On September 23, Ho Chi Minh City hosted NVIDIA AI Day, part of a global series of large-scale events where participants experienced the latest AI technologies, explored strategies, and discussed pathways to turn AI from theory into practice.

Building a sovereign AI foundation for Vietnam

In a special panel titled “Building sovereign AI for Vietnam: Vision, capabilities, and opportunities,” Vo Xuan Hoai, deputy director of the National Innovation Center (NIC), noted: “The government is focusing on building Vietnamese-language datasets for AI research and development. At the same time, developing AI talent is a top priority, with a target of over 50,000 engineers in the coming years. Comprehensive policies to support enterprises, startups, and human resources are essential to shape a strong AI startup ecosystem within the next five years.”

Advancing AI development has been identified as a core mission of both the Party and Government, aiming to position Vietnam as a leading center of AI innovation and deployment in ASEAN.

Le Hong Minh of VNG said: “Sovereign AI means building applications and business models that are authentic and uniquely valuable. Vietnam still faces two major challenges: limited investment compared to the world and a lack of research expertise and core technology platforms. To address this, Vietnamese tech companies must focus on product development and concrete business opportunities. AI cannot be forced on users - it must integrate naturally into their experiences.”

He also stressed speed as a decisive factor. VNG successfully commercialized AI Cloud within six months, and after two years of testing, 20% of Zalo users are already using AI-powered features. VNG continues to seek collaboration with the government, universities, researchers, and startups to expand high-value AI applications.

“Today’s AI is like the internet in 1995–1996: full of doubts and uncertainties. Developers must remain inspired by what technology can achieve. Unlike Silicon Valley startups that raise funds before products exist, Vietnamese companies must both explore and survive,” Minh remarked.

Vietnamese LLM debuts on NVIDIA NIM

During the event, GreenNode introduced GreenMind-Medium-14B-R1 (GreenMind), the first open-source Vietnamese large language model (LLM) optimized for NVIDIA NIM and capable of running on a single NVIDIA H100 Tensor Core GPU.

Leveraging self-improving data loops and human involvement throughout training, GreenMind is designed for practical use cases such as enterprise AI assistants, intelligent chatbots, document retrieval systems, advanced search tools, and Vietnamese reasoning tasks.

Vo Trong Thu, head of AI Lab at GreenNode, explained: “We focus on mastering core technologies while building GreenNode AI Platform as an open ecosystem, making AI more accessible to the community. Launching GreenMind on NVIDIA NIM is a major step for Vietnam to take control of local AI development and help businesses and startups adopt AI in a cost-effective, practical way.”

Developed by Vietnamese AI engineers, GreenMind provides an affordable, locally compliant solution that integrates easily with existing infrastructure, reducing reliance on foreign technology.

“New-generation AI models are evolving beyond mechanical Q&A. They are learning to ‘reason’: understanding context, distinguishing concepts, and producing more precise answers. To achieve this, we must build Vietnamese reasoning datasets, trained on Vietnamese knowledge, language, and culture. Only then will we create models truly designed for Vietnamese users,” Thu added.

Investing in core AI capabilities

Alongside open-source initiatives, VNG continues to invest heavily in proprietary LLMs trained from scratch, pursuing the goal of Vietnamese-built AI that can compete globally.

Dr. Chau Thanh Duc, head of R&D at Zalo AI, underlined that developing a sovereign Vietnamese LLM requires high-quality data, large-scale deployment, and training methods adapted to limited resources.

These lessons come directly from Zalo’s experience in creating AI products tailored for Vietnamese users, aligned with its “AI-First” vision. Zalo AI now operates dedicated R&D labs focused on advanced AI technologies.

Zalo’s success has made Vietnam one of the few Southeast Asian countries to develop a domestic LLM, a milestone of strategic importance in today’s highly competitive global tech race.

“Our goal for Zalo is that within five years, 100% of Vietnamese people will use messaging and intelligent AI assistant services developed by Zalo,” Minh said.

Thai Khang