OpenAI signs with Firmus in Malaysia as total tops 900 MW

Firmus signed a multi-year strategic agreement with OpenAI on 8 September 2026 in Sydney to supply dedicated artificial intelligence compute capacity from two data center sites in Malaysia. The transaction establishes OpenAI as an anchor customer across the infrastructure developer’s computing platform. This commercial arrangement officially elevates the total contracted volume of Firmus across all global clients beyond 900 MW.

OpenAI will secure computing power from the planned Malaysian facilities to sustain processing requirements for its commercial systems across the Asia-Pacific territory and worldwide. Sachin Katti, vice president of compute strategy at OpenAI, confirmed that the deployment addresses expanding market appetite for the developer’s tools.

Capacity surpassed 900 MW.

Strategic computing agreement in Malaysia

The long-term contract secures capacity from two specific Firmus AI Factory sites situated in Malaysia as part of an ongoing regional development program. Tim Rosenfield, co-founder and co-chief executive officer at Firmus, noted that the enterprise spent seven years constructing the specialized manufacturing, regional engineering footprint, and technical workforce needed to deliver large-scale computational infrastructure at rapid commercial speed. Under this operational structure, the facilities will supply raw hardware power to handle high-density artificial intelligence training and continuous operational execution. Rosenfield stated: “This multi-year partnership marks the moment Asia-Pacific becomes a producer of intelligence, not just a consumer of it.”

OpenAI intends to rely on the dedicated server clusters to maintain service delivery for its core digital products. Katti said: “These new data centers in Malaysia will help us serve growing demand for OpenAI’s products across the region and around the world.”

Operational footprint across the Asia-Pacific region

Firmus manages an asset base that spans seven dedicated computational hubs across four separate jurisdictions comprising Australia, Singapore, Indonesia, and Malaysia. Two data centers currently deliver active commercial processing in Australia and Singapore, while five additional infrastructure projects remain under direct construction. The five facilities in the development pipeline are targeting full operational readiness over the next 24 months.

The expansion covers four nations.

• Two facilities operating today in Australia and Singapore
• Five facilities under construction across Australia, Singapore, Indonesia, and Malaysia
• Target completion within 24 months for all pending builds
• Total customer capacity commitments exceeding 900 MW across the network

The entry into Malaysia extends the physical operating boundary of Firmus beyond its initial locations established in Australia, Singapore, and Indonesia. Company representatives stated that the multi-country platform provides the operational foundation required to support further regional server installations.

Technical integration with NVIDIA hardware

Firmus will deploy NVIDIA Vera Rubin hardware at scale across the Asia-Pacific geography to introduce the manufacturing partner’s next-generation accelerated computing architecture. The infrastructure developer integrates NVIDIA Vera Rubin NVL72 rack-scale systems into its proprietary HyperCube platform by utilizing the standardized NVIDIA DSX AI Factory Platform. Regional New South Wales serves as the dedicated industrial manufacturing location where Firmus fabricates and preassembles the physical HyperCube modules, combining mechanical equipment, primary electrical systems, and specialized liquid cooling technology into a uniform structural design before dispatching the components to target data center construction sites. This physical prefabrication model aims to improve installation velocity, optimize deployment capital efficiency, and maintain system performance while driving down the overall cost of token production.

NVIDIA corporate leadership indicated that modern regional infrastructure requires engineered full-stack facilities built specifically for durability, high-density performance, scalable capacity, and fungible processing workloads. Nico Caprez, vice president of global AI infrastructure growth at NVIDIA, stated that the technical architecture combines NVIDIA DSX and NVIDIA Vera Rubin NVL72 hardware to accelerate operational deployment schedules and support regional industrial token economics.

Caprez confirmed that the modular builds enable faster system delivery across target territories.

Research initiative planned for Australia

Alongside the commercial deployment for OpenAI, Firmus plans to launch the Australian AI Access Program to provide computing resources to qualified domestic researchers and institutions. The initiative targets computational projects focused on issues of national and regional priority across Australia. Participating organizations will deploy computational power toward specialized work in science, educational methods, agricultural techniques, energy efficiency, resource management, and climate resilience.

Firmus Grid Limited, which trades under the commercial name Firmus Technologies, operates corporate headquarters in Sydney and manages artificial intelligence cloud operations across Asia-Pacific territories. The enterprise maintains an engineering model centered on direct optimization between utility power intake and computational token output.