The US Commerce Department is preparing an interim rule that extends AI-chip export controls to remote computing services. Exporters would need licenses to send or transfer chips and servers to data centers anywhere in the world when that infrastructure may provide compute to companies in China, Russia, Iran, North Korea or other countries in the D:5 embargo group.
The policy is intended to close a route around existing controls: restricted companies can rent capacity from hardware physically located in another country. It also covers indirect access through brokers and requires exporters and data-center operators to examine customer ownership and activity.
The threshold applies to networks able to supply more than 12 million total processing performance, an export-control measure that TechCrunch estimates at roughly the maximum compute of 750 Nvidia H100 or H200 chips. Existing operators covered by the rule would have 90 days after it takes effect to obtain licenses. Commercial recommendation systems used for services such as shopping and social media are excluded.
Additional licensing requirements apply to advanced-chip transfers to Malaysia, Singapore, Indonesia, Thailand and Brazil because of diversion concerns, though US-company destinations are exempt. The rule also withdraws the broader 2025 AI Diffusion Rule, replacing country-level limits with controls focused on restricted users and remote access.