple Inc. and Qualcomm Inc. It controls more than half the made-to-order chip foundry market and has a lock on top-end technologies used in components for smartphones, servers, games consoles, cellphone base stations and even weapons systems. Rivals are as much as a decade behind and toil away on more mundane chips. But that edge could be quickly eroded by political concerns if governments insist that components be sourced locally and TSMC lacks the factories to do so, losing orders to competitors with fabs on the ground.
More than 90% of TSMC’s revenue comes from clients that are based outside of Taiwan. Yet more than 94% of production capacity is within that 100-mile radius. TSMC makes all its best stuff at just this handful of factories. That wasn’t a problem two decades ago, when more chipmakers did their own manufacturing. But now that fewer companies have fabs, and most rely on TSMC, the excessive concentration is no longer sustainable.
You’d be hard-pressed to find any company in any industry that operates at such a global scale at this level of geographic concentration. Even Foxconn Technology Group, another Taiwan behemoth that gets half its revenue from Apple, has facilities on all four corners of the planet.
What makes matters worse is management’s refusal to see the folly of maintaining this strategy or undertake to change it at a time when international relations put TSMC at the center of a geopolitical storm.
“Our center of R&D and majority of production lines will continue to be located in Taiwan,” Chief Executive Officer C.C. Wei told investors April 15. Beyond plans for a small factory in Arizona, “We currently have no further fab expansion plans in other areas such as Europe, but we do not rule out any possibility.”
Founder Morris Chang, known as the godfather of Taiwan’s chip industry, followed up a week later by pouring cold water on the notion of the United States as a base for semiconductor manufacturing. Speaking at a forum in Taipei, Chang talked up Taiwan’s engineering talent, convenient transport links and local management as reasons why the foundry business can only possibly succeed there. The US simply lacks the workforce and interest to develop advanced manufacturing, he was cited as saying, and its people prefer instead to focus on venture capital or marketing.
His comments about the lack of American talent for chip manufacturing are incongruous. The US produces managers and engineers with the requisite skills and education, plenty of whom work for TSMC, including Wei. Furthermore, Chang and his successor as chairman, Mark Liu, are both US citizens and graduates of universities there.
Intel CEO Pat Gelsinger is betting that Chang is wrong. With $20 billion earmarked for expansion over the next few years, the California-based chipmaker appears confident it can find the local talent needed to run the high-end fabs that will rival TSMC.
In reality, the Taiwan company is facing a sort of innovator’s dilemma — holding steadfast to a business model that worked for three decades, and refusing to recognize a changing landscape that may render the old ways redundant.
TSMC needs to recognize that there are significant economic and political forces pushing to have more semiconductors manufactured in Western nations. Taking over from Donald Trump, President Joe Biden is eager to return the US to its former chip glory and is seeking $37 billion from legislators to help fund that effort. Among the administration’s concerns is the possibility that China will attack Taiwan and put the supply of chips at risk.
Intel’s Gelsinger has tapped into the zeitgeist and played up the fear of a supply chain dependent on Asia — despite most nations being steadfast American allies — to push for greater government investment in the US chip sector. This would likely lead to subsidies for Intel.
European leaders are also putting on pressure. In January, Germany’s economy minister, Peter Altmaier, weighed in on the shortage of automotive chips by appealing directly to Taiwan’s government. Last week, European Commissioner Thierry Breton met with TSMC executives and followed up with a cryptic Tweet that implied the EU would form some kind of partnership to bring the Taiwanese company to its shores.
A company official noted that the meeting showed commitment to Europe and desire to support customers, but stopped well short of pledging to actually open a factory there.
A European fab would be logical. The company has a smattering of clients in the region, but more importantly, it’s the home base of some of TSMC’s most important suppliers, including Dutch equipment maker ASML.
Europe is already considering plans to go it alone on an advanced facility anyway, one which my colleague Alex Webb notes would be wildly misplaced. Yet if the EU is willing to write 30 billion euro ($36 billion) checks to buy its way back into the chip race, TSMC may as well jump aboard and offer to do it.
In the US, the company will need to do more than open a token factory in Arizona by 2024. That fab will have a monthly capacity of 20,000 wafers, equivalent to what's currently available from TSMC’s largest overseas plant — in Nanjing, China — representing a paltry 1.9% of the firm’s current global capacity.
An additional five Arizona fabs may be in the works, Reuters reported this week. This was always likely, but the company has yet to elucidate any plans.
Bluntly, management is happy and comfortable in Taiwan, where talent is plentiful and the supply chain mature. Moving a significant proportion of capacity overseas would be difficult. But surviving in a world that’s increasingly scared of geographic risk will be even more challenging.
It’s time now for TSMC to recognize this new world order, and commit publicly to being a part of it.