Hardly a week goes by now without a new development in the great US-China tech war. However, the outcome of this war hinges on a single critical industry: semiconductors. More specifically, it hinges practically exclusively on who makes them.
This is the true subtext behind recent escalating tensions over Taiwan: China's real goal is not to invade and occupy the island, but to control its world class chipmaking industry. Over a comparatively short period of time, Beijing intends to transfer the lion's share of industrially critical medium-end - and even some high-end - chipmaking from the self-ruled island to the mainland. This will give it absolute strategic advantage over the US as well as strategic dominance over Taiwan itself: rendering the ROC a geopolitical extension of the PRC that will cease to be of any use or value - but rather a growing liability - to Washington.
Ultimately, China's plan to absorb Taiwan - and force the US to accept the "One China Principle" as Beijing defines it - boils down to a simple scheme to force the island's semiconductor factories to obey the Communist Party's industrial directives as the price of avoiding wholesale hollowing out, i.e. mass migration to the mainland.
Taiwan currently dominates high-end semiconductor manufacturing: led by TSMC, the island's foundries (or "fabs") account for nearly two-thirds of the world's output of advanced chips - defined as being fabricated at or below the resolution of 7 nanometers (7nm). This has made the island the prize in a tug-of-war between Beijing and Washington for the crown jewel of the lifeblood of current and next-generation technology.
However, whereas the US is relying on antiquated legislation and trade negotiation - the CHIPS Act and Chips-4 Alliance (with Japan, South Korea, and Taiwan) - to assert control of the chip supply chain, China already enjoys actual physical control of it on the ground.
For all its heavy intellectual capital requirements that are still practically monopolized by US and US-allied companies - such as ASML, Lam Research, Synopsys, etc. - the chipmaking industry is just as beholden to upstream Chinese dominance of the primary elements extraction and processing industry as any other.
China supplies two-thirds of the world's silicon and three-fifths of its "rare earth" metals; between these dozen or so elements alone, Beijing essentially determines who gets to make how many chips - for whom - and at what price point.
The global semiconductor shortage since late 2020 has been a reflection of Chinese leverage over the ostensibly US-dominated global chip sector. It largely reflects a growing opacity in Beijing's dealings with the industry's top players - who even as they follow the letter of Washington's growing litany of export restrictions against Chinese chipmakers are becoming steadily more in thrall to the Communist regime's semiconductor localization policy.
Early in 2021, the fledgling Biden administration was forced by the growing chip supply crunch to temporarily grant Chinese firms like SMIC a license to import lithography machines - the centerpiece of the entire semiconductor fabrication process - capable of producing chips up to the 10nm sophistication level. In hindsight, this was enough of an opening for the Communist Party to secure the tools for its preferred domestic champions to rapidly climb the value-added ladder for the most complex commercially manufactured components of the 21st century.
In late July 2022, it was revealed that SMIC had been making bitcoin mining chips at the 7nm process node for a whole year - though not necessarily at full commercial scale, i.e. "yield." This was a total gamechanger: it indicated that the leading Chinese chipmaker had leapfrogged even Intel and Global Foundries of the US in establishing an initial proven 7nm capacity. But the backstory behind this great leap forward - and that's exactly what it is, even if China continues to allow speculation otherwise, i.e. that it's only a modest and incremental industrial upgrade - is far more significant.
In late 2020 and early 2021, as the Trump era's sanctions and restrictions on Chinese tech firms carried over into his successor's term, the semiconductor world was fixated on whether China could quickly develop its own alternatives to the lithography hardware and electronic design automation (EDA) software necessary to manufacture chips below the 40-45nm resolution. If the answer was no, so the thinking went, Beijing's tech ambitions would be set back years, possibly even a decade, by simple US export bans.
By mid-2021, however, there was little indication in public news releases that China had either made any commercially significant domestic supply chain breakthroughs or imported lithography machines capable of fabricating at even the 28nm node - let alone the 14nm or 7nm ones. Rather, the tech press was awash with reports of Chinese imports of older model tools - including second-hand ones - and noted that some of them were obtained at a hefty premium despite their age and maturity.
Looking back, however, the picture was already far more nuanced: China had begun, no later than the second quarter of 2021, to systematically assert full-spectrum upstream control of the chipmaking supply chain.
When in the summer of 2021 the chief executive of ASML, the Dutch company that dominates the market for the highest end lithography machines - the deep ultraviolet (DUV) immersion and extreme ultraviolet (EUV) varieties - publicly voiced concerns that China could upend the entire chipmaking equipment market within just a few years, the game was already up insofar as it concerned the industry.
What really happened over the course of 2021 was that China - which produced 359 billion integrated circuits that year, or virtually a billion a day - acquired a dominant position in the bottom half of the global annual $450 billion market. In typical Chinese fashion - as had happened in so many other industries over the preceding couple of decades - the mainland ruthlessly exploited its comparative advantages of labor, capital, and generous state subsidy to rapidly undercut its foreign competitors, specifically by muscling in on the vast home market.
