Limits of Economic Deterrence in the US-China Tech Competition

A week before reaching a deal, US President Donald Trump said he wanted $500 billion worth of Ukraine’s critical minerals as compensation for having supported the country following the full-scale invasion by Russia in February 2022. The offer, then rejected by Ukraine’s president Volodymyr Zelensky, was followed by threats to cut the country’s access to Starlink’s satellite communications system if it did not reach a deal. At the core of the US government’s pressure lays, among other things, its desire to diminish US dependency on the country that holds almost half of the world’s critical minerals reserves essential for tech development: China.

Only 12 days following the inauguration of Trump’s second presidency, the administration announced a fresh set of 10% tariffs on imports from China—shortly thereafter increased by an extra 10%. China followed with a retaliatory set of measures including antitrust probes into US tech companies, 10-15% tariffs on farm products, coal, crude oil and farm equipment, and expansion of export controls on critical minerals that are essential in producing everything from smartphones to F-35s and solar panels.

Trump’s latest rush to secure critical minerals in Ukraine, a country that currently does not produce them (despite having them) and cannot ensure easy access to them during an ongoing war, raises the question of how ready they are to deal with pressures from their trade war with China – particularly in realms of high technology. How quickly and effectively can they respond to, and anticipate changes to critical supply chains feeding key sectors such as defence and technology? How effective have US economic deterrence measures been to stop Chinese tech?

China’s Response: A Full-Spectrum Digital Competitor

Washington is grappling with the fact that China has emerged as a full-spectrum digital competitor. It has a commanding position in the global manufacturing chain for digital goods and dominates the production and refinement of rare-earth materials and critical minerals.

Despite attempts from the US to stop Chinese access to critical technologies in AI supply chains, both Chinese companies and the government have taken action to respond. DeepSeek’s R1 model release and low training cost has shown that China’s tech company ecosystem could become more quickly and effectively resilient to export controls than expected. After years of having Apple’s iOS and Android as the main operating systems in China, a homegrown one, Huawei’s HarmonyOS, is finally starting to capture a significant slice of the Chinese market, perhaps providing a basis for future internationalisation. Chinese companies have also sought to find other ways to secure access to high-end technologies, especially chips – either rushing to buy high-end chips prior to the enforcement of export controls and tariff or curbing restrictions to access NVIDIA-banned chips (A100 and H100) through a growing underground market.

In contrast to even the recent past, the Chinese government is responding far more quickly and with more alacrity. Within days of US tariffs being announced by Trump, China swiftly announced expanded export controls on critical rare earths and minerals, and retaliatory tariffs on a range of US imports, including coal, oil and gas products, as well as large-engine cars.

Furthermore. Chinese regulators are taking targeted measures against US businesses as well: competition regulators have launched an investigation into Google’s Android operating system as well as into US chipmaker giant NVIDIA over violations of anti-monopoly laws, and are reportedly considering probing Apple’s App Store. Not only has China’s Ministry of Commerce added more companies to its Unreliable Entity List – a countersanctions tool established right after the US 2019 addition of Huawei to its own Entity List – but it is going beyond its previous, symbolic approach. Hitherto, all listed businesses were US defence companies with little or no business operations involving China anyway. Now, apparel conglomerate PVH and biotech company Illumina have been added to the list.

The Limitations of US Economic Tech Deterrence

Sanctions tend to have two potential objectives: denying certain capabilities to their targets and deterring them from certain conduct. US tech sanctions against China have, over the past few years, expanded from merely targeting companies such as Huawei and ZTE to covering the entirety of the country, and from a relatively limited range of high-end semiconductors to an ever broader scope that by now has come to encompass sales and distribution channels, as well as equipment manufacturers in a range of third countries.

Whether those sanctions will achieve their primary goal, of putting a hard ceiling on Chinese semiconductor development and thereby the country’s ability to progress in cutting-edge applications such as AI, is subject to vociferous debate. Huawei has been able to advance the development of its mobile phone chips in the absence of access to advanced manufacturing facilities and equipment, albeit more slowly and at greater cost. Despite the labelling of DeepSeek’s R1 as a national security risk and OpenAI’s reference to it having been trained by ‘distillation’ practices, China has since mounted a nationwide push for adoption across local governments, public services and even private companies such as Tencent – which have deployed DeepSeek as part of the famous WeChat app. International companies are adopting the technology too: DeepSeek’s AI models are now available through Amazon Web Services.

Any demonstration of competence from Chinese firms or government authorities often meets with the response that they must have been cheating one way or another, whether by stealing intellectual property, bending international trade rules or, in this case, smuggling semiconductors. There is, of course, justification for those assertions, but at the same time, both the determination of Chinese policymakers and the growing ability of scientists, researchers, engineers and businesspeople to succeed should not be underestimated.

Both Biden and Trump administrations have pursued a policy of economic deterrence using different tools in the toolbox such as sanctions, export controls, tariffs, among others. To the extent that these tools are meant to deny, reroute, and contain Chinese tech prominence, the results are still quite limited. They seem even to have been counterproductive, in the sense that they have galvanized an ecosystem of Chinese businesses to finally come on board with policy initiatives to indigenize high-end technology capabilities. The emergence of DeepSeek is one illustration of this, as well as of another trend that merits attention: the constraints that China faces, including a relative shortage of funds and the aforementioned technology sanctions. These incentivise developers to explore alternative technological paths that, at least in the case of DeepSeek, have led to a product that is closely comparable to its American counterparts, but far more efficient in terms of inference and training costs. As seen in this case, constraints can enable efficiency and innovation in model training as much as disable.

