Say what you will about the original Cold War, the looming threat of annihilation and the physical border infrastructure that split a continent in two: it was at least relatively easy to disentangle the two main parties. Trade between the Soviet Union and the United States was minimal and movement of people rarer still.
The struggle for dominance between China and America is far less clear cut, and attempts to make it more so have proved difficult at best.
As Donald Trump’s decision to postpone the imposition tariffs – and Treasury secretary Scott Bessent’s indication of a likely further deadline extension – have demonstrated, trade between the two powers has become deeply entrenched no matter the policy preferences of those in the White House, or indeed the state of competition between the nations.
As Elbridge Colby, now US under secretary of defence for policy, wrote in these pages in 2023, America’s priority is making sure it can “defeat China in a conflict”.
1004 Trade volumes
Norman Angell’s 1911 thesis that interdependence could make war economically irrational was superseded by 1914’s demonstration that wars are not necessarily rooted in economic logic, but the coexistence of increasing military risk with large trade flows is curious, and all the more so given the technological arms race taking place in the background.
Even focusing on just two fields – AI and quantum computing – the scale of the risk is considerable.
Getting to sufficiently powerful AI systems first would be highly economically destabilising, permit massive parallel cyberattacks and information operations efforts, smooth out the operation of domestic systems, and potentially offer a pathway to self-improving systems that might make this advantage permanent.
Quantum computing, meanwhile, would undermine the encryption techniques which keep data and systems – from healthcare to banking to messaging apps – secure.
As a helpful note from the US National Cybersecurity Centre of Excellence sets out, unnamed “adversaries” are deploying a “harvest now, decrypt later” strategy against America, gathering sensitive data today to decode in the near future when viable quantum computing technologies are developed.
Investments into quantum technologies are well underway, with $15bn (£11bn) in Chinese state funding announced prior to 2023, against roughly $3.8bn in the United States.
Even allowing for a gap between headline figures and reality, and a $1bn-plus private sector investment gap in favour of the US, Beijing is at the least competitive in funding, and looking to consolidate efforts: private sector firms previously conducting quantum research have in recent years transferred assets to “state-linked institutions”.
The risks and advantages of powerful AI systems, meanwhile, have been discussed at length elsewhere.
The sight of US firms inching away from sensitive fields in the Chinese market – with Amazon, IBM and Microsoft reducing their presence in the country – is at least an indicator that some private sector firms are uneasy about the coexistence of an intensive race for a technological edge alongside economic cooperation.
As much as AI safety advocates might hate the “arms race” terminology, it still seems to be an accurate description of the world as seen from Washington: JD Vance explicitly referred to an AI arms race with China earlier this year, noting that any unilateral pause in research risking finding “ourselves all enslaved to PRC-mediated AI”.
JD Vance explicitly referred to an AI arms race with China earlier this year – Yasin Ozturk/Anadolu via Getty Images
You don’t need to believe in superintelligence, however, to see risks: competent coding agents that work tirelessly, day and night, year-round, in massively parallelised fashion, would confer a deeply intimidating cyber-offensive edge to the Chinese government.
The description of the neutron bomb as a weapon that kills people and leaves buildings standing was captivating. Weapons that shut off capabilities without destroying either would be better still.
As a Rand report argues, the People’s Liberation Army views “warfare as a confrontation between opposing operational systems rather than between units, arms, services and platforms”; paralyse the command structure, shut down information gathering and distribution, and disrupt infrastructure.
It’s not particularly hard to see how asymmetric cyberwarfare or one-sided decryption of communications would fit into this model.
If there’s a pattern to recent Chinese cyberoperations, with malware attacks on US bases in Guam and civilian infrastructure, it’s obtaining the “ability to physically wreak havoc on [US] critical infrastructure at a time of its choosing”, in the words of then-FBI director Christopher Wray.
Selective US retrenchment should be seen in this light; moving sensitive areas of technology into relatively safe harbours. Whether it will work, however, is a little less clear.
Fifteen years ago, an agency within the US Department of Defence produced a report noting that one of the most distinctive features of Chinese intelligence operations was “ethnic targeting” of Chinese Americans and appeals to “residual patriotism towards China”.
Most Chinese spies in America, at that time, hadn’t been bribed or blackmailed. Instead, they were simply profiled by ethnic background, and approached based on obligation to friends, family and nation.
This approach had formed the basis for roughly 98pc of the Chinese security service’s recruitment efforts, and “divided loyalty” was the primary motivation in roughly 57pc of espionage cases.
Given that US AI firms are increasingly “dependent on China for top-tier AI researchers”, merely shifting location may not be enough to guarantee the security of technologies developed by America’s private sector tech giants. And that brings us, again, to the strength of the economic ties between the two rival powers.
It’s one thing to start winding down economic operations overseas or to start developing alternative supply chains. But what do you do if your fundamental research capabilities are increasingly dependent on bringing in citizens from your rival power?
OpenAI already appears to be wrestling with this question, imposing “controls on sensitive information” and undertaking “enhanced vetting of staff”, which in turn may make hiring researchers harder.
The result seems to be an asymmetry that favours China, with the flow of information the other way appearing relatively limited. If America wants to win the race – and retain its lead – it will need to find a way to close its talent gap, or shut off the leaks.
