By Justin Heyes. 28th May 2024
As datacenter development continues in Johor at a considerable pace, the question must be raised, when will Malaysia look to other parts of the country to establish new datacenter hubs as viable alternatives, in a strategic and thought through plan to capitalise on this booming industry while avoiding risking overloading one state?
The demand and subsequent development in Johor, has been unprecedented. From a total of 10 datacenters in Malaysia in 2021, to 51 datacenters total in 2024, with 32 operational and 19 currently in construction, with the majority as well as new builds being in the southernmost state. While this boon has been fantastic news in establishing Malaysia at the forefront of the digital economy in Southeast Asia, the rate of development has taken a toll on the resources within the state.
At the time of writing water tariffs in Johor have spiked comparatively to other states, rising to > 35m3 @ RM 3.55/m3 for non-domestic use, and developers being advised to source sites along the national gridlines for power owing to datacenter parks starting to reach capacity. With more investors looking to capitalise on Malaysia, as a strategic location to increase capacity, there needs to be a more concerted effort in encouraging new developments to be carried out in alternative locations.
It is understandable why Johor has become a priority for datacenter developers looking to enter the Southeast Asian market. Its position as the neighbouring state to Singapore made it an ideal location to take on the overspill of data flowing into Singapore, without sacrificing latency, as the regional hub began to reach its own limits. It is no secret that Johor’s explosion into the market coincided with the Singaporean Moratorium, initially enacted in 2019, this as a response to the considerable energy consumption associated with datacenters.
In Singapore, seven percent of the total electricity consumption goes to datacenters, and this number was projected to reach 12 percent by 2030. As a country with limited space to build, the demand for more capacity and in turn, more power has become a national talking point and will take time to recover from. In January 2022, the Singaporean government lifted the moratorium, however there is currently no news about future allocations beyond the 80MW awarded by EDB-IMDA’s pilot DC-CFA exercised last July, the stall in allocation for development has meant that interest is spilling over to neighbouring countries.
This has been the figurative ‘canary down the coalmine’ for the rest of the APAC region. However the key takeaway, in hindsight, has been somewhat misplaced. Investment in renewable energies and green datacenter build practices have been commendable, and at the time were a logical solution to prevent a repeat of what happened in Singapore, however with the dawn of generative AI, the paradigm has shifted.
Conventional datacenter designs are simply not capable of storing and processing the volume and density of the data being produced. New hyperscale AI datacenters are emerging as the future of the industry, and as it currently stands, it could be a while before these behemoths have technological innovations that will make them less resource hungry.
Server racks for AI computing can consume four times the power currently used in cloud processes, according to Newmark Group. That calls for datacenters that can support surges in workloads, and more advanced cooling systems to handle the heat emitted. Adding this demand to the already increasing resources required to keep Johor running could cause problems.
One proposed solution would be to start allocating areas for development based on the scope of the proposed hyperscale datacenter. Collocations focused on servicing populations that require proximity to reduce latency could be allocated locations around Johor and KL, and would not necessarily require AI racks to function. While more resource intensive AI hyperscale datacenters could be strategically placed in more remote areas such as Kedah, where this is access to a density of subsea cables running through the Songkhla and Satun landing stations. These facilities would be focused on processing and tuning AI, which typically is not burdened by latency requirements and could functionally relay the data within the standard latency for AI processes in under 16ms to Singapore and other countries.
While this proposal at first sounds logistically difficult, as dictating where developers can and can’t build may frighten investors, the reality is with clearly laid out incentives and further development in the areas suitable for hosting AI hyperscale facilities, Malaysia could broaden its appeal as a datacenter hub, while managing resource demands, providing alternative routes for international data and showcasing how the optimal core, regional and edge datacenter organisation system can be successfully managed in the APAC region.