DEMAND for Asean data centres is likely to remain resilient as Chinese artificial intelligence (AI) DeepSeek is set to disrupt the industry and increase demand for it.
The region is “still in the early stages of its data centre build cycle”, said Maybank analyst Hussaini Saifee on Friday (Jan 31).
He expects more efficient large language models to accelerate AI adoption, boost demand and ease graphic processing units (GPU) supply concerns.
China AI startup DeepSeek rocked the tech industry last week after it launched its latest AI models, which it says are better and cheaper than rival models in the US.
While DeepSeek’s AI bots may have some optimisation in training, Saifee pointed out that their primary advantage is their ability to process data more quickly and efficiently.
“Optimisation of compute requirements enables broader accessibility, reduced reliance on high-end GPUs and lower entry barriers for AI development,” he said.
BT in your inbox
Start and end each day with the latest news stories and analyses delivered straight to your inbox.
Saifee added: “As AI becomes more efficient, adoption speeds up, fuelling greater demand for data centre capacity while easing concerns over hardware constraints.”
Moreover, the analyst also believes that DeepSeek’s free platform could lead to higher adoption by mid to low tier companies, thus boosting compute demand.
As at Jan 29, the company’s AI Assistant, powered by DeepSeek-V3, has overtaken rival ChatGPT to become the top-rated free application available on Apple’s App Store in the US.
The breakthrough could lead companies, such as Microsoft and OpenAI, to reduce their pricing, said Saifee, who was citing founder and chief advisor of Vivander Advisors Michael Goldrich.
With US AI companies “moving to the next level of AI innovation”, demand for higher compute resources, which are used for model training and near-limitless memory, is set to rise, said Saifee.
He also pointed out that the next wave of physical AI applications, such as autonomous vehicles, will be “highly GPU-intensive” due to real-time processing and inference requirements.
“DeepSeek’s breakthroughs may possibly help to accelerate such innovations and as such doesn’t risk slowing down compute demand,” said Saifee.
