DeepSeek may fuel data centre growth in Malaysia
04 Feb 2025
Instead of disruption or distraction, the arrival of “cheaper” Chinese artificial intelligence (AI) model DeepSeek may drive demand for data centres in Malaysia, industry experts said.
The country, they added, remains in a strong position to capitalise on sustained regional data processing due to its geographic location that is deemed as attractive.
BMI Research head of technology and telecoms research Andrew Kitson said Malaysia already recorded above-average mobile and wireline data traffic volumes at 13.21 exabytes (EB) and 13.25 EB respectively in 2023.
The data for the first nine months of 2024 showed further robust increases in data traffic over these channels and BMI forecasts robust annual increases in the traffic over the next 10 years.
“Malaysia remains in a particularly strong position to capitalise on sustained regional data processing, storage and transfer demands,” Kitson told Business Times.
“This is given its attractive geographic location as a hub for submarine cable connectivity, its strong relationships with key US allies, its proximity to Singapore as a leading financial services hub, and its remarkable progress with regards to digital transformation that cannot be matched by neighbours such as Indonesia and Vietnam.”
He, however, said DeepSeek’s ability to match or outperform existing AIs built on larger and costlier computing stacks will give investors and developers of digital infrastructure pause for thought.
“We believe that countries currently positioning themselves as regional data processing hubs including Malaysia will want to take stock with regards to current and planned data centre developments.
“This may include downsizing compute capacity at data centre sites still at the planning stage or even deferring making firm commitments to data centre projects that have only been mooted so far,” said Kitson.
Despite that, he said there will not be one single AI used in every country.
There will be considerable demand for high-performance computing in all countries that exceed current and planned capacity as AI applications and use cases become broadly embedded across all layers of the digital economy.
“Basically, the more that people and companies use AI, the more computing power is going to be needed,” added Kitson.
Juwai IQI co-founder and group chief executive officer Kashif Ansari said the firm foresees continued and increased demand for data centres in Malaysia and the Asia Pacific region instead of disruption from DeepSeek.
“Any efficiencies DeepSeek can deliver compared to other large language models will drive growth in the use of AI services that will far outpace efficiency gains. When you make something cheaper, people consume more of it.
“The tech industry even has a term for this, the ‘Jevons paradox,’ which states that increasing the efficiency of a resource usually leads to its greater consumption.”
Kashif added that the local companies should be happy that DeepSeek offers a more affordable large language model as expensive AI models will not be accessible to all Malaysian companies.
A more affordable AI enables local companies to compete with global players.
“All the factors that have made Malaysia into Southeast Asia’s fastest growing data centre market still apply. There is still growing demand for data centres, and Malaysia is still one of the best locations in the Asia Pacific for those data centres.
“We have relatively inexpensive land, skilled labour, water and energy. We have a supportive government. We have a large domestic market and another important one next door, in Singapore. And we have a strategic location with excellent connectivity to the rest of Asia,” he added.
Meanwhile, RHB Investment Bank Bhd (RHB Research) said although DeepSeek’s large language model claimed to have been trained at just US$5.6 million, it does not necessarily translate into a significant reduction in the need for data centres.
“Instead, it means that the AI model gets 30 per cent more power. Certain companies have lamented that AI is not able to deliver targeted return on investments and, hence, a more efficient AI model could enable such aimed returns to be met,” it said in a note.
The firm said Jevons paradox could come into play, where technological progress makes using a resource more efficient, and overall consumption of that resource tends to increase.
“Assuming demand for AI is relatively elastic, falling prices due to efficiency improvements create higher AI adoption. We understand that one factor that slowed AI adoption within big organisations so far has been how expensive the AI models are to run,” it added.
Additionally, it said Meta Platforms Inc and Microsoft Corporation have not changed their plans to invest heavily in AI hardware in data centres for this year.
It added that the tech companies have defended their AI-related investments for the current fiscal year, saying it is crucial to remain strategically competitive in AI over time.
DeepSeek recently took centre stage in the tech world as it claims to operate on a fraction of the resources used by its competitors.
Its performance against industry leaders such as Google and OpenAI has unsettled big technology stocks, particularly in the semiconductor sector.
Digital Minister Gobind Singh Deo recently said the government was studying the impact of the platform on Malaysia.
He said the government was giving serious consideration to DeepSeek and its model before it can be adapted for local use to keep pace with the rapid development of the AI landscape.
Source: NST
