When foreign companies want to buy advanced AI chips from the United States, they will soon face a choice: accept dictation from Washington or look elsewhere. That, at least, is what a sprawling 129-page draft regulation circulating through the Trump administration suggests.
The Commerce Department is preparing rules that would make US government approval mandatory for nearly all exports of AI chips, from small research projects to massive data centre buildouts. For large-scale deployments, foreign governments would need to guarantee investment in American AI infrastructure as the price of access. The approach treats semiconductors as geopolitical leverage, not as goods to be bought and sold in a functioning market.
A draft making its way through the government is the sixth iteration from the Commerce Department's Bureau of Industry and Security, and it has already generated tension within the Trump administration itself.The Commerce Department is moving ahead with draft rules to expand federal oversight of AI chip exports, but President Trump opposes any approach that mirrors former President Biden's restrictions. A White House official told reporters the draft "does not reflect what President Trump has said on export controls".
Yet here is where policy and politics diverge.Trump views semiconductor exports as a "bargaining chip and leverage in bilateral negotiations with other countries". The administration has already demonstrated this approach in the Middle East.UAE chip licenses were contingent on a $1-for-$1 investment condition. The new proposal simply extends that template globally.
The Tiered Shakedown
The proposed system would operate on a sliding scale.Shipments involving up to 1,000 Nvidia GB300 GPUs would pass through a simplified review and may qualify for limited exemptions. Orders between 1,000 and 200,000 chipswould require pre-approval from the U.S. Department of Commerce, as well as an export license. But cross the 200,000-unit threshold, and the rules shift fundamentally.For any deployment exceeding 200,000 GB300 GPUs by a single company in one country, that nation's government would need to make security commitments and matching investments in American AI infrastructure directly.
This targets the hyperscalers.Clusters powered by 200,000 GB300 GPUs operated by a single company within one country such as those currently deployed by AWS, Microsoft, Oracle, Open AI, or xAI would trigger direct intergovernmental arrangements. If Britain, France, or Germany want their tech giants to build frontier AI capability, they will need to negotiate with the White House and commit capital to American data centre buildouts.
That is not policy. That is coercion dressed in regulatory language.
Fiscal Cost and Market Distortion
U.S. export controls on semiconductor sales reduce U.S. chipmakers' revenues, and ITIF estimates that U.S. firms could lose about $77 billion in semiconductor industry sales in the initial year after a hypothetical, one-time full decoupling with China. Even without full decoupling, licensing uncertainty has real costs. Companies must navigate complex approval processes. Investors demand reassurance about supply continuity. Startups cannot afford the compliance burden.
Semiconductor export controls, in theory, protect national security. In practice, they are proving difficult to design fairly.The draft regulations are essentially "diffusion 2.0," an industry source said, referring to Biden's export control regime that industry decried as overly restrictive and harmful for U.S. competitiveness. Yet the Commerce Department denies any return to Biden-era thinking, claiming these rules represent a new approach built around bilateral negotiation rather than universal tiers.
The Counterargument Has Weight
The case for export controls is not without merit. Advanced AI chips represent genuine national security concerns."The rule could help the U.S. government address chip diversion to China and ensure a more secure buildout of the most powerful AI supercomputers," said Saif Khan, a former national security official in the Biden administration now at the Institute for Progress. Preventing authoritarian regimes from acquiring frontier-level computing power makes strategic sense.
The problem is scope.The license requirements are overly broad, applying globally, raising concerns that the administration intends to use the controls as negotiation leverage with allies rather than for security. Australia, Japan, and South Korea are security partners. They are not adversaries plotting chip diversion. Requiring their governments to invest in American infrastructure as the cost of market access treats them as hostages rather than allies.
For Australia specifically, the picture is mixed.Australian data centre operators anticipate advantages over regional competitors due to proposed US export controls on artificial intelligence chips, although concerns are raised regarding the sector's dependence on unpredictable US policy. Canberra benefits from privileged status in the current framework. But that status is conditional on political alignment, not economic need. When policy swings, reliance on US approval becomes a liability.
The Economic Reckoning
Uncertainty itself is a cost. Companies planning global infrastructure need confidence that chip supply will continue.For chip manufacturers, the planning question is no longer whether a permit regime arrives but how quickly. Those operating AI data centres need approval before GPUs can be ordered, and the process must begin well before any rule takes effect. That calculus pushes investment decisions toward the United States, which is the administration's goal. But it also reduces competition and innovation elsewhere.
The practical implication is that global AI development will increasingly depend on Washington approval. That is not a bug in this system; it is a feature. But it comes with a genuine cost: less distributed innovation, less competition, and ultimately a world where American policy uncertainty cascades into global economic risk.
Whether these draft rules become final law remains unclear. The White House has signalled resistance. But the mere fact that a Commerce Department has drafted such rules signals the administration's intent to use semiconductor access as leverage in bilateral negotiations. That intent, once acted upon, will reshape global data centre investment for years.
The fundamental tension is straightforward.The proposed policy is by no means a total export ban, but an extremely powerful tool that lets the U.S. government influence how global AI infrastructure develops. Fiscal responsibility and sound economic management require asking whether that influence is worth the cost in competitiveness, predictability, and allied trust.