en.Wedoany.com Reported - On July 12, Indonesia's Polibeli Group Ltd. (Polibeli) and US-based Amazon Web Services (AWS) are in advanced discussions regarding artificial intelligence computing infrastructure in Southeast Asia, planning to build an AI computing center to serve large model inference services. The project aims to leverage AWS's existing cloud computing and AI service systems to provide inference computing power for large model developers within and outside the region, with potential clients including AI companies such as Anthropic. Currently, publicly available information has not disclosed the specific location, server count, chip configuration, or official construction timeline of the computing center.
The focus of this project is not large model training, but continuous inference after model deployment.
When users invoke large models through applications, data must be computed in server clusters, generate results, and return to terminals. As user scale expands, the inference center needs to handle a large number of concurrent requests simultaneously, placing ongoing demands on GPU servers, memory capacity, high-speed storage, cluster networks, and task scheduling systems. Unlike one-time centralized large-scale training, inference services often require round-the-clock operation and dynamically allocate computing resources based on traffic changes. Therefore, equipment utilization, response latency, and computing power consumption per task will directly impact the operational capability of the computing center.
According to the current disclosed direction, Indonesia's Polibeli will participate in organizing regional computing resources and advancing the project, while US-based AWS is expected to provide the cloud platform, AI service systems, and corresponding technical environment. The two parties plan to configure computing resources around large model inference scenarios, enabling the computing center to handle inference tasks with different models, user scales, and response speed requirements. If the project proceeds to the implementation phase, the scope may cover server cluster deployment, data storage, high-speed network connections, data center power supply and distribution, cooling systems, and computing management platforms, but specific equipment plans remain to be announced by both parties.
Another disclosed construction lead is located in Thailand.
On July 2, Indonesia's Polibeli signed a non-binding memorandum of understanding with Thailand's AUTHAIKAM COMPANY LIMITED to initiate a preliminary assessment for a large-scale AI computing center. The project plans for a maximum power capacity of approximately 100 megawatts, with current work including project document exchange, site inspections, land condition analysis, power supply assessment, and technical feasibility studies. The project remains in an early evaluation stage, and whether it enters formal development will depend on subsequent results regarding site, power, technical conditions, and regulatory approvals.
A 100-megawatt-level computing center imposes significantly higher demands on power systems compared to typical enterprise data centers. When AI servers operate continuously, not only do the computing devices themselves require substantial power, but cooling, power conversion, backup power, and network facilities also create long-term loads. Therefore, site selection must simultaneously consider grid access capacity, power supply stability, land conditions, communication networks, and cooling environments. The fact that Indonesia's Polibeli and its Thai partner are first conducting land and power assessments in the preliminary phase indicates that whether power resources can support large-scale server cluster operations will be a key factor determining if the project meets construction conditions.
It is currently unconfirmed whether the 100-megawatt project in Thailand and this AWS collaboration refer to the same computing center. However, both arrangements revolve around AI computing infrastructure in Southeast Asia: one has entered site and power condition evaluation, while the other focuses on large model inference services and AWS technical systems. These two project leads suggest that Indonesia's Polibeli is attempting to extend its digital supply chain business experience into server resource organization, computing center construction, and AI computing services. Subsequent project progress will require close observation of actual milestones such as site confirmation, grid connection, server procurement, data center design, and initial computing power deployment.
US-based AWS has established a cloud infrastructure network covering multiple markets in Southeast Asia. The AWS Singapore region launched in 2010, the Indonesia region began operations in 2021, the Malaysia region launched in 2024, and the Thailand region went live in 2025. Existing regional nodes provide local enterprises with computing, storage, database, and AI services, and also provide a foundation for the new inference computing center to connect to the AWS service system.
AI applications in Southeast Asia are transitioning from model experimentation to business system deployment, driving increased demand for inference computing power. After enterprises deploy customer service bots, content generation, intelligent search, data analysis, and AI agents, they require continuous model invocation, rather than concentrated computing power usage only during the model training phase. The focus of computing center construction is also shifting from simply expanding GPU numbers to coordinated operation of servers, storage, networks, and software scheduling. How to improve concurrent processing capabilities, control response latency, and maintain cluster stability will become key technical issues that the project must address as it moves from planning to actual operation.
US-based AWS's current AI applications in Southeast Asian enterprises and public sectors already cover scenarios such as financial services, real estate, image processing, and government systems. The increasing number of AI applications in the region will further drive demand for local inference resources. If the computing center being advanced by Indonesia's Polibeli and US-based AWS is completed, it could add a layer of local computing capacity between AWS's regional infrastructure and AI applications, providing computing nodes closer to end users for large model invocation.
At this stage, the project remains primarily focused on collaboration discussions and preliminary preparations, with no announcement yet on construction scale, equipment suppliers, data center design firms, start time, or initial computing power deployment. Whether it can materialize into an actual project will depend on the simultaneous implementation of site selection, power access, server configuration, technical solutions, and client demand.
