TSMC’s Geopolitical Dance

Last week, it agreed to build an advanced foundry in the U.S.—at the same time that the U.S. sought to further restrict the foundry’s business with Huawei

4 min read
Photo of microchip held in tweezers with TSMC logo on a background.
Photo: Ascannio/Alamy

The global semiconductor supply chain is having an interesting year. Having adjusted to the potential and realities of a U.S.-China trade war, it is now faced with an economy-halting pandemic. Friday’s news seemed a microcosm of what is emerging from this moment: a combination of less concentrated advanced manufacturing and attempts to pressure companies to bend to geopolitical objectives.

On 15 May, the world’s largest semiconductor foundry, TSMC, announced that it planned to build a US $12 billion fab in Arizona, which would begin production in 2024. (The $12 billion investment is 2021–2029.) That same day, the Trump administration said it would now require TSMC and other non-U.S. chipmakers to get a license from the U.S. Commerce Department if they want to ship chips to Huawei that were made using U.S. software and technology.

First, the new fab: According to VLSI Research’s Dan Hutcheson, a U.S. fab is partly a ploy to keep Apple happy. The iPhone-etc.-maker’s CEO Tim Cook has been pushing for such a move for some time to ensure supply continuity for the processors that go in the company’s products. These processors have historically used leading-edge chipmaking technology. Currently that’s TSMC’s 7-nanometer process, but the company says the next generation process, 5 nm, is in production now.

TSMC, of course, has other important customers for its leading-edge technologies. AMD, Xilinx, Qualcomm, and Nvidia are among them; and more recently, so are cloud giants such as Google, Microsoft, Facebook, and Amazon, which have been developing their own server and AI designs.

To keep them happy, the Arizona fab will have to operate at the most advanced process available. TSMC is promising a 5-nm fab there, but by 2024 when production is set to begin, TSMC may be moving to another process generation, 3 nm. But fabs are built to be upgraded, Hutcheson points out. They aren’t built around a particular technology, and it seems assured that whatever 3 nm and more advanced processes entail, it will still mainly rely on extreme-ultraviolet lithography, the same tech central to the 7-nm and 5-nm processes.

However, transferring a manufacturing process to a new location and getting it to the point that it yields a profitable proportion of wafers is never easy. Hutcheson notes that TSMC struggled with that when it first built fabs in Tainan, which is little more than an hour away by high-speed rail from its headquarters in Hsinchu. However, depending on where in Arizona the fab is located, the company may benefit from infrastructure and experienced employees related to Intel’s advanced fabs in Chandler.

The plant’s projected 20,000-wafer-per-month capacity figure is actually quite low compared to that of other facilities. It matches the company’s recently built 16-nm Fab 16 in Nanjing, China. But it’s not in the same league as the company’s planned 5-nm Fab 18 in southern Taiwan, which will have a nameplate capacity of 120,000 wafers per month. Still, 20,000 wafers per month is in line with the first phase of other new fabs, says Joanne Itow, managing director at Semico Research. And that capacity could translate to 144 million applications processors per year, according to Itow. That’s enough to partly supply several customers and generate about $1.44 billion in revenue for TSMC.

That’s all assuming this fab actually happens. “Right now, it’s a Powerpoint fab,” says Hutcheson. TSMC’s own press release gives a very conditional feel: TSMC “announced its intention to build and operate an advanced semiconductor fab in the United States with the mutual understanding and commitment to support from the U.S. federal government and the State of Arizona.”

“Technically, it probably doesn’t matter where the chips are manufactured; however, in today’s tense trade arena the optics of having a fab in the United States provide a more positive partnership atmosphere,” says Joanne Itow, managing director at Semico Research.

The other TSMC news is much less of a win-win. The U.S. government has sought to starve Huawei of advanced semiconductors. Its Bureau of Industry and Security (BIS) added Huawei and its affiliates, particularly its semiconductor arm HiSilicon, to its list of entities that U.S. firms can’t sell to without a license in 2019. Huawei got around this by stepping up its own chip design capabilities, though it relies on foundries, especially TSMC, to manufacture its advanced chips. BIS is now seeking to tighten the screws by extending the licensing to foundries using U.S. software and tools to make Huawei’s chips.

In effect, the rule boils down to one country specifying which tools can be used in a factory in another country to produce goods for a customer in a third. TSMC is among the largest customers of U.S. chip tool makers and they have reason to worry, according to the Semiconductor Industry Association. “We are concerned this rule may create uncertainty and disruption for the global semiconductor supply chain, but it seems to be less damaging to the U.S. semiconductor industry than the very broad approaches previously considered,” the organization’s CEO John Neuffer said in a statement.

The new rule will likely accelerate Huawei’s ongoing shift away from U.S. technology, says Nelson Dong, a senior partner in charge of national security at the international law firm Dorsey & Whitney and a board member at the nonprofit advocacy group the National Committee on US-China Relations. Indirectly, “this move may well force the global semiconductor industry to look away from U.S. suppliers of semiconductor design tools and semiconductor production equipment and even to create new rival companies in other countries, including China itself,” he says. He cites the example of export restrictions in the satellite industry, which ultimately led to the growth of competing businesses outside the United States and higher prices for U.S. satellite makers due to their suppliers’ smaller market.

It is difficult to imagine how the U.S. could enforce such a rule in an advanced fab. “Fabs are an extreme version of ‘What happens in Vegas, stays in Vegas,’ ” quips Hutcheson. Manufacturing processes are proprietary and very closely guarded. “How would they even know it was going on?”

The Conversation (0)

A Circuit to Boost Battery Life

Digital low-dropout voltage regulators will save time, money, and power

11 min read
Image of a battery held sideways by pliers on each side.
Edmon de Haro

YOU'VE PROBABLY PLAYED hundreds, maybe thousands, of videos on your smartphone. But have you ever thought about what happens when you press “play”?

The instant you touch that little triangle, many things happen at once. In microseconds, idle compute cores on your phone's processor spring to life. As they do so, their voltages and clock frequencies shoot up to ensure that the video decompresses and displays without delay. Meanwhile, other cores, running tasks in the background, throttle down. Charge surges into the active cores' millions of transistors and slows to a trickle in the newly idled ones.

Keep Reading ↓ Show less