Skyvast Corporation, a Malaysia-registered tech firm, recently unveiled a project it claims will be the country’s first ‘sovereign, full stack AI infrastructure’ – a domestically controlled and vertically integrated system built and operated by either state or private entities, encompassing everything from hardware layers such as data centers and servers to AI frameworks and local cloud infrastructure. Operated by its subsidiary Skyvast Cloud Sdn Bhd, the initiative is built around the Altermatic DT250 AI Server, powered by Huawei’s Ascend GPU and the Chinese-developed DeepSeek large language model. Skyvast aims to scale this infrastructure to over 3,000 GPUs deployed across multiple geographic zones by 2026. While the Malaysian government’s direct involvement in the project was unclear, Deputy Communications Minister Teo Nie Ching attended the announcement event and expressed support and echoed the company’s claim.
Yet within days of the announcement, Minister Teo retracted her previous statement, clarifying that no Ascend chips had been sold to, nor would be deployed by, the Malaysian government. Huawei itself corroborated this, stating that no Ascend processors were part of any government-endorsed initiative.
This retraction underscored how quickly a country’s AI policy can become entangled in geopolitical calculations, especially when underlying technology and infrastructure are sourced from Chinese firms. It also highlights a deeper challenge: as Malaysia and other Southeast Asian states work to localize AI capacity, they risk reproducing external dependencies under a different form.
In this article, we argue that even with the implementation of a “sovereign, full-stack AI infrastructure,” a country may still face the risk of digital dependency. Despite efforts at localization, Southeast Asian nations remain vulnerable to reliance on Chinese AI technologies for several key reasons.
First, Chinese AI data remains under the control of Chinese firms even when deployed overseas. The Digital Silk Road (DSR) – first introduced in the 2015 Belt and Road Initiative (BRI) Whitepaper and gaining prominence by 2017 – has funneled substantial funding into the global expansion of Chinese investments in fiber optics, data centers, and cloud infrastructure. While DSR financing initially prioritized telecommunications and fintech, it has pivoted since the pandemic toward accelerating the deployment of AI systems abroad.
Chinese AI systems, built by major firms such as Huawei, Alibaba Cloud, Tencent, and SenseTime, are highly centralized and tightly integrated with cloud-based infrastructure. While regional centers outside China are necessary for Chinese actors to expand AI capacity and reach, the technical and intellectual makeups of these overseas facilities are essentially under the control of their headquarters. For instance, Huawei Cloud has built data centers in Thailand, Singapore, the Philippines, and Indonesia. The cloud backends of these facilities are configured, maintained, and periodically updated from China-based service centers. Receiving countries do not have the right or capacity to independently audit these systems. Instead, audit access requires explicit permission from Huawei.
In most cases, Huawei conducts internal audits or hires a China-based third-party firm to audit the infrastructure within the receiving country, but this auditing process does not include the backend systems in China. Receiving countries receive reports of what Huawei chooses to send them. In other words, receiving countries lack direct access and the ability to audit these facilities independently and fully. Alibaba Cloud built and used regional nodes, but has its software stack in Hangzhou, updating the systems and maintaining metadata while interacting with China-controlled systems abroad.
With a centralized end-to-end solution, Chinese vendors have access to monitor their remote systems, including but not limited to system logs, metadata, training outputs, and retrieval of anonymized data. The core code and control lie with the vendor. As such, updates, access rights, and feature expansions are centrally managed. One example is SenseTime’s smart city platforms, which are installed in host countries but whose license renewal, algorithmic transparency, and backend updates remain under the control of its China-based facility.
The ramifications of Chinese AI systems are far-reaching. They create a form of geostrategic lock-in: once a country builds and adopts these systems, it becomes increasingly dependent on them, making transitions to alternative providers exceedingly costly and complex. Switching to Western systems would require billions in investment to replace hardware, migrate software, transfer and reformat data, and retrain personnel. This entrenchment grants Beijing latent coercive leverage, especially in sensitive policy areas such as Taiwan, the South China Sea, or Hong Kong. As governments and businesses across Southeast Asia become embedded in Chinese AI ecosystems, their ability to cultivate independent technological capacities diminishes. The result is a fragile illusion of stability—one that can be abruptly withdrawn when political tensions rise.
