OpenAI Anchors Firmus Expansion Into Malaysia

Firmus Technologies has signed a multi-year strategic partnership with OpenAI that will put dedicated AI compute capacity in two new Firmus AI Factory sites in Malaysia, extending the Australian infrastructure company deeper into Southeast Asia. OpenAI will become an anchor customer, giving Firmus a major workload commitment as it expands an infrastructure portfolio designed around increasingly dense accelerated computing systems. The agreement pushes Firmus’ total contracted capacity across all customers above 900 megawatts, according to the company. It also positions Malaysia as the fourth country in a regional footprint already spanning Australia, Singapore and Indonesia.

The significance goes beyond adding another data center market to Firmus’ map because the agreement connects a major AI model developer directly with an Asia-Pacific infrastructure platform. Firmus now has seven AI factories across the four countries, with two operational sites in Australia and Singapore and another five under development. The company is targeting ready-for-service status for those five sites over the next 24 months. OpenAI’s role as an anchor customer gives that development pipeline a demand-side commitment at a time when access to power, accelerated computing and supporting infrastructure increasingly shapes where AI workloads can expand.

“This multi-year partnership marks the moment Asia-Pacific becomes a producer of intelligence, not just a consumer of it,” said Tim Rosenfield – Firmus Co-Founder and Co-CEO. “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.” The statement captures Firmus’ broader proposition: regional infrastructure should do more than host conventional digital workloads as AI demand grows. Instead, the company is positioning its facilities as production infrastructure for computational output at industrial scale.

NVIDIA Vera Rubin Moves Into the Infrastructure Plan

Firmus plans to deploy NVIDIA Vera Rubin technology at scale across Asia-Pacific as part of its infrastructure expansion. The AI factories developed under the partnership will use the NVIDIA DSX AI Factory Platform to integrate NVIDIA Vera Rubin NVL72 rack-scale systems with Firmus’ HyperCube platform. That combination brings accelerated computing, networking and software closer to the underlying cooling, mechanical and electrical infrastructure required to operate high-density AI systems. Moreover, it reflects an industry shift in which compute architecture and physical infrastructure increasingly need to be engineered as a coordinated system rather than procured as separate layers.

Firmus describes HyperCube as a repeatable infrastructure system combining advanced liquid cooling, mechanical systems and electrification, with manufacturing and prefabrication taking place in regional New South Wales. The model aims to standardize more of the physical infrastructure needed to support dense accelerator deployments before equipment reaches an AI factory site. Firmus expects tighter integration across the AI factory value chain to improve deployment speed, capital efficiency and operating performance. For customers buying large quantities of AI compute, those factors ultimately matter because infrastructure performance contributes to the economics of producing and serving AI tokens.

“AI infrastructure demand is accelerating across Asia-Pacific, and meeting it requires full-stack AI factories engineered for performance, durability, efficiency, scale and fungible workloads,” said Nico Caprez, vice president of Global AI Infrastructure Growth at NVIDIA. “Built with NVIDIA DSX and powered by NVIDIA Vera Rubin NVL72 systems, Firmus’ HyperCube-based AI factories will help bring advanced capacity online faster and produce intelligence at industrial scale with stronger token economics across the region.” NVIDIA’s involvement therefore reaches beyond supplying accelerators, with the DSX architecture forming part of the infrastructure design that Firmus plans to deploy.

OpenAI Secures More Asia-Pacific Compute Capacity

For OpenAI, the Malaysian agreement adds dedicated infrastructure as demand for AI products continues to drive requirements for large-scale computing capacity. The partnership gives the company access to capacity from two Firmus sites without either company publicly specifying the individual site locations, megawatt allocations or financial value of the contract. Those omissions matter when assessing the precise scale of the Malaysian deployment, even though Firmus has disclosed that total contracted capacity across its entire customer base now exceeds 900 MW. The agreement therefore provides a clear infrastructure commitment while leaving several commercial and site-level details private.

“We’re excited to work with Firmus as we expand our global compute capacity. These new data centers in Malaysia will help us serve growing demand for OpenAI’s products across the region and around the world. We look forward to bringing this capacity online together,” said Sachin Katti, VP of Compute Strategy at OpenAI. The statement frames Malaysia as part of OpenAI’s wider compute expansion rather than simply a local capacity purchase. It also places the Firmus agreement within the larger infrastructure challenge facing AI developers: securing enough physical compute capacity to support growing product usage across multiple markets.

Malaysia Gains Weight in the AI Infrastructure Map

Malaysia’s addition gives Firmus a broader geographic platform for serving AI demand across Asia-Pacific while complementing its existing operations and development pipeline. The country has already emerged as a rapidly expanding Southeast Asian data center market, attracting infrastructure development as operators search for large power allocations, suitable sites and connectivity into major regional digital markets. Firmus is entering that environment with an architecture specifically designed around accelerated AI computing rather than treating GPU infrastructure as an addition to conventional data center capacity. The OpenAI commitment gives the expansion an immediate hyperscale AI workload relationship.

The arrangement also illustrates how the competitive unit in AI infrastructure is expanding beyond the GPU itself. Power delivery, liquid cooling, networking, prefabricated mechanical systems and software integration increasingly determine how quickly operators can turn accelerator allocations into usable compute. However, Firmus will still need to convert its contracted pipeline into operating capacity as five of its seven announced AI factories remain under development. That execution gap makes deployment timelines, infrastructure readiness and operational performance important measures of the platform as its contracted portfolio grows.

Firmus Plans Australian AI Access Program

Alongside the OpenAI agreement, Firmus said it intends to establish an Australian AI Access Program aimed at extending computing access to eligible researchers and organizations. The proposed initiative will focus on AI applications in areas including science, education, agriculture, energy and resource efficiency, and climate resilience. Firmus has not disclosed a budget, capacity allocation, launch date or selection process for the planned program. Its current status is therefore an announced intention accompanying the broader expansion of the company’s AI infrastructure platform.

The initiative adds another dimension to Firmus’ strategy because it connects commercial AI factory development with potential access to accelerated computing for research and nationally relevant applications. For Australia, that could make the location of AI infrastructure manufacturing and access to compute nearly as important as the physical location of the data centers themselves. Firmus’ Malaysian expansion combines Australian-engineered infrastructure, NVIDIA’s next-generation computing architecture and OpenAI demand within a single regional deployment model. If the five sites under development reach service on schedule, the company’s move beyond 900 MW of contracted capacity would translate from a commercial pipeline into a considerably larger operating AI infrastructure footprint across Asia-Pacific.