Counting the Costs of US-China Technology Decoupling

Much like the breathless World Cup final between Argentina and France with its dramatic twists and turns, the US-China technology decoupling saga has had its fair share of convolutions. But where the football analogy ends is that the latter cannot be resolved by penalty kicks but will likely be an interminable affair with ramifications extending far beyond two countries. Fortuitously, charting the full range of consequences of this multifaceted geopolitical controversy is now possible with an emerging research approach known as computational diplomacy. 

To grasp fully the value of such accounting, we need only highlight some notable developments in this still unfolding dispute involving the two superpowers. The first salvo was fired by the US when it banned Huawei’s technology, followed by its promulgation of the Creating Helpful Incentives to Produce Semiconductors (CHIPS) and Science Act in August 2022. Through the law, the US is effectively blocking exports of some chip manufacturing equipment and limiting sales of certain semiconductors to China. US companies that manufacture products in China are now unable to import chips for that purpose and must relocate their production. Meanwhile, China is seeking to develop its own chip design and fabrication capacity to circumvent these barriers. 

In the past week alone, three notable developments have occurred to deepen this rift between China and the US. First, the administration of US President Joe Biden extended its “entity list” to include another 36 Chinese companies. To sell specific technologies to circumscribed enterprises, American firms must first obtain hard-to-secure government permission. One immediate consequence was UK-based chip designer Arm being prohibited from selling its most advanced designs to Chinese firms such as tech giant Alibaba due to US and UK controls. 

Second, on December 15, China filed a complaint with the World Trade Organization (WTO) accusing the US of abusing its export controls on semiconductors and other related technology effectively to maintain “its leadership in science, technology, engineering and manufacturing sectors”. It added that American actions undermined “the stability of the global industrial supply chains”. The US strenuously denied these charges, justifying their measures by citing national security interests. (This came days after Washington rejected a WTO ruling against the US in challenges brought by China, Norway, Switzerland and Turkey to tariffs levied by the then administration of president Donald Trump in March 2018 on steel and aluminum imports on national security grounds.)

As these recent developments clearly illuminate, US-China decoupling involves more than the two protagonists. Instead, there are distinct implications for many affected economies, industries and companies. Taiwan is a critical player. Viewed by China as its sovereign territory, Taiwan’s status as global leader in semiconductor manufacturing is undisputed, a critical reason for why the US is intent on extending its protection to and guarantee of defense of the island – the vaunted “Silicon Shield” – in the event of an attack or invasion by mainland Chinese forces. Nearly all of the most advanced microprocessors worldwide are currently manufactured by a single company – Taiwan Semiconductor Manufacturing Company (TSMC) – with production solely domestic. 

With semiconductor chips now indispensable in all kinds of hardware linked by the Internet of Things, from mobile phones to smart cars, collaborative robots and military equipment, the timely supply of chips is a commercial and security issue of global proportions. More dramatically, the lack of access to the newest nano-scale chips will constitute a major impediment for the development of new strategic technologies in artificial intelligence (AI) and robotics all over the world, including in China and the US. 

By dint of its size, the semiconductor industry is far from straightforward and features many other companies scattered across diverse sectors with specializations in integrated circuit (IC) chip design, wafer fabrication and backend equipment manufacturing. The implications of the CHIPS and Science Act for this long tail of companies will play out in different ways with knock-on effects for yet other industry sectors.  

In its efforts to curb China’s access to this technology, the US government is essentially cutting its nose to spite its face. This action will hurt its own tech giants such as AMD and NVidia. Its hawkish approach of pushing ahead with unilateral controls and pressuring its allies to fall in line may also undermine the latter’s support in the long run. Even if the US can enjoy some near-term gains by hobbling China’s manufacturing sector, China could retaliate by seizing Taiwan by force, thus controlling such key entities as TSMC, though that would be predicated on the assumption that a mainland invasion would be able to secure and preserve the relevant facilities. 

The nascent field of computational diplomacy could offer a helpful perspective. Grounded in complexity theory and computational social science, it provides an operational toolkit for analyzing, modelling, and simulating geopolitical incidents. It entails mining data generated by diplomatic negotiations such as official correspondence, formal agreements, meeting transcripts, media reports and online campaigns. By trawling and making sense of large data sets, computational diplomacy throws up possible scenarios and simulates different courses of action for all contingencies considered. It leverages recent advances in data science, deep learning, and massive computations of highly complex systems to facilitate the quantitative forecasting of intricate international relations including the fraught US-China technology decoupling. 

By calculating the costs of different courses of action for any given scenario, and by identifying the key actors of interest – states, companies, individuals, non-governmental organizations, and other players – the full range of diplomatic moves could be robustly assessed just as AlphaGo, the computer program developed by Google’s DeepMind Technologies, did when it mastered the challenging board game Go.

Experiments of computational diplomacy are already ongoing in institutions such as the University of Geneva and ETH Zürich, which together established the interdisciplinary Lab for Science in Diplomacy (SiDLab) in collaboration with the independent Geneva Science and Diplomacy Anticipator (GESDA), a foundation focused on combining the anticipative power of science with diplomacy. Other natural hubs could emerge to conduct similar interdisciplinary research that marry insights from complexity science with political science and international relations. Critical nodes in the international diplomatic circuit come to mind including New York which hosts the United Nations, the Hague with the International Court of Justice (ICJ) and the International Criminal Court (ICC), or Singapore where the secretariat of the Asia Pacific Economic Cooperation (APEC) forum is located. These cities already have thriving ecosystems of universities, research institutes and think tanks that could embark on interdisciplinary research in computational diplomacy. 

While computational diplomacy is still a nascent field, the “proof of concept” for such an approach was demonstrated by scientists at Meta (parent company of Facebook) with their recent success with Cicero, the first AI agent to attain human-level performance in Diplomacy, a strategic board game released in 1959. Players must strike a fine balance between cooperation and competition with others to vanquish opponents. Cicero deployed natural language negotiation and tactical coordination skills that characterize diplomatic engagement. It made use of human-AI interactions and then outperformed human participants, ranking among the top 10 percent of those who played more than one game. 

Sceptics may argue that international relations are no board game. But as with many other complex systems, it would be helpful and most likely highly revealing to consider the prospects for computational methods in deriving deeper insights into intricate dependencies. The CHIPS and Science Act may trigger collateral damage and unanticipated consequences with an uncertain timeline. But the computing power supplied by these ubiquitous chips – along with analytical frameworks offered by computational diplomacy – could help us avoid these pitfalls.