简体中文(中国) 
A comprehensive analysis of Firmus Technologies’ two-pronged strategy: securing multi-gigawatt compute commitments from dual hyperscale anchors Meta and OpenAI across Australia and Southeast Asia, underpinned by a US$300 million express capacity investment on SUBCO’s transpacific APX East subsea cable.
1. Executive Overview
Australian AI infrastructure developer Firmus Technologies has executed a vertically integrated model to establish a sovereign compute generation and optical export engine across the Asia-Pacific region. By pairing megawatt-scale liquid-cooled AI factory builds with a US$300 million (~A$418 million) cornerstone commitment on SUBCO’s transpacific APX East subsea hypercable system, Firmus addresses both ends of the hyperscale equation: high-density GPU compute and dedicated multi-terabit subsea transport.
This infrastructure architecture is anchored by two of the world's leading AI companies—Meta Platforms and OpenAI—pushing Firmus’s contracted pipeline past 900 Megawatts (MW) across seven AI factory projects in Australia, Singapore, Malaysia, and Indonesia, while securing up to 150 Terabits per second (Tbps) of dedicated express subsea bandwidth between Australia and the United States.
2. Meta Platforms & Firmus Asia-Pacific Infrastructure Deal
Meta Platforms has entered into a multi-year strategic agreement with Firmus to access high-density GPU compute capacity across Firmus’s expanding network of AI factories in the Asia-Pacific region to power foundational research, LLM training, and consumer AI workloads.
Meta serves as the cornerstone tenant at Firmus’s flagship Melbourne AI factory, which deploys the Southern Hemisphere's largest installation of NVIDIA GB300 NVL72 (Blackwell Ultra) architectures, designed specifically for complex frontier model training.
The agreement grants Meta expansion options across upcoming Firmus AI factory campuses in Singapore, Malaysia, and Indonesia, built on NVIDIA’s DSX AI Factory platform layout to provide seamless multi-regional cluster scaling.
All Firmus compute nodes utilize its proprietary HyperCube immersion/liquid-cooling platform. Co-designed with intelligent grid-integration software, HyperCube maximizes compute throughput per watt while drastically reducing total energy draw, water consumption, and physical land footprint—enabling sustained 100+ kW per rack densities.
3. OpenAI Anchor Commitment & Southeast Asian AI Expansion
Preceding the Meta announcement, OpenAI signed a landmark multi-year agreement to become an anchor customer for Firmus, contracting dedicated AI compute capacity across two new AI Factory sites in Malaysia. This partnership marks Firmus’s expansion into its fourth Asia-Pacific market and established the foundation for its regional gigawatt pipeline.
The Malaysian facilities built for OpenAI will deploy NVIDIA Vera Rubin NVL72 rack-scale systems integrated with the NVIDIA DSX AI Factory platform. This marks one of the first regional deployments of NVIDIA's next-generation accelerated compute hardware in Southeast Asia.
The OpenAI anchor commitment expanded Firmus's total portfolio to seven AI factory facilities across Australia, Singapore, Indonesia, and Malaysia (two operational, five under development), pushing total contracted capacity across all clients past 900 MW.
"This multi-year partnership marks the moment Asia-Pacific becomes a producer of intelligence, not just a consumer of it. Firmus has spent seven years building for this—a footprint that runs from Southeast Asia to Australia, and the manufacturing, technology, and teams to bring it online at speed."
4. The $300M Subsea Pipeline: Securing SUBCO’s APX East Cable
To deliver petabits of AI inference tokens and training model states generated for both Meta and OpenAI to global end-users, Firmus committed US$300 million (~A$418 million) to become the lead cornerstone customer on SUBCO’s APX East subsea cable system.
| Subsea Cable Metric | Project Specification |
|---|---|
| System Name | SUBCO APX East (Transpacific Express) |
| Firmus Capital Commitment | US$300 Million (~A$418 Million) |
| Secured Optical Capacity | Up to 150 Tbps over a 25-Year IRU Term (~6 Fiber Pair equivalent) |
| Geographic Route | Sydney (AU) → US West Coast (via Hawaii and Fiji express spurs) |
| Fiber Pair Architecture | 16 Fiber-Pair Spatial Division Multiplexing (SDM) Trunk |
| Target Ready-for-Service (RFS) | Q4 2028 |
Unlike traditional subsea investments driven by enterprise cloud backhaul, Firmus and SUBCO are pioneering dedicated optical infrastructure for AI token exports. Australia and Southeast Asia leverage local land, power, and liquid cooling to generate digital intelligence tokens, which are then exported instantly over APX East’s 150 Tbps express pipeline to North American and global enterprise consumers.
5. Integrated Stack Dynamics: "Energy In, Tokens Out"
Firmus’s vertically integrated architecture solves the primary physical bottleneck facing standalone data center operators: securing guaranteed, low-latency, multi-terabit subsea connectivity directly from high-density compute nodes to international markets.
Grid integration across Australia, Malaysia & Singapore
HyperCube Cooling + NVIDIA Blackwell & Vera Rubin
150 Tbps Transpacific Express Subsea Pipeline
Meta, OpenAI & Global AI Consumers
- Domestic Subsea Interconnect (SMAP & Bernacchi-1)
In addition to APX East, Firmus works closely with SUBCO to underwrite domestic subsea routes, securing undersea cable capacity to Tasmania (Bernacchi-1 system, RFS Q2 2027) and SUBCO’s 5,000 km SMAP transcontinental coastal ring.
- Commercial Trajectory & Planned A$5B IPO
With anchor commitments secured from both Meta and OpenAI, Firmus’s active development pipeline exceeds 900 MW. Backed by long-term optical transport agreements through 2053, Firmus is preparing for a planned AUD $5 billion initial public offering (IPO) on the Australian Securities Exchange (ASX).
The combination of Meta and OpenAI anchor commitments alongside Firmus’s $300M investment in SUBCO’s APX East subsea cable establishes a new paradigm in Asia-Pacific digital infrastructure. By bridging high-density liquid-cooled compute with dedicated 150 Tbps transpacific subsea capacity, Firmus delivers a scalable model for regional AI token generation and international export.