Whereas TSMC and Samsung - the longstanding duopoly of high end chips - had largely stopped fabricating ICs above 28nm, this was not the case for their laggards such as UMC, Intel, and Global Foundries. They now found ferocious new competition at the 40nm, 45nm, 65nm, and 90nm nodes: yet the effect on the market as a whole would hardly be limited to just these lower segments.
In addition to silicon, rare earths, and other primary elements, advanced chipmaking requires - unsurprisingly - a lot of less advanced chips. Not only that, but even ASML still earns over 40 percent of its revenue from previous generation lithography machines - dry argon fluoride (ArF) DUV and krypton fluoride (KrF) devices. With China quietly collaborating with ASML's Japanese rivals Nikon and Canon - exclusively in the older lithography space - Beijing suddenly acquired the means to deny second-tier global chipmakers the very heart of their businesses while simultaneously threatening to wipe the floor from even the ASML monopoly.
This gave ASML little choice but to secretly begin selling its best DUV immersion machines to SMIC and (for all we might speculate at this point) possibly other Chinese fabs: this would've happened no later than June 2021 - as in July 2021 SMIC was already shipping basic 7nm chips for bitcoin mining application.
In fact, realistically ASML had already begun shipping DUV immersion lithography to SMIC in April or even March: it would've sent an initial small batch of just 5 or 6 units as a preliminary trial of sorts - under the mutual understanding that all beneficiaries would be confidential mainland clients whose purchases of SMIC 7nm product were to be tightly guarded by state authorities.
As of August 2022, therefore, it's reasonable to estimate - based on additional technical information released after the shocking July disclosure of SMIC's 7nm crypto processors - that the leading Chinese champion alone has been churning out no less than 10,000 to 20,000 such wafers monthly, albeit nearly all of them labeled under the 14nm or even 28nm process nodes (at least until now).
Additionally, it should now be apparent that no later than the end of 2021, when China's primary domestic supplier of lithography equipment, Shanghai Microelectronics (SMEE), was poised to unveil a homegrown 28nm DUV device - even a rudimentary dry ArF one - ASML came under enormous new pressure to stabilize its supply of DUV immersion machines to the mainland market. It has since been revealed that 23 ASML lithography devices were sold to the mainland in the first quarter of 2022 alone - presumably all or most being DUV ArF immersion, per SMIC's requirements as of second half 2021. The true number could well be much higher - and under-the-radar sales must be assumed to be ongoing to this day, especially after the early July ban announced by the Biden administration on DUV as well as EUV devices to mainland chipmakers (firmly opposed in public pushback by ASML).
For all we know, by early 2023 SMIC and a select group of other Chinese fabs may well have monthly 7nm capacity of 50,000 to 100,000 wafers - even if they continue labeling most of it 28nm or 14nm - which would seriously jeopardize TSMC and UMC, the two leading Taiwanese foundries, in their residual business above 7nm.
That would mean Taiwan - which has already been hit with sanctions on imports of mainland sand (a vital raw material in the chipmaking process, along with water) in the wake of the Pelosi visit - would face new shortages of chipmaking inputs just as competition with the mainland heats up. It already seems as though Beijing intends to penalize Taiwan's strengthening economic ties with the US - specifically the push for a free trade agreement - by engineering a semiconductor recession on the island.
In fact, if China really played hardball, it could promptly cut Taiwan's chip industry in half by mid decade - and shrink it by three-quarters by late decade. Presumably, the mainland would make up the entire difference in such a scenario. All that Taiwan - and the US, by extension - would be left with in this full-scale chip war would be a small slice of premium manufacturing of 5nm, 3nm, and future 2nm process nodes.
Not only would reducing Taiwan's share of chipmaking boost China's own: it would also render what capacity remains on the island even more dependent on both mainland suppliers and clients. This would allow Beijing to quietly hoard even more output of TSMC's 5nm and (soon to be mass rollout) 3nm chips than it realistically already has been.
With such a grim outlook for the very heart and pride of its tech-dependent economy, Taiwan will be under enormous strain to make major concessions to China to avoid rapid hollowing out. TSMC, for one, will face financial pressure to transfer its advanced 5nm and 3nm manufacturing to the mainland to retain any significant capacity at or above 7nm - which as of today still accounts for 70 to 80 percent of its revenues.
Coercively onshoring Taiwan's semiconductor industry is by far China's best strategy to simultaneously end the century-long US superiority in global technological leadership as well as absorb the island itself. So potentially effective is it, in fact, that Beijing is aggressively pursuing no less than three other avenues to achieve this strategic master stroke over the remainder of this decade: advanced chip packaging, silicon photonics, and new semiconducting materials.
