Meta has signed agreements to rent AI chip capacity at Firmus-operated data centers across Southeast Asia, positioning the social media giant to scale its AI model training and innovation in the region.
- Meta rents AI chip capacity at Firmus data centers in Southeast Asia
- Deal includes Nvidia DSX hardware inside modular HyperCube units
- Firmus eyes major IPO raising about $5 billion in October
What happened
Meta has entered into agreements to rent AI chip capacity from Firmus, an Australian company that builds and operates data centers in Southeast Asia. The partnership allows Meta to reserve a specific amount of capacity at multiple Firmus sites, with flexibility to increase as needed. Meta intends to use this capacity to research, build, and train advanced AI models, leveraging the latest hardware technologies.
The data centers utilize Nvidia’s DSX architecture integrated within Firmus’s HyperCube modular units, which include built-in power and cooling systems manufactured in rural New South Wales. This marks an expansion beyond Meta’s existing use of Firmus data centers in Australia, where Nvidia’s Blackwell Ultra chips are already deployed. Financial terms and the exact operating sites in Southeast Asia remain undisclosed.
Why it matters
This deal highlights Meta’s growing investment in regional AI infrastructure to support its AI ambitions across Asia-Pacific. By partnering with Firmus, Meta gains access to scalable, state-of-the-art computing capacity designed specifically for AI workloads, positioning the company to accelerate AI model development closer to key markets in Southeast Asia.
For Firmus, this agreement further validates its business model ahead of a major share market debut planned for October. The firm aims to raise approximately $5 billion in one of Australia’s largest IPOs, capitalizing on strong demand for AI-specific data center capacity from multiple anchor customers, including OpenAI and Nvidia.
What to watch next
Market watchers should monitor Firmus’s October IPO and the extent to which it can scale its infrastructure to meet increasing demand from AI companies like Meta. Expansion of its capacity in Southeast Asia, including Indonesia, Singapore, and Malaysia, will be critical to sustaining growth and capturing additional regional business.
Additionally, the deployment of Nvidia’s next-generation Vera Rubin chips in Firmus’s HyperCube units will be a key technological milestone. This will enhance performance and energy efficiency for AI workloads, making Southeast Asia an increasingly important hub for AI research and cloud computing services.