The Research Institute for Democracy, Society, and Emerging Technology (DSET) Economic Security Program recently published an op-ed titled “Courting the ‘Compute Neutrals’: How the US Can Pull Southeast Asia from China’s AI Orbit” on The Pilot, a commentary platform under Pacific Forum. The article analyzes how, as the United States strengthens export controls on advanced AI chips to China, Southeast Asian countries—particularly Malaysia—have become key nodes in global compute diversion and infrastructure reconfiguration, and how Washington can consolidate the effectiveness of export controls and reduce supply chain spillover risks through institutionalized cooperation mechanisms.
The article was co-authored by DSET Economic Security Program Policy Analyst Jin Chian Seer and Non-Resident Fellow Chih-Hua Tseng. It builds on DSET’s recent policy report, “A Shared Future? Economic Security Challenges from Malaysia–China Economic Cooperation and Data Center Development.” The authors point out that since the United States began restricting exports of high-end AI training GPUs to China in 2022, Chinese firms have gradually rebuilt a “black compute network” through third countries—deploying data centers and server supply chains outside China so that Chinese entities can effectively access restricted advanced chip computing power.
The article notes that Malaysia, with its strategic location, mature semiconductor backend industry, and long-standing non-aligned foreign policy, has increasingly become a representative “compute neutral.” On the one hand, Malaysia is viewed as a potential chip transshipment node; on the other hand, it has attracted Chinese firms to legally invest in high-performance data centers, creating a gray zone. The DSET report categorizes regional AI data center deployment into four models: Shadow Greenfield investment, Joint Venture, Colocation, and Hybrid Cloud. It highlights that both the Joint Venture and Hybrid Cloud models have already shown concrete cases of compute spillover risks.
The article further cites empirical cases to illustrate these risks. For example, under the joint venture model, Malaysian and Chinese firms have co-developed large-scale data centers deploying servers built on NVIDIA’s Blackwell architecture, serving clients with Chinese capital backgrounds. While such cooperation contributes to local digital economic development, it may also allow Chinese-linked entities to access advanced compute resources outside China. Under the hybrid cloud model, Chinese engineers have transported large volumes of training data to Malaysia, rented servers equipped with restricted chips to conduct model training, and then brought the results back to China. This shows that an export control regime centered solely on hardware exports struggles to address the cross-border movement of “compute use locations.”
Malaysia’s policy stance has also shifted in recent years. The article observes that while Malaysia has deepened memoranda of cooperation with China, it has also followed guidance from the US Bureau of Industry and Security (BIS), cooperated with US authorities in investigating illegal GPU transshipment, and established a licensing system for advanced chip imports. This development highlights Malaysia’s strategic dilemma between capturing AI-driven economic benefits and avoiding entanglement in great power competition.
In response to these challenges, the DSET Economic Security Program suggests that US policymakers should not view Malaysia merely as a frontline transshipment risk, but rather as a partner in AI infrastructure diplomacy. In addition to strengthening supply chain verification and investigations into indirect transfers, the United States should establish an AI Infrastructure Partnership Framework to protect overseas data centers that adopt US-origin technologies through supply chain disclosure and end-user transparency mechanisms. The article also recommends promoting a Trusted AI Infrastructure Whitelist mechanism, combined with a cross-verification system for usage declarations, to enhance transparency and prevent compute spillover.
The article concludes that advancing conditional cooperation while maintaining technological security will be critical for future policy. By building secure, transparent, and resilient overseas AI infrastructure, the United States can strengthen supply chain resilience and sustain institutional leadership in setting the standards for the next phase of AI development.