With AI-enhanced 3D chip packaging, China will be able to multiply the performance of lower-end chips - often no narrower than the commoditized 40nm node - by stacking them up and combining their processing in efficient and power-saving configurations. While this increases the use of silicon itself, it reduces the cost of advanced chipset production by up to an order of magnitude. All chipmakers whose primary lines of business remain above 10nm may be existentially threatened if this new trend of 3D packaging takes off in as little as two to three years: especially if the massive server and cloud markets which rely on mature central processing (CPU) and general processing (GPU) units typically between 14nm and 10nm are replaced by multi-packaged 28nm alternatives that may ultimately breach 10nm or even 7nm performance equivalents.
Silicon photonics, meanwhile, holds out the opposite promise: much more high-throughput calculation per unit of chip. With revolutionary quantum mechanics applications, parallel processing for cryptographic and high-volume data crunching becomes feasible at much lower levels of power consumption. With Chinese firms and research institutions already practically neck-and-neck with global leaders in this field, mass production and deployment are a matter of less than five years.
Finally, China is investing heavily in R&D on new semiconducting materials: particularly graphene, gallium nitride, and silicon carbide. All hold potential to significantly boost performance per chip transistor - meaning even a 28nm carbon-based chip could match a 7nm silicon one in speed while conserving 5 to 10 times the power. If eventually combined with 3D packaging, the new semiconducting materials will offer an extraordinary boost to low-cost, high-performance, and low-energy computing.
With all the aforementioned initiatives, China's de facto control of Taiwan's semiconductor sector is already little more than a matter of time. And yet, it behooves us to understand just how and why both the US and Taiwan arrived at such a dire juncture - where Chinese dominance is already approaching so rapidly on the horizon.
The bottom line is that the Trump era US trade war with China has been nothing short of an all-round strategic catastrophe for both America and the global democratic order it underpins.
Had Trump struck a comprehensive deal with Beijing when he had the upper hand in the intensive negotiations of early-to-mid 2019, instead of going for an unrealistic total kill of China's state-led economic model - thereby expanding the trade war into a tech war, as well - China would've remained entirely beholden to US-controlled chipmaking supply chains to the present day, as it only gradually ascended the value-added semiconductor manufacturing ladder.
Instead, by attempting to wipe out 5G telecoms giant Huawei with blanket chip export sanctions, Washington tipped off to the Communist regime that American semiconductor supplies would thereafter all be at risk of cutoff - and had to be replaced with domestic or at least non-American alternatives as quickly as practicable.
From there, further US restrictions against SMIC ensured that China would accelerate its quest to establish a fully indigenous chipmaking industry chain, to boot: though this would've happened regardless, American efforts to deny China modern fabrication equipment pulled forward by at least several years an all-out scramble by the mainland to reduce and eliminate the bottleneck of foreign lithography equipment and EDA software; thereby forcing the lucrative chipmaking supplier industry to cater to the Chinese market much earlier than would've otherwise been the case - that is, for all but the uppermost tier of EUV lithography.
The net result, in less than three short years, is that the entire chipmaking supply chain is primed for a wave of commoditization driven by Chinese industrial policy on a scale and timeframe rivaling that of primary bulk manufacturing in the early 2000s - that is, in the wake of China's entry into the WTO and an injection of fresh international capital into its cumbersome state-owned enterprises (SOEs).
In other words: by the second half of the present decade, Taiwanese, South Korean, Japanese, and yes, even American, technological prowess and primacy will have been relegated to a thing of the past: the entire future of semiconductors and chipmaking will belong to mainland China.
But if anything, this terrifying trajectory alone will create panic - sooner rather than later - in Washington and Taipei. The specter of the end of US primacy and its replacement by a Chinese ascendancy in East Asia and the Western Pacific is so traumatizing to both that they're liable to overcompensate with new gestures of diplomatic and defense cooperation in the very near term that unequivocally cross Beijing's red lines on the One China Principle.
In that case, Taiwan's de facto independence can already be measured not in years or months, but mere weeks or even days.
While stepping up military blockade exercises again - perhaps to an even greater intensity than in early August - China will now threaten to commandeer the island's maritime commerce and freeze its supplies of semiconductors to the US: and to make the latter stick, Beijing will also brandish its "nuclear option" - a rare earth ban on American customers in both the military and civilian sectors.
This will force Washington to nix additional arms sales to the island and possibly even scale back the symbolic diplomatic visits by Congress members and state governors - which will finally be understood to have been a spectacular strategic blunder. The ensuing domestic political crisis in Taiwan, meanwhile, could well destroy Tsai Ingwen's pro-independence government - faced as it will be with a steep and probably irreversible recession.
Taiwan will be absorbed - and America will have been replaced as the leading superpower - not so much in fact, but in sheer psychological intimidation; years ahead of schedule, one should add.