China vs the US: who will win the chip war?

There is a joke in Taipei that if China invades Taiwan, the best place to shelter will be in microchip factories, because they are the only places the People’s Liberation Army can’t afford to destroy. The country that controls advanced chips controls the future of technology – and Taiwan’s chip fabrication foundries (‘fabs’) are the finest in the world. Successful reunification between the mainland and its renegade province would give China a virtual monopoly over the most advanced fabs. Given Xi Jinping’s designs on Taiwan, it is no wonder that the US government is worried.

For this reason, in recent months the United States has taken various steps to thwart China’s attempts to make advanced semiconductors. New fabs are technically challenging to build, and eye-wateringly costly. A recent study by Boston Consulting found the cost of a large chip plant is now more than a next-generation aircraft carrier or a new nuclear power station. In Taiwan, Fab 18, which was built for the production of advanced chips, cost US$17.5 billion.

The US attack on China’s bid to dominate future technology is twofold. The first part of the strategy is to rebuild America’s share of the global chip fab business, which has fallen from a peak of 37 per cent in 1990 to just 12 per cent now. Last month, Joe Biden signed an Innovation and Competition Act with an extraordinary $52 billion to finance technological research and the fabrication of chips in the US. The EU is making similar noises, but America has made it a national priority.

It’s an attempt by Biden, supported by the Republican party, to reverse the hollowing out of US manufacturing – and is a policy that was first identified and pursued by President Trump. With bipartisan support, the new Act specifically aims to punish US companies that invest in China especially in areas where there is a struggle for supremacy, such as chipmaking.

Xi, clearly alarmed, has urged China to increase its technological self-reliance and to ‘accelerate the pace of legislation in the fields of digital economy, internet finance, artificial intelligence, big data, cloud computing etc’. America’s aim to disrupt China’s technological advance is being strongly supported by a private sector in the West that has become increasingly aware of the geopolitical risks of having such a high percentage of its global fab capacity based in Taiwan.

Intel has, in recent years, eased away from the semiconductor industry that it once dominated, but it is now in all-out catch-up mode. In January, the company poached Apple’s director of Mac system architecture and announced plans to produce system chips to compete with Apple’s top-of-the-line M1 series. Intel aims to regain its position as the world’s leading integrated design and fab company within three years.

But Intel’s battle is not just with Apple. It needs to take on the Taiwan Semiconductor Manufacturing Company, which has a 50 per cent global share in advanced chips and, even more importantly, a 92 per cent share in the most advanced five-nanometre chips. Under its new CEO Pat Gelsinger, Intel wants to become a leading contract chip maker and regain the upper hand.

As well as completing the building of a fab in Oregon, Intel is planning new fabs in Arizona and Ohio. Some $80 billion is also earmarked for Europe, with sites in Ireland, Germany and Italy being negotiated. It is a vast project, with mind-boggling sums involved. Some analysts doubt Intel’s ability to pull it off.

But while the first part of the US scheme is to match or exceed China’s state-funded model for hi-tech industries, the second part is more nefarious. China is desperate to join the ranks of the advanced chip manufacturers (which include Samsung in Korea). America’s plan is to prevent China from importing the machinery that enables the manufacture of advanced chips.

Key technologies for semiconductor manufacture include ultra-clean rooms, deposition, coating, etching, ionisation and packaging. But the ultimate determinant of advances is lithography – the fine printing of electronic circuits on silicon wafers. The most advanced chips are currently printed at a gauge of just five nanometres: that is, five billionths of a metre. (By way of comparison, the average human hair is 70,000 nanometres wide.)

At present there is only one company in the world that can make lithography machines to print wafers at the five-nanometres gauge. Based in a nondescript suburb of Eindhoven in the Netherlands, Advanced Semiconductor Materials Lithography (ASML) is perhaps the world’s least well-known hi-tech business. Yet it ranks just behind Shell as the fourth largest company in Europe.

Its highest-tech machines use a process called ‘extreme ultraviolet’ lithography, which makes them the only systems that can do lithography below 13.5 nanometres. The company produces approximately 50 machines a year, at $150 million a pop plus service contract. As a result, it owns that rare commodity – a market monopoly.

Last month, the US successfully put pressure on the Dutch government to block the export of such machines to China. And for good measure, America is now applying pressure on the Dutch not to export even the previous generation of machines to China.

Back home, Biden has also banned Nvidia – America’s global leader in graphic processor chips – from selling its latest high-end products to Chinese customers, who made up a large part of its customer base. This is quite a setback for Beijing’s ambitions in artificial intelligence: there are only a small number of companies whose processors are optimised to run the complex algorithms needed for AI. Some of the biggest customers of Nvidia’s A100 graphics processing units were the Chinese giants Alibaba, Tencent and Baidu, all providers of cloud computing services which can also be used to program AI applications.

Will Biden succeed in this attempt to throttle China’s technological ambitions? At the moment there is no indication that any other company has the knowhow to duplicate ASML’s extreme ultraviolet machines. Not only do they require design expertise, but their manufacture comprises a complex web of hi-tech subcontractors worldwide; replicating that would be remarkably difficult. But perhaps the main barrier for competitors is the billions of lines of software code which seem to double every four years.

However, the 60-year experience of the industry is that there are generational leaps in chip-manufacturing technology. By 2025 Intel is planning to enter what it has labelled the ‘angstrom era’ of the sub-nanometre chip. At this level, the etching process will need to be reduced to the width of single atoms. (An angstrom is one ten-billionth of a metre.) Intel has already placed a $300 million order for ASML’s next-generation machine.

As long as the Dutch play ball – and there is no reason to suppose they won’t – the US is likely to defeat China in the chip war, at least in the short term. But there are caveats. For one thing, China has a window of opportunity to invade Taiwan and take control of its chip factories before the bulk of new ones are transferred to America and Europe.

Secondly, China will keep trying to develop technologies of its own. It has already announced the development of a seven--nanometre chip using the last generation of lithography machines combined with various technological tricks, although experts doubt whether these chips could be harvested at an economic level of yield.

America’s Semiconductor Industry Association has concluded that, for the time being, ‘more advanced technology is still out of [China’s] reach’. Nevertheless, China has demonstrable world-class scientific ability. Candidates to replace ASML’s extreme ultraviolet techniques include the use of X-rays, electron beams, ion beams and nanoimprint lithography. China could get there first.

The possible unintended consequences of the chip war, and other technological wars, should be noted. If globalisation has brought enormous economic benefits to the world since Deng Xiaoping’s deregulation of the Chinese economy in the 1980s, de-globalisation, the retrenchment to national interest and autarky, is likely to bring the world slower growth, or worse. Both America and China will be losers in their chip war even if the end result is not a real global war.