NVIDIA and Firmus partner on AI infrastructure for Batam data center

Firmus will use NVIDIA infrastructure to sell cloud services to AI-native companies, enterprises, and ISVs through its Batam AI Factory campus.

Firmus Technologies has entered into a partnership with NVIDIA that will allow Firmus to procure NVIDIA infrastructure and sell NVIDIA-powered cloud services for AI-native companies, enterprises, and independent software vendors.

The agreement covers up to 170,000 NVIDIA AI accelerators across the Grace Blackwell, Vera Rubin, and Vera platforms through 2027 and 2028. The infrastructure will support Firmus' AI Factory campus in Batam, Indonesia, which the company is developing with DayOne, a Singapore-headquartered digital infrastructure platform.

Firmus expects to receive between US$25 billion and US$30 billion from committed offtake agreements during the first six years of the partnership, based on existing customer commitments.

McKinsey estimated that meeting global AI-related data center demand alone would require US$5.2 trillion in data center investment by 2030, based on projected AI-related capacity demand of 156 GW.

Cloud revenue model

The partnership uses a revenue-sharing and credit-support model. Under the arrangement, Firmus will sell cloud services powered by NVIDIA infrastructure, while NVIDIA will receive product revenue and a share of cloud revenue from the supported capacity.

Firmus said the structure is designed to give AI-native companies, enterprises, and ISVs access to large-scale AI compute without requiring them to contract directly for capital-intensive infrastructure at the same scale.

Tim Rosenfield, co-CEO of Firmus Technologies, said AI-native companies need access to scalable, energy- and cost-efficient compute infrastructure to compete globally. He said the partnership gives those customers access to advanced AI accelerators with greater certainty, scale, and flexibility.

Cooling and efficiency

The Batam campus will use NVIDIA DSX, NVIDIA's full-stack AI factory platform, together with Firmus' HyperCube platform. HyperCube is a liquid-cooled AI Factory architecture developed in Australia and co-designed to NVIDIA DSX blueprints.

Dense accelerator systems require more cooling capacity than conventional server environments. NVIDIA says its liquid-cooled GB200 NVL72 rack design increases compute density, reduces floor space requirements, and supports high-bandwidth, low-latency GPU communication.

NVIDIA also says GB200 can deliver 25 times more performance at the same power compared with H100 air-cooled infrastructure.

NVIDIA DSX provides a reference architecture for designing, simulating, and operating AI factory infrastructure. Firmus said the integration will allow the campus to operate under a common architecture, with the aim of improving deployment speed, energy efficiency, and resiliency.

The companies said the combined platform is designed to improve tokens per watt and reduce cost per token for customers using the Batam campus. Tokens per watt measures how much AI model output is generated for each unit of energy consumed.

Firmus' regional capacity

The Batam project follows other Firmus AI infrastructure deployments. In March 2026, Firmus said it had signed a long-term contract for dedicated AI infrastructure capacity at Project Southgate in Australia, involving about 18,400 NVIDIA GB300 GPUs.

The agreement expands Firmus' planned AI Factory capacity in Asia-Pacific.