CSPs in Southeast Asia Look to AI Cloud Services to Drive New Revenue Streams

In November 2024, Radian Arc and COMIT announced a joint effort to meet the increasing needs for Artificial Intelligence (AI) and Machine Learning (ML) capabilities among enterprises in Vietnam and the wider Southeast Asian (SEA) region. By leveraging Radian Arc’s Graphics Processing unit (GPU) infrastructure for AI model training and inference processing, enterprises are able to leverage the company’s platform to perform compute-intensive applications to support their AI workloads.

Also, in today’s world where data sovereignty is becoming a key concern and requirement across nations, Radian Arc’s GPU-as-a-Service solution is able to overcome this, as its GPU infrastructure is installed within local Communication Service Provider (CSP) networks, such as edge data centers, thereby ensuring that data do not leave the territory. This is where COMIT’s large network of CSP clients in the region—including but not limited to Viettel, VNPT, MobiFone, etc.—come into play.

While not new, the move to capitalize on the increasing demand for AI compute capabilities is gaining momentum among CSPs in the SEA region. In this regard, Singtel has been one of the most active CSPs trying to capitalize on this opportunity and has announced multiple partnerships with NVIDIA, the Bridge Alliance, Nscale, SK Telecom, and GMI Cloud, to build up its GPU-as-a-Service capabilities. In October 2024, Singtel also launched its AI Cloud Service, RE:AI, which provides enterprises and organizations affordable access to AI applications and technologies without the need for them to maintain the underlying infrastructure to support these AI workloads.

Elsewhere, Lintasarta, a company owned by Indonesian CSP Indosat, has also launched its AI cloud service named GPU Merdeka. In collaboration with NVIDIA, this service offers up to 3.35 Terabits per Second (Tbps) bandwidth with the deployment of eight NVIDIA H100 SXM Tensor Core GPUs. In Malaysia, YTL Group, TM Global, Maxis, etc., have all announced plans to offer GPU-as-a-Service capabilities to their customers.

With growing demands for data and AI sovereignty, local CSPs—with their extensive domestic networks and infrastructure—are well positioned to capitalize on this opportunity. The above examples and trends clearly illustrate that CSPs recognize the opportunity, too.

While CSPs could simply provide the underlying GPU infrastructure to support AI workloads, this may not be sufficient to ensure commercial success for the operator in a highly competitive market environment. Many large players, such as hyperscalers and cloud service providers, are also keen to grab a slice of the pie. For example, Google Cloud and Gulf Edge (headquartered in Thailand) have executed an agreement to develop a sovereign cloud in Thailand. SIAM.AI also announced its intentions to develop a sovereign cloud in the same country.

ABI Research is of the view that to effectively compete in this space, CSPs will need to add value. Some positive examples that ABI Research has observed include:

• Offering an All-in-One Turnkey Solution: CSPs can leverage their expertise in connectivity by providing an AI development platform that is coupled with network solutions. Singtel’s RE:AI provides a good example of how this can be done, with the platform providing developers with a combination of AI compute infrastructure and a diverse range of networking options like 5G, fixed, and/or quantum safe networks.
• Supporting Localized AI Implementations: CSPs can also provide a seamless AI implementation process for enterprises by offering advanced AI models and capabilities on their platforms. For example, Singtel has partnered with Scale AI to integrate its data annotation capabilities into its RE:AI platform (see IN-7651 for more information). In addition to its AI cloud platform, Indosat has taken this a step further and launched its self-developed open-source Large Language Model (LLM), Sahabat-AI, that is designed to address the local languages and context in Indonesia.

All in all, ABI Research believes that CSPs are well positioned to capitalize on the growing demand for flexible AI computing services. However, to succeed in this space, service differentiation will be key. To encourage greater AI adoption among enterprises, CSPs can perhaps even consider working with hyperscalers to integrate their AI marketplaces, such as the Amazon Bedrock Marketplace or Microsoft Commercial Marketplace, on their respective AI cloud platforms.