Tech denial strategy needs a rethink

Export controls sometimes may inadvertently accelerate the very capability-building they were designed to prevent

It is almost ironic to see the world’s most powerful economies using trade as a weapon. The United States bans advanced chips from reaching China. In response, China cuts off exports of gallium and germanium. The US restricts drone components, but China still leads the drone supply chain.

The Joe Biden administration created complex rules for AI diffusion, which the Donald Trump administration later rolled back. Each side keeps trying to block the other, but those being denied usually find a way around it.

This is a pattern with a structure that deserves attention, especially since most technology-denial strategies today rely on assumptions that the evidence does not support.

In a recent paper published on technology-denial regimes, this author introduced two frameworks to explain why these situations keep repeating. The first examines denial from the perspective of the country imposing it, weighing supply-chain control against coalition-building. The second looks at the targeted country’s ability to adapt, focusing on its own R&D strength and the partnerships it can form.

The first mistaken belief behind most denial regimes is that controlling supply means controlling outcomes. For a denial regime to work, two things must happen simultaneously: the imposing country needs enough supply chain control to make workarounds difficult, and it must bring enough countries on board to block alternative routes.

When both conditions are met, you get something like an ‘iron gate’ - similar to how the Nuclear Suppliers Group kept nuclear-weapons technology tightly controlled for decades. If only one condition is met, the barrier is far weaker.

China’s export controls on gallium and germanium illustrate this clearly.

China produces roughly 99 per cent of the world’s primary gallium, which looks like total dominance. But the controls were less effective than Beijing expected. The US found new suppliers in Canada, Japan, and Germany. The US’s gallium use not only held steady during the restrictions, it even grew slightly. Prices rose, but higher prices without a drop in consumption are only a partial victory. China had supply chain control but lacked international support, putting it in the ‘fragile bloc’: strong on paper, weak in practice.

Applying the same logic to the US semiconductor export controls raises harder questions. The restrictions coordinated with the Netherlands and Japan around ASML’s EUV lithography equipment are more sophisticated and closer to a genuine ‘iron gate’. But semiconductors differ from nuclear technology: they are not just military assets. They underpin almost every part of the modern economy, which makes it harder to hold the coalition together as the economic costs mount for the allies supposed to sustain these controls.

The second mistaken assumption concerns how well the targeted country adapts. Denial regimes typically assume the target will simply absorb the blow. Countries with strong domestic research and development and alternative partnerships — or ‘resilient transformers’ — often emerge stronger. China is the clearest example. It has top research institutions, a massive domestic market that can sustain local alternatives, and deepening ties with countries outside the Western tech bloc. This is very different from a country like North Korea, which has neither domestic capability nor willing partners. The right denial strategy depends entirely on which category the target falls into.

DeepSeek made this point sharply. A Chinese AI lab produced a frontier model using far less compute than Western equivalents, without relying on the most advanced Nvidia chips. Their efficiency gains suggested that restricting hardware access may actually be pushing Chinese AI development toward more resilient methods, not weaker ones. The US export controls may have inadvertently accelerated the very capability-building they were designed to prevent.

The drone industry follows a similar arc. Restrictions on DJI have not reduced its global dominance. Instead, allied militaries and commercial users have found workarounds, settled for inferior alternatives, or quietly continued using DJI equipment despite the bans. Ukraine’s battlefield use of Chinese-made drone components even caused a minor diplomatic headache. The denial regime created disruption but failed to generate credible alternatives fast enough to matter.

This is where the framework becomes practically useful. Before launching a denial regime, a country should answer two questions. First: Do we control enough of the supply chain, and can we assemble a coalition that genuinely blocks the target’s options? If not, expect leakage through third countries.

Second: How capable is the target of adapting? If its adaptive capacity is high, a focused and time-limited approach on the most critical chokepoints will outperform a broad ban that simply gives the target years to develop the skills you wanted to deny it.

The Biden administration’s AI diffusion rules ran into exactly this problem. They were complicated, drew strong objections from allies who felt unfairly ranked, and were nearly impossible to enforce given how flexible cloud computing is. The Trump administration’s revision of those rules confirmed the unsustainability of the original framework.

Yet the core challenge remains: How do you control access to a technology that is predominantly software, evolves rapidly, and is being open-sourced by private companies beyond any government’s reach?

The answer is that you probably cannot control it fully. You can slow things down at specific chokepoints, buy time to build local alternatives, and avoid sprawling denial regimes that end up stimulating exactly the innovation you hoped to suppress. The history of technology-denial is largely a history of strategic overreach — the assumption that blocking access would permanently weaken a rival, only for that rival to emerge more self-reliant and more competitive.

The real question was never whether to deny; it was what to deny, to whom, and for how long, before those denied simply build their own tools.

(The writer is a research analyst in the High-Tech Geopolitics
Programme at the Takshashila Institution)

Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH.