Meta Taps Australia’s Firmus for AI Computing Capacity in Southeast Asia
Meta Platforms is expanding its access to artificial intelligence computing capacity in Southeast Asia through a new infrastructure partnership with Australian AI infrastructure company Firmus.
Firmus said on September 29 that it had signed AI infrastructure agreements with Meta for graphics processing unit (GPU) computing capacity at its data-centre facilities in Southeast Asia. Financial terms were not disclosed. The agreement will give Meta contracted capacity along with expansion options based on Nvidia’s full-stack DSX platform for AI research, model development and training.
The deal comes as Meta accelerates spending on AI infrastructure globally. For investors, it also highlights a broader shift in the AI industry: access to computing power, electricity, cooling and data-centre capacity is becoming as strategically important as the AI models themselves.
What the Meta-Firmus Deal Includes
Under the agreement, Firmus will provide Meta with contracted GPU capacity and options to expand that capacity at its Southeast Asian facilities.
The infrastructure will use Nvidia's DSX platform, a full-stack architecture designed to integrate computing, networking, cooling, power and other components required for large-scale AI workloads.
Meta will use the capacity for AI research, model development and training, according to Firmus and Reuters. The companies have not disclosed the value or exact capacity of the latest agreement.
The arrangement builds on Meta's existing relationship with Firmus. Meta has already deployed Nvidia GB300 NVL72 systems at Firmus' Melbourne data centre. Firmus described that installation as the largest deployment of Nvidia Blackwell Ultra systems in the Southern Hemisphere.
This means the latest agreement is not an isolated cloud purchase. It is part of a wider relationship between Meta and a specialised AI infrastructure provider.
Why Southeast Asia Is Becoming Important for AI Compute
Southeast Asia is emerging as an important location for AI infrastructure because the region combines growing digital demand with access to land, power and connectivity.
Firmus currently has a broader AI factory portfolio spanning Australia, Singapore, Indonesia and Malaysia. The company said in September that it had more than 900 MW of contracted capacity across customers, with seven AI factories in its portfolio. Two were operational and five were under development, targeting readiness over the following 24 months.
Malaysia has become particularly important for Firmus. Earlier in September, OpenAI signed a multi-year agreement to contract dedicated computing capacity from two Firmus AI Factory sites in Malaysia, making OpenAI an anchor customer.
The arrival of both Meta and OpenAI customers shows how Southeast Asia is becoming part of the infrastructure layer supporting global AI companies.
Meta's AI Spending Is Already Massive
The Firmus agreement needs to be viewed against the scale of Meta's broader infrastructure investment.
Meta's latest SEC filing says the company expects $130 billion to $145 billion in 2026 capital expenditures, including principal payments on finance leases. The company said the spending is intended to support its AI efforts and core business.
Meta spent about $49.11 billion on property and equipment in the first six months of 2026, primarily on servers, data centres and network infrastructure.
Its infrastructure commitments have also increased sharply. As of June 30, Meta reported $349.31 billion of non-cancelable contractual commitments, most of which related to third-party cloud capacity and investments in servers, network infrastructure and data centres.
That scale explains why Meta is using a mix of owned infrastructure, leased capacity and external AI infrastructure partners.
Why Meta Needs More Computing Capacity
Training advanced AI models requires enormous amounts of computing power. But training is only one part of the equation.
Once an AI model is released, users continuously interact with it. Every request requires computing resources, a process commonly referred to as inference.
As Meta expands its AI products, including its broader AI assistant and enterprise ambitions, demand for inference capacity can grow alongside user adoption.
This creates a structural infrastructure requirement. Meta needs enough GPU capacity not only to develop future models but also to serve AI workloads reliably and at competitive costs.
Using specialised infrastructure providers such as Firmus can provide additional flexibility without requiring Meta to build every facility itself.
Firmus Is Building an AI Infrastructure Business
Firmus is positioning itself as an "AI factory" operator rather than a conventional data-centre company.
Its platform combines high-density computing with power, cooling, networking and other infrastructure. Firmus says its systems are designed to reduce power, water and space requirements for AI workloads.
The company has also partnered with Nvidia on a major AI factory programme in Batam, Indonesia. In June, Firmus announced plans for a 360 MW Nvidia DSX AI Factory campus in Batam, with the wider project targeting 170,000 GPUs.
Firmus has raised significant capital to support its expansion. In April, the company announced a planned $505 million strategic equity investment led by Coatue, with Nvidia participating subject to certain conditions. The funding was intended to accelerate AI factory deployments across Asia-Pacific.
The company's expansion therefore sits directly at the intersection of three major AI trends: Nvidia's accelerated computing ecosystem, hyperscalers' demand for capacity and the growing data-centre buildout across Asia-Pacific.
What It Means for Nvidia
Nvidia is another company indirectly positioned to benefit from the expansion of AI infrastructure.
Firmus' latest Meta agreement will use Nvidia's DSX platform, while Firmus has separately partnered with Nvidia on AI factory deployments across the region.
This illustrates Nvidia's broader strategy of supplying not just GPUs but increasingly complete infrastructure architectures around accelerated computing.
For Nvidia, the more AI factories that are built and the more customers that require high-performance computing, the larger the potential market for its hardware and software ecosystem.
However, the ultimate financial impact of any individual Firmus agreement on Nvidia cannot be quantified from the disclosed information.
The Infrastructure Challenge: Power and Cost
The rapid expansion of AI data centres also creates risks.
AI facilities consume substantial amounts of electricity, while high-density GPU systems generate significant heat and require sophisticated cooling systems.
Australia's data-centre industry is already facing concerns about whether electricity supply and grid infrastructure can keep pace with planned AI projects. Firmus itself has been developing energy-efficient infrastructure to address some of these challenges.
For Meta, infrastructure availability is therefore not simply a technology issue. Power availability, construction timelines, GPU supply, cooling efficiency and regional connectivity can all influence how quickly new AI capacity comes online.
There is also a financial consideration. Meta is committing enormous amounts of capital to AI infrastructure, so the return on that investment will increasingly depend on whether AI products generate enough additional revenue and engagement to justify the spending.
What Investors Should Watch Next
For investors tracking Meta, Nvidia and the broader AI infrastructure sector, several developments will be important.
First, watch how quickly Meta's AI infrastructure spending translates into higher usage and monetisation across its products.
Second, monitor the growth of AI data-centre capacity in Malaysia, Indonesia and Singapore. Deals involving OpenAI, Meta and other large AI companies could accelerate the region's role as a global computing hub.
Third, infrastructure efficiency will become increasingly important. Better cooling, lower power consumption and higher GPU utilisation can reduce the cost of running AI workloads.
Finally, investors should watch Firmus' planned public-market expansion. Reuters reported that the company is preparing for a roughly $5 billion IPO, which could make it one of Australia's major upcoming listings.
Bottom Line
Meta's agreement with Australia's Firmus is another sign that the AI race is increasingly becoming an infrastructure race.
The companies have not disclosed the financial value of the Southeast Asian agreement, but the strategic significance is clear: Meta is securing additional GPU capacity for AI research, model development and training while Firmus is expanding its position as a regional AI infrastructure provider.
For investors, the development reinforces a broader theme across the technology sector. The next phase of AI growth will require enormous investments in GPUs, data centres, electricity, cooling and networking.
Meta's ability to turn those infrastructure investments into profitable AI products — and Firmus' ability to efficiently deliver the computing capacity customers require — will be key factors to watch as the industry expands.
Follow AIViralHub for more updates on AI companies, technology infrastructure, global markets and business developments.
This article is for informational and educational purposes only and should not be considered investment advice.