An investigation traces how Shanghai-based AI startup INF Tech accessed advanced Nvidia chips at an Indosat data center in Jakarta, despite US export controls
and how thin our track record of enforcement. If we're going to keep relying so heavily on these tools, that has to change.
Context & Ripple Effects
This adds a data-center access route to earlier coverage of Chinese AI companies moving training data to Malaysian facilities with advanced Nvidia hardware, using overseas data centers to reach restricted compute. It also follows reporting on supply-chain blind spots used to move Nvidia chips toward China, shifting attention from physical diversion to who can use chips once they are installed abroad.
First-order effects
- INF Tech’s access arrangement and Indosat’s Jakarta facility face heightened scrutiny as potential enforcement touchpoints for restricted Nvidia compute.
- The case exposes that controlling a chip’s sale or physical destination does not necessarily control remote or colocated access to its processing capacity.
Second-order effects
- Data-center operators and Nvidia’s channel partners may face pressure to strengthen customer screening, usage monitoring, and contractual controls around advanced systems.
- Chinese AI developers have another demonstrated route for obtaining compute outside China, making enforcement dependent on cooperation from overseas infrastructure providers rather than border controls alone.
Third-order effects
- If similar arrangements persist, export controls will increasingly need to govern access to AI compute as a service, not only the movement of chips through supply chains.
- The pattern could deepen a split between jurisdictions that host advanced AI infrastructure and those whose firms seek access to it, with enforcement credibility determined by visibility into end use.
The trend: AI-chip restrictions are evolving into a contest over controlling cross-border access to compute, not merely the shipment of hardware.