NVIDIA’s GPUaaS Partnership with Singtel—a Sign That Southeast Asia Is Primed for Digital Transformation

In March 2024, Singtel announced that it will be launching its GPU-as-a-Service (GPUaaS) in Singapore and Southeast Asia in 3Q 2024, providing enterprises with access to NVIDIA’s Artificial Intelligence (AI) computing power to drive greater efficiencies that accelerate growth and innovation. Powered by NVIDIA H100 Tensor Core GPU-powered clusters that will be operating in upgraded data centers, the AI-ready data centers are specifically tailored for intense compute environments with purpose-built liquid-cooling technologies for maximum efficiency and lower Power Usage Effectiveness (PUE)—a benchmarking metric used to measure a data center’s energy efficiency—catered to enterprises.

By eliminating the need for heavy investments in setting up and managing expensive data center infrastructure, this partnership between Singtel and NVIDIA envisions that enterprises of all sizes will be able to fully optimize operations with AI. The partnership will enable NVIDIA to leverage Singtel’s regional data center business NXERA to run its GPUaaS in three new sustainable, hyper-connected, and high-density AI data centers located in Singapore, Thailand, and Indonesia when they begin operations.

With Singtel helming the push to democratize access to AI infrastructure within its own country and the region, the company can leverage its extensive fixed broadband network, submarine cables, 5G high-speed network connectivity, and patented cloud platform to orchestrate AI workloads in a multi-network and multi-cloud environment. Its new GPUaaS structure, accompanied by a carrier neutrality agreement, can form a foundational structure for Micro, Small, and Medium Enterprises (MSMEs) to accelerate digitalization efforts, particularly in the form of leveraging AI in Industrial Internet of Things (IIoT) and 5G deployments.

As AI’s list of industrial use cases grows, Singtel-NVIDIA’s GPUaaS will form an attractive avenue for all levels of industry (primary such as agriculture and mining, secondary such as manufacturing, and tertiary such as logistics, finance, and transport). Industries dominant in Southeast Asia, such as manufacturing, construction, and agriculture, stand to gain from easier access to AI computing power via the GPUaaS offering, especially as Internet of Things (IoT) use cases, such as digital twins, predictive maintenance, and Autonomous Mobile Robots (AMRs), are integrating AI and Machine Learning (ML) to enable more intelligent applications. Introducing GPUaaS to the region will also stimulate 5G growth, considering that edge AI cloud and 5G are able to operate on the same plane and consistently collaborate on their integrative functions, collectively enabling more applicable industrial use cases for enterprises to explore.

As ABI Research observes strong potential for the use of AI and 5G technology in sectors critical to the economy of Southeast Asia, offering GPUaaS could provide a definitive roadmap to accelerating digital transformations in MSMEs across Southeast Asia, starting with Singapore, Thailand, and Indonesia, and possibly extending thereafter to other Southeast Asian economies. This lines up with ABI Research’s forecast that AI gateways on AI chipsets and edge AI cloud powered by Graphics Processing Units (GPUs) in Asia-Pacific will grow to 9.2 billion shipments by 2028, with an average Year-over-Year (YoY) growth rate of 33.3%. Manufacturing in Asia-Pacific is expected to account for 55% of all GPU-based AI chip shipments by 2028.

While enterprises can capitalize on the GPUaaS offering from the get-go, its full potential for intense computing and AI-intensive large language processing can go unrealized without proper preparation to integrate and maximize its potential in industrial deployments. ABI Research believes that with new developments in providing cloud computing through GPUaaS, enterprises and industries across Southeast Asia are at the doorstep of highly scalable and cost-effective digital transformation. However, for enterprises to successfully digitalize and optimize their revenue with GPUaaS, they should focus on two key aspects of planning.

First, enterprises should focus on identifying the key impacts and measurements of a successful AI and IoT implementation within their goals. These measurements of success can be decided based on the following:

1. Identify an enterprise/industry’s key pain points as an essential step to spotting an area of improvement that incentivizes changes to existing workflows. By understanding an enterprise’s pain points, it can develop and explore use cases that directly tackle and solve its problems. For instance, commonly identified issues in primary industries such as mining and construction would be the lack of communicative connectivity, along with worker safety. By identifying such pain points, an industry facing such issues could explore GPUaaS-ready IoT use case such as 5G-powered AMRs and digital twin technology that can safely collate massive amounts of data by collating, compiling, and computing actionable insights for its engineers.
2. Understand the Return of Investment (ROI) and Total Cost of Ownership (TCO) expected to form a realistic agreement on what implementing GPUaaS can contribute to increasing revenue and the company’s profitability. A comprehensive ROI and TCO guideline will set expectations for Capital Expenditure (CAPEX) and Operational Expenditure (OPEX), which can manage the long-term strategic digitalization of the enterprise.
3. Explore other integrative use cases that can capitalize on GPUaaS in both IIoT and 5G deployments. Doing so can open new possibilities for an enterprise to capitalize on its existing/planned digital infrastructure deployment to enable other use cases and/or applications.

By setting clear standards and measurements of success for integrating and implementing new technologies in an enterprise structure, enterprises can look to develop key strategies that can prime them for optimizing and maximizing GPUaaS when it comes in 3Q 2024. These strategies should include the following:

1. Develop roadmaps for progressive deployment to help form an overview of expected expenditure, deployment, and implementation that primes an enterprise to adopt and maximize the GPUaaS offering from the moment it launches in 3Q 2024. Introducing data scientists early in an unmanaged and non-digitalized industry can help lay the foundation and lower technical debt in the long run with the service.
2. Establish key partnerships with stakeholders with strong implementation expertise from Communication Service Providers (CSPs) and advisory partnerships that will benefit the overall optimization and implementation of GPUaaS and its promised benefits.
3. Understand the use cases’ Operation Cost Savings (OCS) and how other technologies can further boost ROI in the long run as another viable step to strategizing which use cases are effective solutions to the key industry pain points identified. Decisions on which use case to deploy would involve understanding one’s OCS priority in the following: (i) increasing CAPEX or OPEX savings, (ii) increasing production efficiency and volume, (iii) reducing downtime, or (iv) saving energy.