China’s infrastructure advantage could shape Southeast Asia’s AI choices
China’s infrastructure advantage could shape Southeast Asia’s AI choices
China is positioning itself to become a key provider of the infrastructure underpinning Southeast Asia’s AI ambitions. Chinese companies are embedding AI into telecommunications networks, cloud services and critical infrastructure across the region. Their advantage is not necessarily better models but the equipment and relationships through which AI services can be delivered. As governments and businesses adopt these technologies, they risk developing dependencies that could become increasingly difficult to reverse.
This dimension of China’s growing influence has received less attention than the risks posed by AI-enabled cyber threats. This year’s Singapore International Cyber Week, held from 12 to 15 October, centred on AI’s influence on the cybersecurity and threat landscape. This included the technology’s implications for Singapore’s new cybersecurity strategy, launched this month, and Southeast Asian regional cooperation. But the question of who builds, operates and maintains the infrastructure supporting Southeast Asia’s AI ambitions may prove just as consequential for the region’s technological autonomy.
China’s approach has become more organised. The first ministerial roundtable on AI held between China and the Association of Southeast Asian Nations (ASEAN) took place in September 2025 and led to the launch of the China-ASEAN AI Application Cooperation Centre in Nanning, Guangxi. A second meeting, in September 2026, brought together representatives from all 11 ASEAN members. Guangxi provincial officials report 44 AI cooperation agreements with ASEAN countries and joint large-model research with eight.
At the fourth China-ASEAN AI Cooperation Forum on 16 September, organisers unveiled five outputs, including a Lao-language model for Laos’s state electricity corporation. The model was described by Chinese officials as their first overseas AI project in the power sector. More revealing is the delivery model, which retains computing power in Guangxi while exporting services. Laos receives a locally adapted application, but the infrastructure behind it remains in China.
The financial scale remains modest. Organisers of the 23rd China-ASEAN Expo reported nearly 100 preliminary agreements worth around US$150 million (A$210 million). Officials describe a division of labour: research in Beijing, Shanghai and Guangzhou; integration in Guangxi; and application across Southeast Asia.
Chinese companies are also working through infrastructure already in the region. Huawei has long worked with operators from Thailand to Brunei, and AI is introduced as an extension of those relationships. At Mobile World Congress 2026, Huawei and Indonesian telecommunications company Telkomsel announced an agreement combining Indonesian 5G and broadband upgrades with embedded AI, while Telkomsel also expanded its cooperation with ZTE on AI-native networks.
Others are building positions elsewhere. Chinese technology conglomerate Tencent has outlined further data-centre investment in Indonesia, and ByteDance’s TikTok has secured approval to expand data infrastructure in Thailand. None of this requires standalone decisions to procure Chinese models. AI can arrive through a telecommunications contract, a cloud agreement or a software upgrade. Each may make commercial sense; together, they deepen reliance on one ecosystem.
Open-weight models reinforce the trend. DeepSeek showed that capable models could be offered at low cost, and Alibaba’s Qwen has become an important foundation for development beyond China. In May, Singapore released Sea-lion v4.5, a collection of large language models optimised for use in Southeast Asia. The release included models adopted from China’s Qwen and Gemma. This reflects how Chinese models have entered the wider ecosystem, rather than Singapore falling under Chinese influence.
But sovereignty at the model layer doesn’t mean independence across technologies. A government may replace its model while remaining dependent on particular chips, cloud infrastructure or technical support.
Malaysia illustrates the complications. At the May 2025 launch of its private sector AI-infrastructure initiative, Deputy Communications Minister Teo Nie Ching described plans involving Huawei’s Ascend AI hardware and DeepSeek. Within two days, the trade ministry distanced the government from the plan, saying it had not been developed, endorsed or coordinated by the government. Huawei said it had not sold Ascend chips in Malaysia. Chinese technology may offer an affordable route to greater AI capability, but it can also carry geopolitical and supply-chain risks.
These risks should not be overstated: Chinese suppliers offer advantages in cost and delivery. The more immediate concern is dependencies that governments cannot easily assess or reverse. An operator reliant on proprietary AI network management faces costs in switching. If utilities use externally hosted models, they’ll have limited understanding of how they are updated or secured. Rather than ownership, the issue is who can audit, maintain, modify and replace a technology.
China is also extending into AI governance, with five ASEAN members among the founding signatories of the Shanghai-based World Artificial Intelligence Cooperation Organisation. Two developments will test Southeast Asian governments’ choices. ASEAN concluded negotiations on its Digital Economy Framework Agreement in May, and its implementation will shape cross-border data flows and digital trade. Malaysia’s sovereign computing decisions offer a more immediate test of how governments balance Chinese technology against US restrictions.
China’s growing role has emerged through commercially attractive investments, infrastructure upgrades and technical partnerships. Collectively, they could shape who provides the computing power, operates the networks and sets the terms for AI. Governments may believe they are keeping options open by avoiding an explicit choice between Chinese and Western AI, but the most consequential choices may already be taking place in other areas of the technology. By the time those dependencies become apparent, changing course could be harder.
About Author
.png "53% vs.19%: The budget divide separating AI leaders from everyone else")
