Australian AI infrastructure company Firmus is expanding its presence in Malaysia after securing a multi-year agreement with OpenAI to provide computing capacity from two planned Malaysian datacentres. The deal places Malaysia within Firmus' growing Asia-Pacific AI infrastructure network and adds another major international technology investment to the country's rapidly expanding datacentre sector.
Although Firmus remains relatively unknown outside the infrastructure industry, the company already operates through leased datacentre capacity in Malaysia, Singapore and Australia. It is also developing its own facilities in Tasmania. Details surrounding the two new Malaysian sites remain limited, however, with Firmus yet to disclose their exact locations, expected operating dates or individual power requirements.
OpenAI Deal Significantly Expands Firmus' Contracted Capacity
The scale of the OpenAI agreement becomes clearer when looking at Firmus' overall contracted power capacity. Across its Asia-Pacific operations, the company now has agreements representing as much as 900MW of contracted capacity.
That is a dramatic increase from around 155MW reported in April, suggesting Firmus is rapidly moving from a relatively specialised AI infrastructure provider toward a much larger regional operator.
Contracted capacity refers to the maximum amount of electrical power customers have agreed to use rather than the amount being consumed immediately. Even so, reaching 900MW illustrates just how large the potential infrastructure build-out could become if those agreements are fully deployed.
Two Malaysian Datacentres Will Support OpenAI
Under the agreement, OpenAI will receive computing capacity from two Firmus datacentres planned for Malaysia.
The arrangement reflects the enormous infrastructure requirements behind modern generative AI systems. Training and operating advanced models requires thousands of specialised accelerators alongside substantial power, cooling and networking resources.
For OpenAI, securing additional datacentre capacity in Asia-Pacific provides another way to expand compute availability closer to customers in the region. For Malaysia, it reinforces the country's growing role as a regional destination for hyperscale and AI-focused infrastructure.
NVIDIA Hardware Will Power the Infrastructure
Firmus' AI datacentres are expected to rely heavily on NVIDIA accelerator technology. The company's most powerful current offering uses systems based on NVIDIA's Blackwell architecture, which was specifically designed for large-scale AI training and inference workloads.
Firmus has also reportedly expressed plans to deploy NVIDIA's next-generation Vera Rubin processors across its Asia-Pacific infrastructure once those systems become commercially available at scale.
That suggests the company is planning beyond today's hardware generation. Rather than building facilities around one fixed accelerator platform, Firmus appears to be positioning its infrastructure to adopt increasingly powerful NVIDIA systems as they become available.
900MW Does Not Translate Directly Into NVIDIA Revenue
The headline 900MW figure may sound enormous, but it should not be interpreted as a direct indication of how much NVIDIA will earn from the relationship.
Power capacity only indicates how much electricity the contracted infrastructure could potentially consume. NVIDIA's actual revenue will depend on how many accelerators Firmus ultimately purchases, which GPU generations are deployed and how quickly the datacentres reach their planned capacity.
A facility could have significant contracted power long before every rack is populated with AI hardware.
The real commercial opportunity for NVIDIA will therefore depend on actual processor deployment rather than headline megawatt figures alone.
Why Malaysia Is Becoming Attractive for AI Datacentres
Firmus has not revealed where its two planned Malaysian facilities will be located, but Johor would be an obvious candidate given the state's rapidly expanding datacentre ecosystem.
Johor's proximity to Singapore provides an important geographical advantage. Singapore remains one of Southeast Asia's biggest technology and financial hubs, but limitations involving available land and power have encouraged datacentre operators to expand into neighbouring Malaysia.
Johor offers considerably more space while remaining geographically close enough to provide high-speed connectivity into Singapore and the wider regional network.
This has helped transform the state into one of Southeast Asia's fastest-growing datacentre markets.
Government Incentives Are Helping Malaysia Attract Technology Investment
Malaysia has also been actively encouraging investment in digital infrastructure through federal and state-level initiatives.
The country offers relatively attractive land availability, expanding power infrastructure and access to regional fibre connectivity. These advantages are becoming increasingly important as AI datacentres require significantly more electricity and physical capacity than traditional enterprise facilities.
The arrival of major international technology companies has further strengthened the ecosystem. Microsoft, Oracle, NVIDIA and other global technology groups have already announced or participated in substantial Malaysian cloud and AI infrastructure initiatives.
