• NVIDIA is expanding Nemotron-based AI deployments across Southeast Asia.
• Sea will be the first Southeast Asian enterprise to adopt NVIDIA Vera Rubin.

NVIDIA is expanding the use of its AI models, software, and computing infrastructure across Southeast Asia, with organisations in Singapore, Malaysia, Vietnam, Thailand, and Brunei deploying or testing the technology in public services, enterprise applications, and locally developed AI systems.

The announcements around NVIDIA AI Day in Singapore cover public-sector projects, region-specific AI models, physical AI applications, and computing infrastructure. Sea Limited is also set to become the first enterprise in Southeast Asia to adopt NVIDIA’s Vera Rubin platform, according to Sea and NVIDIA.

Public-sector AI and localised models

In Singapore, the Home Team Science and Technology Agency, or HTX, is researching the use of NVIDIA Nemotron 3 Super and Nemotron 3 Nano Omni models for public-safety applications.

HTX is exploring Nemotron 3 Super for reasoning and agent-based workflows, while Nemotron 3 Nano Omni supports vision, audio, and language inputs. The agency is assessing how the models can support AI systems developed around operational data used by Singapore’s Home Team.

HTX has separately made sovereign control a central part of its approach to public-safety AI, particularly for sensitive operational workloads and data. “Sovereign AI is non-negotiable in public safety because societal norms, operational context, and risk thresholds are not always universal,” HTX Chief Executive Chan Tsan said in June.

Singapore established the National AI Council in February this year, chaired by Prime Minister Lawrence Wong, and updated its National AI Strategy in May with 10 refreshed priorities for AI development and adoption.

Singapore-based technology services company NCS is also using Nemotron models and NVIDIA’s Video Search and Summarization Blueprint in AI deployments across enterprise and public-sector environments. The company is also working on physical AI applications involving humanoid robotics.

ST Engineering is using NVIDIA NeMo tools and cuOpt software in its AI Studio platform. The company is applying agent-based AI to operations including marine maintenance, repair, and overhaul.

AI Singapore is incorporating Nemotron open models and NVIDIA NeMo tools into its SEA-LION model family. SEA-LION is designed around languages and cultural contexts used across Southeast Asia.

AI Singapore expanded the family on September 18 with Nemotron-SEA-LION-v4.8, developed with NVIDIA and based on the Nemotron 3 models. AI Singapore said the models were trained using its regional data and cover 10 Southeast Asian languages alongside English.

The release is SEA-LION’s first use of a mixture-of-experts architecture and its first model family based on Nemotron.

In Vietnam, Viettel AI has been fine-tuning Nemotron 3 Super for Vietnamese-language and agent-based applications. NVIDIA said the resulting model ranked first on the VMLU Vietnamese language benchmark and on Viettel’s internal product benchmark.

The model is set to be used in Viettel’s Legal AI system, which is designed to support employees within Viettel Group as well as external customers.

FPT Smart Cloud has also worked with NVIDIA on Nemotron-Personas-Vietnam, an open dataset based on Vietnamese demographic and cultural data. The dataset is intended to support the post-training and evaluation of AI models developed for Vietnamese users.

In Thailand, the Big Data Institute and iApp Technology are exploring Nemotron models through the ThaiLLM Collaboration. iApp Technology is adapting Nemotron 3 Nano using Thai-language legal data and NVIDIA’s NeMo framework.

The resulting OpenThai 2.0 Legal model has been released as open source and is used in Thanoy, iApp Technology’s legal-assistant chatbot. NVIDIA said the service has around 43,000 users.

Malaysia established AI Malaysia Berhad under the Ministry of Digital in July this year, building on the National AI Office created in December 2024. The government said the organisation will coordinate national AI policy and implementation and support the National AI Action Plan 2026–2030.

Malaysia’s YTL AI Labs is fine-tuning Nemotron models for enterprise and citizen-facing services.

Singapore-based Hummingbird Bioscience, together with LynxKite, is using Nemotron 3.5 Lightning and NeMo Retriever to develop an explainable toxicity knowledge graph for drug research. The project is intended to combine public and proprietary data from multiple file formats for analysis and in silico simulations.

Other projects apply NVIDIA technology to physical infrastructure and urban systems. Malaysia’s ITMAX is using NVIDIA Cosmos models with the company’s Video Search and Summarization Blueprint for traffic-management applications.

Thailand’s AS-TECH is applying the same technology stack to passenger-flow management in airports. In Brunei, Antrique has developed an AI platform aimed at applications in the country’s food sector.

NVIDIA is also working with regional developers through its Nemotron open-model ecosystem. Bitdeer AI recently co-hosted an Open Models AI Codefest with NVIDIA, providing GPU cloud infrastructure for developers working with Nemotron models, datasets, and training tools.

Sea adopts Vera Rubin

Sea Limited, the Singapore-founded company behind Shopee, Monee, and Garena, plans to use Vera Rubin to expand the computing infrastructure supporting AI development across its businesses. Sea said the investment is intended to give its teams more computing capacity to develop AI models and intelligent agents.

Li Kuangli, senior director of partnerships and strategy at Sea, described building AI capabilities across the technology stack as part of the company’s competitive strategy. “We see this as a core differentiator in how we choose to compete and create value in the AI era,” Li said at NVIDIA AI Day Singapore.

Vera Rubin is NVIDIA’s rack-scale AI architecture, combining the Vera CPU, Rubin GPU, NVLink 6 interconnect, ConnectX-9 networking, BlueField-4 data-processing units, and Spectrum-6 Ethernet switching.

NVIDIA said seven chips used across the Vera Rubin platform entered full production in March this year. By May 31, the company said the platform was ramping into full production, with server makers and supply-chain partners manufacturing Vera Rubin-based systems at scale.

Rubin GPUs use HBM4 memory with up to 288GB of capacity per GPU, according to NVIDIA.

NVLink 6 provides up to 3.6TB/s of bidirectional bandwidth per GPU, while a Vera Rubin NVL72 system connects 72 Rubin GPUs through the interconnect.

Sea is already using AI across Shopee, Monee, and Garena. At Shopee, AI tools are used for product-listing creation, product discovery, customer engagement, and seller decision-making.

Sea said AI-generated content tools at Shopee supported a 14% year-on-year increase in purchase conversion rates in the first quarter. At Monee, recent changes to transformer-based credit-risk models increased approval rates by around 10% while maintaining a similar level of risk, according to the company.

Garena teams have created more than 1,300 AI chatbots, workflows, and agents, according to Sea, alongside in-game AI companions designed to respond to player intent.

Sea plans to use Vera Rubin to support further development and deployment of AI models and agents across the three businesses.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

Tech Wire Asia is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.