At the policy level, technological progress is a key element of government’s policy, both to realise its development objectives at home, as to better compete and defend itself on the global stage against a United States that Beijing (not wholly unjustified) perceives to be fundamentally opposed to its attaining near-peer status. Attempts to deter China from those objectives will probably strengthen resolve to achieve them.

US domestic challenges

In the meantime, US decision makers must also start considering the growing ramifications of sanctions against China on their domestic politics, as their cost will increasingly start hitting home. TikTok provides an illustrative example: it is one of the most popular apps among the US population, a significant part of which is less concerned about great power politics. Spotting a political opportunity with that broader part of the US population, Donald Trump provided the company with a stay of execution, delaying the application of the law mandating its divestiture or ban for 75 days.

But TikTok is far from the only dossier with a domestic impact. Both Biden and Trump administration have sought to boost domestic AI supply chain production in the US to diminish reliance on China and other countries by boosting and relocating fabs domestically – but strategies have so far been incremental and with mixed results.

Former president Biden’s CHIPS Act sought to address Taiwanese semiconductor manufacturing chokepoint by encouraging TSMC to relocate to Phoenix, Arizona, but has fallen short of its promises. Firstly, production costs will remain at least 50% higher when fabs open following substantial delays. Secondly, the lack of US workforce skills for chip manufacturing can endanger the pace and sustainability of TSMC US. Thirdly, in seeking to onshore to lower the risks of a semiconductor supply chain crisis in the event of a Chinese invasion of Taiwan, the US also lowered the strategic protection that market concentration provides to Taiwan – as any small disruption to TSMC as a result of a China-driven blockade, for example, could be catastrophic for the global economy, including for China.

Domestic chipmaking industry returns on investment will take longer than expected and there is a risk that Trump’s volatile policymaking can further slow the pace of AI supply chain economic growth and robustness. While, on the one hand, Trump recently announced Stargate – a $100 billion private joint venture led by OpenAI, Oracle, SoftBank and others – to build AI infrastructure in Texas, on the other he has threatened to end the CHIPS Act after a massive onshoring effort of TSMC – who relies on the North American market for 70% of the revenue.

Contradictory policies come at a time where the US might not be able to afford miscalculations in the AI competition. In 2022, the US represented less than 1% of the global fabrication capacity in advanced logic chip manufacturing. Even though there is an expectation of it growing up to 28% by 2032, the prospects look tricky. US chipmaker Intel Corp has been struggling to maintain its side of the bargain to boost domestic semiconductor manufacturing after a drop in revenue in the fourth quarter and the abrupt departure of its CEO in December.

Moreover, Trump’s overreliance on tariffs is a short-lived strategy that could further exacerbate the inflation and cost-of-living issues that did much to bring him back to the White House in the first place. Following tariffs on Canada, Mexico and China, it is expected that the average middle class US household annual costs could rise anywhere from $2,500 to $3,900—something that is arguably not ‘America first’.

Secondly, the use of tariffs as a political tool since Trump’s first presidency has reflected on the fall China’s total share of total US imports – making it more prepared to endure additional tariffs and readapt supply chains.

Thirdly, tariffs will continue to create disincentives for global supply chains and increase prices of goods in the U.S. Given that consumer electronics amount for 27% of China exports, the US will still be selective and potentially reluctant to target that sector. Moreover, as mentioned, diversification of imports away from China does not mean less exposure to Chinese industries. Since the early days of the trade war, in 2018, Chinese manufacturers have moved production lines to Vietnam – a country that has been largely benefitting from the rerouting of US imports – showing that derisking goes beyond reduction of direct reliance but also includes indirect reliance through other trading partners.

The road ahead

If Trump’s hawkish approach to China persists, it might backfire or lose steam, if not backed by a well-calibrated assessment of the consequences domestically and internationally. In his previous administration, US pressure on EU member states concerning China led the bloc to name the country as a ‘systemic rival’ back in 2019. Threats to impose tariffs on EU goods, and the US decision to initiate ‘peace talks’ with Russia, have been far from well-received in Brussels. The regional bloc does not have a unified position on China, with some, such as Spain, favouring a pragmatic approach: ‘Europe must make its own decisions, and on its own. And we have to decide when China can be a partner and when China is a competitor’.

Ultimately, the key problem of the US deterrence strategy against China’s emergence as a tech power is that it is maximalist, and therefore unattainable. It includes objectives such as preventing harmful uses of technologies, but also capping China’s further economic growth potential. As China itself faces a critical juncture in its development path, it will double down on technological innovation as an engine for future prosperity. It may be difficult for people who spent their entire careers in the post-Cold War environment to accept, but it seems the notion of US primacy in the technological field is impossible to sustain. The practical point here is that economic deterrence is not a silver bullet nor a sustainable long-term strategy domestically or internationally if not followed by other actions.

At some point, Washington must accept the existence of China as a near-peer power. Isolating it can only increase its independence form US-linked tech supply chains (or move towards indirect dependence) and stimulate China to work towards a more effective strategy with Global South countries as done in infrastructure development through the Belt and Road Initiative, for example. As argued by Australia’s former cyber and tech Ambassador, if China ‘redefines cost structures by offering AI at a fraction of Western subscription costs’ then DeepSeek and the likes will disrupt AI business models and market expectations. In other words, the goal of US technology policy should not be to deny technology wholesale, but to recognise that technology will be diffused anyway, calibrate long-term economic deterrence expectations, and to collaborate in curbing harmful uses of some of these technologies – especially the average consumer-facing AI models.

© RUSI, 2025.

The views expressed in this Commentary are the authors’, and do not represent those of RUSI or any other institution.

For terms of use, see Website Ts&Cs of Use.

Have an idea for a Commentary you’d like to write for us? Send a short pitch to commentaries@rusi.org and we’ll get back to you if it fits into our research interests. View full guidelines for contributors.