Second, Chinese AI infrastructure often embeds the logic of algorithmic governance shaped by China’s surveillance-intensive domestic context. Chinese firms are global leaders in technologies such as facial recognition, predictive policing, and behavioral monitoring – tools that have been instrumental in surveilling political dissidents and controlling sensitive regions like Xinjiang and Hong Kong. When exported to Southeast Asia, where democracies are often young, elite-driven, or institutionally fragile, these technologies risk enabling incumbents to entrench power and suppress dissent, gradually eroding democratic checks and consolidating executive authority.
Recent developments across the region illustrate how ruling elites have begun to adopt and weaponize these tools. In the Philippines and Indonesia, leaders like Rodrigo Duterte and Joko Widodo rose to power through promises of rapid development, populist redistribution, and streamlined governance. However, under both administrations, there have been credible reports of state-linked efforts to manipulate public discourse, including the use of troll farms and coordinated digital campaigns to silence critics and shape public narratives. The introduction of AI-driven surveillance in such a context risks reinforcing these authoritarian tendencies, allowing ruling coalitions to marginalize opposition, close remaining democratic channels, and legitimize power consolidation under the guise of public security or national development. The result is a quieter yet more durable erosion of democratic institutions.
One-party states such as Laos and Cambodia have embraced AI under the banner of “digital transformation,” backed heavily by Chinese support. Companies like Huawei and Hikvision have supplied Cambodia with smart city and surveillance infrastructure, while Laos has begun integrating similar technologies in its capital, Vientiane. In both cases, AI is being deployed not merely for modernization but as a tool to consolidate bureaucratic control and expand state coercion over society. Although these two countries lack the strength of more developed regional peers, their increasing strategic alignment with Beijing – and their growing political willingness to adopt Chinese surveillance tools and models – has made them key frontiers in China’s expanding AI footprint in Southeast Asia.
Third, for Malaysia and some other countries in the region, China’s promotion of the DSR and AI development has been underway for nearly a decade, often in collaboration with host-country governments. More recently, G3 Global Berhad, a Malaysian AI company, emerged as a key champion for China’s AI expansion in Malaysia. In 2019, G3 entered a strategic partnership with the Chinese firms SenseTime and China Harbour Engineering Company to build the country’s first AI park, aiming to develop AI applications in facial recognition, smart surveillance, and autonomous systems, as well as to establish AI research and development centers and AI labs for local talents.
The primary investor in G3 Global Berhad, Aminul Islam, is reportedly connected to the United Malays National Organization and influential figures within key government agencies, including the Ministry of Immigration. His case exemplifies how business elites in the region leverage partnerships with Chinese firms to advance their commercial ventures, simultaneously expanding their economic reach and political capital. Importantly, discussions around the region’s pressing need for robust data governance must not be divorced from the dynamics of patronage capitalism, which foreign technology partnerships may further entrench.
Finally, while Western AI systems present their own challenges, they are primarily driven by a model of data commercialization. Companies like OpenAI, Google, Meta, and Amazon operate under shareholder capitalism, which raises important concerns around data privacy and ownership. However, they are generally less likely to expose users to direct geopolitical risks, given the fact that they are not extensions of the U.S. government but accountable to investors and subject to some degree of public scrutiny. In contrast, Chinese tech firms maintain far closer ties to the state and are legally obligated under the 2017 Cybersecurity Law to share data with government authorities.
Data governance is becoming an increasingly salient norm in Western societies, which may compel firms to adopt similar standards abroad. The European Union’s General Data Protection Regulation and California’s Consumer Privacy Act represent growing efforts to build comprehensive data protection regimes that could serve as global benchmarks. Western AI governance still suffers from under-regulation, potential regulatory capture, and fragmentation – but unlike the Chinese model, it leaves more room for public input and legal challenges. The ability to challenge misconduct by Chinese firms or shape the governance frameworks they operate under is virtually nonexistent – and often illegal – within China’s digital regime. For Southeast Asian countries, this represents a critical long-term tradeoff that demands serious and strategic consideration.