Bold shift: a fundamental shift in who protects national security and who profits from it—and the result could redefine global tech rules. Now, the United States has authorized the sale of Nvidia’s high-end H200 AI chips to China, overturning a prior export ban. In exchange, Washington will claim 25% of the sales proceeds, a striking example of the administration treating private-sector revenue as a lever in diplomacy. The H200 sits as Nvidia’s second most powerful AI processor, boasting roughly six times the capability of the H20 chips that China could previously obtain. These aren’t gadgets for fun online games or quiz nights; they are the engines behind increasingly sophisticated AI systems that powers autonomous weapons: drones, automated gun systems, and targeting software. In other words, technologies that tilt the balance of modern warfare rather than fanciful science-fiction devices. Reports point to AI-powered targeting systems already deployed in conflict zones, underscoring a present-day reality rather than a distant future.
This decision comes after a year of intense US-China frictions, including a tariff war that saw duties spike to 145% at times. The reversal to permit exports of sensitive AI hardware is, to many observers, a watershed moment that challenges conventional export-control logic and forces allies to rethink their own positions. Australia, for example, is caught between economic links to China and a deeper defence alignment with the United States, highlighting how interconnected security and trade now sit.
How we reached this crossroads
Advanced semiconductors are now a critical front in the AI race. In October 2022, the Biden administration imposed stringent export controls aimed at curbing China’s access to cutting-edge AI chips and the machinery used to produce them. The “small yard, high fence” strategy sought to cap a narrow slice of technology while keeping broader trade channels open. The administration blacklisted around 140 Chinese entities and restricted 24 types of manufacturing equipment, while also restricting US engineers from supporting some Chinese chip facilities. These moves had tangible effects: from 2022 to 2024, Chinese AI firms faced hurdles in securing the computing power they needed and had to push forward with older hardware.
A very different approach
Trump is pursuing a markedly different path. Earlier in the year, his administration allowed Nvidia to sell H20 chips to China in exchange for a 15% revenue share, a concession seen in many circles as linked to broader negotiations over rare-earth minerals. Now, with the H200—much more capable—being cleared for export, the stance appears to have shifted from restraining technology to treating it as a tradable asset in negotiations. The result is a pragmatic, but controversial, pivot away from a rules-based framework toward transactional diplomacy.
AI in warfare today
Chips that power AI are already embedded in modern combat systems. They help guide munitions, optimize targeting, and enable rapid decision-making on the battlefield. In Ukraine, AI-enabled drones reportedly achieved higher strike accuracy by navigating difficult environments. In other conflicts, organizations have noted how AI systems can rapidly identify potential targets, sometimes raising civilian risk concerns. The PLA is also exploring AI for drone swarms and on-the-spot battlefield decisions, while projects like the Pentagon’s Maven aggregate sensor data to propose operations. This is not speculative fiction—it is active military technology in use now.
The new “laundering” of dual-use tech
Semiconductors are inherently dual-use: the same chips that train AI models can guide weapons systems, and the microcontrollers in everyday appliances can enable autonomous devices. Investigations have documented how foreign components appear in weapons and how, in some cases, parts are sourced from consumer electronics and repurposed for military means. The risk of circumventing export controls is real, as makers of chips travel through multiple countries to reach end users, complicating enforcement and creating loopholes that harden more than they soften.
Australia’s position
Australia’s involvement with the AUKUS security pact has reshaped its export-control regime to align with U.S. priorities. Yet the country remains heavily tied to China in trade while facing mounting pressure to synchronize defence and policy with Washington. With the US signaling a more flexible approach to rare-earths and advanced chips, Australia faces a strategic choice: adhere to a stricter, predictable framework or accept a more transactional dynamic that could undercut sovereignty and long-term security planning.
When export controls turn transactional
Export controls work best when they are stable, predictable, and clearly tied to security interests. If they start trading away security for short-term revenue or leverage, trust erodes. The H200 deal represents that kind of shift, turning the “high fence” around sensitive tech into a gate that opens for the right price. For Australia—and for other allied nations—the question is whether aligning with U.S. export policy serves their own security and economic interests, or if they risk surrendering some degree of sovereignty to a partner whose policies can feel arbitrary and transactional. Would you align with a system where national-security tools can be monetized? How should allies balance economic interests with strategic autonomy in a rapidly evolving tech landscape?
