SuperX is expanding its Singapore-based AI Innovation Centre with Model as a Service, allowing enterprises to access locally deployed AI models through APIs without managing model deployment and operations themselves.
SuperX AI Technology is adding a model-serving layer to its Singapore AI infrastructure business, introducing AI Innovation Centre 2.0 (AIIC 2.0) with Model as a Service (MaaS) capabilities.
The upgrade is designed to address a practical gap between AI experimentation and production deployment. Instead of requiring organizations to provision infrastructure and then deploy and operate an AI model themselves, AIIC 2.0 provides access to pre-deployed models through APIs.
The AI Innovation Centre was launched by SuperX and ST Telemedia Global Data Centres in April 2026 at the STT Singapore 5 facility in Tai Seng. Its original purpose was to give enterprises a local environment for testing AI workloads and moving successful pilots toward production. SuperX’s current AIIC platform now offers both GPU-as-a-Service and Model-as-a-Service.
The new MaaS layer effectively moves SuperX further up the AI infrastructure stack. Customers no longer need to treat model deployment as a separate infrastructure project before testing an application. They can call a supported model through an API while SuperX manages the underlying serving environment.
The initial service supports GLM-5.3, an open-weight model hosted by SuperX at the Singapore facility. SuperX says additional open-weight models may be added following technical and compliance reviews.
The architecture is particularly relevant to organizations that need local control over AI workloads. SuperX says that for models hosted at AIIC, customer data and inference remain within Singapore, while customer inputs and outputs are not transmitted to model developers. Local hosting is also intended to reduce network latency for users in the region.
That positioning comes as Southeast Asian companies move AI projects toward larger-scale deployment. A 2026 study from McKinsey, the Singapore Economic Development Board and Tech in Asia found that 46% of surveyed Southeast Asian companies had moved beyond AI pilots to scaling, compared with 35% globally.
Scaling, however, introduces requirements that are different from running an isolated proof of concept. Enterprises need predictable access, security controls, usage monitoring and a way to manage costs as model consumption expands across teams.
AIIC 2.0 addresses some of those requirements through usage management. Organizations can monitor model usage and control access, quotas and costs across departments and users.
That places the offering within the increasingly competitive AI cloud and inference infrastructure market. Hyperscalers such as Microsoft Azure, Google Cloud and Amazon Web Services already offer managed model APIs, while specialist AI infrastructure providers are building services around GPU capacity, inference optimization and model access.
SuperX’s differentiation is its focus on locally deployed infrastructure in Singapore combined with its broader AI compute portfolio. The company already operates GPU infrastructure services through AIIC, including NVIDIA B200 configurations, while its wider platform spans AI servers, GPUaaS and AI inference services.
The company has also been expanding beyond infrastructure into model access. In September, SuperX launched an AI Token Platform designed to provide unified API access to more than 100 models from more than 10 providers, alongside AI applications for professional content creation.
AIIC 2.0 represents a somewhat different model: rather than simply aggregating third-party APIs, the service is built around models deployed locally at the AI Innovation Centre.
That distinction could matter for organizations with data-residency, latency or governance requirements. It also gives enterprises another deployment option between running models entirely in-house and sending workloads to externally hosted model providers.
For SuperX, the broader strategy is to turn AI infrastructure into a layered service. Compute capacity provides the hardware foundation; MaaS supplies managed model inference; and its newer model-access and application products sit further up the stack.
The approach reflects a wider evolution in enterprise AI infrastructure. As organizations move from experimentation toward production, the infrastructure challenge is increasingly less about obtaining a GPU and more about operating models reliably, controlling inference costs, maintaining security and integrating model services into existing applications.
AIIC 2.0 therefore places SuperX closer to the AI cloud platform and managed inference market than a conventional data-center infrastructure provider.
The immediate limitation is model availability. The initial MaaS offering centers on GLM-5.3, and SuperX says future model additions will depend on technical and compliance review. Customer adoption will also determine whether locally hosted MaaS can compete effectively with the breadth and elasticity offered by hyperscale cloud platforms.
Still, the move highlights an important shift in enterprise AI infrastructure: customers increasingly want to consume compute and models as services rather than build every layer themselves. By adding MaaS to its Singapore AI centre, SuperX is attempting to shorten that path from an AI experiment to a production API.
Market Landscape
Southeast Asia is moving relatively quickly toward scaled AI deployment. McKinsey, EDB and Tech in Asia found that 46% of surveyed companies in the region had progressed beyond AI pilots, compared with 35% globally.
That creates demand for managed AI infrastructure spanning GPUs, model serving, APIs and governance. The market includes hyperscalers, AI cloud providers and specialist inference platforms. The competitive focus is increasingly shifting toward latency, model choice, data residency, inference economics, security and operational simplicity.
SuperX’s AIIC strategy sits between GPU infrastructure and model APIs, with local Singapore deployment positioned as a key feature for organizations with regional data and compliance requirements.
Top Insights
- AIIC 2.0 adds managed model inference to SuperX’s existing AI infrastructure environment, reducing the deployment work required between AI pilots and production.
- The initial MaaS offering hosts GLM-5.3 locally in Singapore, with SuperX planning additional models subject to technical and compliance review.
- Local model deployment can address latency and data-residency requirements for Southeast Asian enterprises running sensitive AI workloads.
- SuperX is expanding vertically from GPU infrastructure into model access, inference services and enterprise AI applications.
- The MaaS market places SuperX in competition with hyperscalers and specialist AI clouds that increasingly package models and compute as managed services.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
