Supercomputing sparks AI innovation

January 4, 2026

Cutting-edge technology:

The University of Hong Kong has developed a generative AI capable of real-time 3D scene generation, turning concepts into images.

Unique advantage:

The University of Hong Kong Department of Electrical & Electronic Engineering Associate Professor Ngai Wong says that by making the AI compatible with mobile phones and desktop computers, the team aims to broaden the market reach.

Visionary plan:

Cyberport Head of Business Development Crystal Fok says Cyberport aims to introduce more diverse types of AI models to enhance the AI ecosystem.

Download Video| Transcript

The University of Hong Kong Department of Electrical & Electronic Engineering Associate Professor Ngai Wong:
Create a luxurious hotel with long corridors. Warm golden lighting, and then elegant wall decor with long sense of depth, and then viewed at human eye level. Let's start. 

Voiceover:
With just a few words, this video wall will transform into a 3D animation based on your description. Whether it’s an image of a hotel or a forest, the transformation happens in the blink of an eye.

Researchers at the University of Hong Kong’s Department of Electrical & Electronic Engineering have developed a generative artificial intelligence (AI) capable of real-time 3D scene generation and rendering. 

The department’s Associate Professor Ngai Wong says this generative AI can turn concepts or product ideas into images, creating something from nothing. The technology is suitable for various applications, ranging from professional industry presentations to video game design.

The University of Hong Kong Department of Electrical & Electronic Engineering Associate Professor Ngai Wong:
For example, if we do interior design, we do not need to physically build the interior or the model for it. We can just use a voice prompt to change the interior design into a different style.  

Another big application scenario is in the gaming industry. So, I can create some fancy environment. I can just voice it out and then the AI model will generate and create that for me.

Voiceover:
Ngai Wong said that unlike many global AI initiatives, the department is focused on compressing these models. By making them compatible with low-end edge devices, such as mobile phones and desktop PCs, the team aims to broaden the market reach.

AI development requires substantial resources. The Government has allocated $3 billion to implement a three-year multi-pronged Artificial Intelligence Subsidy Scheme. 

This programme will provide subsidies to eligible users so they can access the computing power of the Cyberport's Artificial Intelligence Supercomputing Centre (AISC). 

The University of Hong Kong is one of the institutions participating in this programme.

Cyberport Head of Business Development Crystal Fok: 
The reason is if every applicant has to purchase the GPU (Graphics Processing Unit) resources and then host the resources and form an IT team to maintain it, it is actually not only a problem of cost and also a problem of time. 

So in order to speed up the AI development in Hong Kong, actually a ready-to-use AI supercomputing centre could help to facilitate all these projects. 

Voiceover:
Deployment of the AISC started in December of 2024 and its average utilisation rate has risen to 80%. 

Projects from nearly 20 universities, institutions and companies have been approved for funding, with the average subsidy reaching 70% of the list price for the services.

To promote the development of the local artificial intelligence (AI) ecosystem, the Government has allocated $3 billion for the launch of a three-year AI Subsidy Scheme to support eligible organisations such as local universities, research institutes and enterprises in leveraging the computing power of Cyberport’s AI Supercomputing Centre (AISC) to achieve scientific breakthroughs.

Popular applications

Projects from around 20 universities, institutions and companies have been approved for funding, with the average subsidy reaching 70% of the list price for the services.

Deployment of the AISC started in December of 2024 and its average utilisation rate has been nearly 80%.

The University of Hong Kong is one of the institutions participating in the programme.

Researchers at the university’s Department of Electrical & Electronic Engineering have developed a generative AI capable of real-time 3D scene generation and rendering.

Concept shaping

The department’s Associate Professor Ngai Wong said this generative AI can turn concepts or product ideas into images, creating something from nothing.

“For example, if we do interior design, we do not need to physically build the interior or the model for it. We can just use a voice prompt to change the interior design into a different style.”

“Another big application scenario is in the gaming industry. So, I can create some fancy environment. I can just voice it out and then the AI model will generate and create that for me.”

The technology is suitable for various applications, ranging from professional industry presentations to video game design.

Mr Wong added that unlike many global AI initiatives, the department is focused on compressing these models. By making them compatible with low-end edge devices, such as mobile phones and desktop computers, the team aims to broaden the market reach.

Research empowerment

Cyberport said in order to speed up AI development in Hong Kong, a ready-to-use AI supercomputing centre could help to support all these projects.

“The reason is if every applicant has to purchase the Graphics Processing Unit resources and then host the resources and form an IT team to maintain it, it is actually not only a problem of cost and also a problem of time,” explained Cyberport Head of Business Development Crystal Fok.

“Cyberport aims to introduce more diverse types of AI models to enhance the AI ecosystem, enabling the technology to benefit society and assist various industries,” she added.