The semiconductor cold war flared once more in mid-May with a volley of announcements by the U.S. aimed at curtailing the ability of China and Huawei Technologies Co. to tap American chip-making knowhow.
Given that the global economy runs on semiconductors, this is one of the most important issues in the world today.
On the same day that Taiwan Semiconductor Manufacturing Co.
said it will invest $12 billion in a fabrication plant in Arizona, the Commerce Department announced that any company selling chips to Huawei, an affiliate of the Chinese military, will require a license if the design and production process uses U.S. intellectual property, software or equipment.
Within three days, TSMC stopped taking new orders from Huawei, according to the Nikkei Asian Review. The No. 2 smartphone maker is believed to have been the fabricator’s second-biggest customer in fiscal 2019, accounting for about 14% of revenue and trailing only Apple Inc.’s
23%. (Read here, page 11.)
Through its HiSilicon Technologies unit, Huawei depends on TSMC to make processors for its mobile handsets, along with artificial intelligence chips and networking processors. It is far from unique in relying on TSMC, the largest contract chipmaker with control over more than half of the global foundry business.
‘A way to coexist’
The fact that almost 70% of semiconductors are manufactured in Taiwan or pass through the island during production is grounds for concern in Washington, and the disquiet increases when you factor in China’s refusal to recognize Taiwan’s sovereignty. During her May 20 inauguration for a second and final term as Taiwan president, Tsai Ing-wen said China and the island of almost 24 million people “have a duty to find a way to coexist over the long term.” The response from China’s Taiwan Affairs Office was that reunification “cannot be stopped by anyone or by any force.”
Beyond Huawei, the U.S. is also seeking to blunt the ability of Chinese chipmakers to catch up with their overseas counterparts. Shanghai-based Semiconductor Manufacturing International Co. has been trying to match TSMC’s capabilities for two decades with the help of the Chinese government and yet still remains five years behind.
The new export-control mandates make bridging that divide even harder, because advanced semiconductor design and production is mind-blowingly complex and impossible without certain tools controlled by U.S. and European companies. These include chip-design software from Cadence Design Systems Inc.
and Synopsys Inc.
both of which are American, or Mentor Graphics, a unit of Germany’s Siemens AG
Another essential element is fabrication equipment made by ASML Holding of the Netherlands — whose machines can each cost hundreds of millions of dollars — LAM Research Corp.
or KLA Corp.
both of the US.
None of this tech can be reverse engineered through typical industrial espionage techniques, so by thwarting China’s ambitions in this highly concentrated section of a critical supply chain, Taiwan’s fabrication capacity is placed even more squarely in Beijing’s crosshairs.
But as incendiary as these measures are, they ultimately fall short of sustainable, long-term risk reduction. TSMC’s fab plant in the U.S. won’t open until 2024, if at all, and will produce about 20,000 5-nanometer (5nm) chips a month. Compared with TSMC’s annual output of 12-13 million chips, the factory doesn’t represent a viable backup should an all-out confrontation over Taiwan’s sovereignty develop. Further, while 5nm technology is leading edge just now, it won’t be in four years — witness TSMC spending $20 billion on a 3nm plant that it plans to open in 2022 or early 2023.
One possible long-term solution would be for leading-edge semi designers such as Apple, Nvidia Corp.
and others to jointly fund a $100 billion foundry in the U.S., perhaps in conjunction with TSMC, on which they already depend almost exclusively for leading-edge fabrication. As tensions escalate, Apple’s over-reliance on China seems naïve at best and existentially devastating at worst.
According to the Wall Street Journal, Intel
is also in talks with the White House about building contract foundry capacity in the U.S., even though it has repeatedly failed in its third-party fabricator ambitions. While it has leading-edge fabrication in Arizona for its own use cases, Intel also relies on TSMC for autonomous driving, field programmable gate arrays (FPGA) and other non-X86 chips, about a fifth of its overall needs. Intel’s advanced chip-making capabilities have also fallen behind TSMC and South Korea’s Samsung, which is exploring how it might build a foundry without using U.S. equipment in a bid to circumvent Commerce Department edicts, according to EE Times, an electronics-industry magazine.
What’s key for investors to remember is that semis are at the epicenter of artificial intelligence, the Internet of Things, 5G mobile and cloud computing. Chips are the critical component of the digital economy and are only becoming more vital and valuable with accelerating profit pools and a reduction in cyclicality as the number of use cases expands.
There is no company or government in the world that doesn’t to some degree rely on TSMC-built Nvidia graphics processing units, Xilinx Inc.’s
FPGAs, mobile-phone chips by Qualcomm, CPUs by AMD
networking chips from Broadcom or many other mission-critical semiconductors.
Our power grids, air, land and sea transport infrastructures, communications and defense systems, computer and health-care networks, along with every other critical system you can think of, cannot function without these components. Our dependence on advanced semiconductors will only increase as the global digital transformation accelerates, and to maintain our advantage in the creation of leading-edge chips, it is crucial to secure the supply chain for the trillions of smart, always-connected devices that the shift will spawn.
Trouble is, China is thinking exactly the same thing.
Brad Slingerlend is the author of the SITALWeek newsletter and co-founder of NZS Capital, an investment company.