US-China chip export debate highlights risks for AI leadership

Techday NZ

07-06-2025

DeepSeek. TikTok. Taiwan. And a White House shake-up on AI rules. The spiralling US-China technology rivalry landed at the heart of Johns Hopkins University last week, as a panel of top experts and policymakers took to the stage to debate whether restricting exports of advanced semiconductors to China can help the US maintain its edge in the race for artificial intelligence.
The discussion, hosted by Open to Debate in partnership with the SNF Agora Institute, comes at a critical time. In Washington, the Trump administration has announced plans to roll back the Biden-era AI Diffusion Rule and introduce new chip export controls targeting China – a move seen by many as a signal that the technology contest between the two superpowers is only intensifying.
On one side of the Johns Hopkins debate were Lindsay Gorman, managing director at the German Marshall Fund's Technology Program, and former CIA officer and congressman Will Hurd. They argued the answer is yes: semiconductor controls can give the US a real advantage in the AI race.
Gorman pointed to DeepSeek, a Chinese AI model whose CEO has publicly lamented the impact of advanced chip bans. "Money has never been the problem for us. Bans on shipments of advanced chips are the problem. And they have to consume twice the power to achieve the same results," she quoted, highlighting how China's AI advances still depend heavily on imported hardware.
"The United States has significant hard computing power advantages – the ability to produce high-end chips, designed specifically for training AI models," Gorman told the audience. She argued that, together with its allies, the US controls a "strategic choke point" on computing power. "Properly implemented controls can have an effect and also have an increasing and compounding effect over time in retarding China's AI advantages and giving the United States a head start," she explained.
Will Hurd, who also served on OpenAI's board before running for US president, compared the AI contest to the nuclear arms race. "Artificial intelligence is the equivalent of nuclear fission. Nuclear fission controlled gives you nuclear power… uncontrolled, nuclear weapons can kill everybody," he said. Hurd emphasised the importance of first-mover advantage, warning that the US cannot afford to lose its technological lead.
He also highlighted a lack of reciprocity in the tech relationship between the two countries. "Chinese companies like Baidu, DJI, and TikTok operate freely in the US, but American companies are not allowed to operate in China," Hurd pointed out. "If there was a level of reciprocity between our two countries, we wouldn't be here having this debate about chip controls."
Yet, on the opposing side, former senior US diplomat Susan Thornton and technology strategist Paul Triolo insisted the US could not outpace China in AI simply by tightening export controls. Triolo argued that the controls are "not working and will not lead to US dominance in AI", describing them as a blunt instrument that creates confusion for industry and disrupts global supply chains. "Most experts believe that Chinese companies are only three months behind US leaders in developing advanced AI models," Triolo said, suggesting any technological gap is vanishingly slim.
Thornton, who spent decades at the heart of US-China diplomacy, warned of unintended consequences. "The main thing we should be asking ourselves about this question… is what is the cost benefit of US policy actions?" she said. "We have to face the reality that China is already building AI… a third of the world's top AI scientists are Chinese. China is one third of the entire global technology market. So it's clearly a player."
She cautioned that blocking China from critical technology could backfire, hurting US companies, alienating allies and raising the risks around Taiwan, the global centre of advanced chip manufacturing. "Certainly, the one thing we need to do is avoid going to war," Thornton warned. "Taiwan, the most sensitive issue in US-China relations, has now been dragged right into the middle of this AI issue because they're the place that produces all the cutting-edge chips that we're trying to control."
Audience members pressed the panel on whether international collaboration on AI safety was possible, and whether the US could ever match China's data advantage, given the size of the Chinese population and its permissive data environment. Hurd conceded that "the US will always have less data because we have a little thing called civil liberties," but argued that superior algorithms and privacy-protective machine learning could level the playing field.
For Triolo, the dynamic nature of the technology means that attempts to wall off China are self-defeating. "There are many ways to get to different ends. The controls have forced Chinese companies to work together, develop innovations, and become more competitive both domestically and globally," he said.
Gorman, in closing, rejected what she called "a defeatism that says America can't out-compete China or slow its progress". "Our companies are doing well. There isn't an issue here with demand, it's with supply. Doing better means that we have to throw what we can at this problem now with a smart application of tools," she argued.
But Thornton had the last word, urging caution. "Making the AI competition with China a zero-sum game, not only will not work, it is dangerous," she said. "We should focus on the things that are going to matter to our children and their children, which is the long-term AI competition, which if not constrained and bounded by international agreements and by cooperation among countries… it'll be a very dangerous world."