Firmus joining that expansion reinforces the idea that Malaysia is becoming more than simply an overflow location for Singapore.
Johor Could Become a Major Regional AI Infrastructure Hub
If Firmus does place its new facilities in Johor, it would join a rapidly growing concentration of hyperscale and AI-oriented datacentre projects.
The state's large industrial areas provide enough room for campuses that may eventually require hundreds of megawatts of power. Its location also provides relatively short network routes into Singapore, making it attractive for companies serving customers across Southeast Asia.
The bigger question will increasingly be whether electricity generation and grid infrastructure can keep pace with demand.
AI datacentres consume substantially more power per rack than conventional facilities, meaning the next phase of development will require not only more buildings but significant investment in energy infrastructure.
Power and Cooling Will Be Major Considerations
Firmus has not disclosed how much electricity each Malaysian datacentre will consume, but the company's overall 900MW contracted capacity provides some indication of the scale being considered.
Modern AI accelerators are extremely power intensive. NVIDIA Blackwell systems can consume enormous amounts of electricity when thousands of GPUs operate together, while upcoming Vera Rubin platforms are expected to continue pushing compute density higher.
Cooling becomes equally important because almost all of that electrical energy ultimately becomes heat.
Large AI datacentres increasingly rely on advanced liquid-cooling systems rather than conventional air cooling alone, particularly when supporting densely packed accelerator clusters.
The OpenAI Deal Could Strengthen Malaysia's AI Ecosystem
Hosting infrastructure used by OpenAI could have benefits beyond the construction of the datacentres themselves.
Large AI facilities require engineers specialising in networking, power systems, cooling, datacentre operations and accelerator infrastructure. They also create demand for local suppliers, fibre connectivity and supporting services.
Over time, the presence of major AI compute infrastructure may also make Malaysia more attractive to companies building services around artificial intelligence.
Rather than accessing all advanced computing capacity from overseas, businesses operating in the region could potentially benefit from infrastructure located much closer geographically.
Firmus Is Expanding Quickly Across Asia-Pacific
The jump from 155MW of contracted capacity in April to as much as 900MW illustrates just how aggressively Firmus is expanding.
The company already operates across Australia, Singapore and Malaysia while developing its own datacentres in Tasmania. Adding two OpenAI-linked Malaysian facilities significantly raises its profile within the regional AI infrastructure market.
That expansion also reflects the larger race to secure computing capacity before demand for advanced AI accelerators grows even further.
AI companies increasingly need to reserve power and datacentre space years in advance because suitable facilities cannot simply be built overnight.
Malaysia Is Becoming Part of the Global AI Infrastructure Race
The Firmus agreement is another indication that Malaysia is increasingly participating in the physical infrastructure behind the global AI industry.
Much of the public conversation around artificial intelligence focuses on models and applications such as ChatGPT, but those services ultimately depend on enormous datacentres filled with specialised computing hardware.
The countries capable of providing land, electricity, cooling and high-speed connectivity for those facilities become important parts of the AI supply chain.
Malaysia's location, infrastructure development and proximity to Singapore are increasingly placing it in that position.
Final Thoughts
Firmus' multi-year agreement with OpenAI represents another significant development for Malaysia's rapidly expanding datacentre sector. Two planned Malaysian facilities will provide AI computing capacity to one of the world's most prominent artificial intelligence companies, while Firmus' total contracted capacity across Asia-Pacific has climbed dramatically to as much as 900MW.
The infrastructure is expected to rely heavily on NVIDIA technology, beginning with Blackwell-based systems and potentially expanding into future Vera Rubin accelerators. Exactly how many GPUs will ultimately be deployed remains unknown, meaning the headline power capacity should not be mistaken for guaranteed hardware deployment.
Firmus has also kept important details about its Malaysian facilities private, including their location, operating timeline and power consumption. Johor appears a plausible destination given the state's proximity to Singapore and rapidly expanding datacentre ecosystem, but that remains unconfirmed.
What is increasingly clear is the larger direction: Malaysia is becoming an important physical hub for the infrastructure powering the AI economy. As OpenAI and other technology companies demand ever more compute capacity, the race is no longer only about building better models—it is also about finding enough land, power and hardware to run them.
Comments 0