GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 —
GROK-46H 865 —
GEM-37FH 865 —
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 —
GPT-6A 820 —
KIMI-K3X 810 —
CL-FAB5H 787 —
CL-OP5H 764 —
CL-OP46H 742 —
CL-OP47H 733 —
GEM-38FH 676 —
CL-OP47 583 -0.7%
INKL 531 —
CL-OP46 496 -0.2%
CL-OP48 490 -0.2%
← Back to feed

Firmus Crosses 900 MW Contracted With OpenAI as Anchor, Deploys Vera Rubin in Malaysia

Firmus Technologies, the Sydney-headquartered AI infrastructure company, has added OpenAI as an anchor customer and expanded its footprint into Malaysia, pushing total contracted AI factory capacity past 900 megawatts. The partnership involves dedicated compute capacity across two Malaysian AI Factory sites with deployment of NVIDIA Vera Rubin NVL72 rack-scale systems.

Firmus now operates or is developing 7 AI factories across 4 countries: Australia and Singapore are operational today; Indonesia and Malaysia join as development-stage sites targeting readiness within 24 months.

The Deal Structure

OpenAI is contracting dedicated AI compute capacity, not purchasing or co-owning sites. Sachin Katti, OpenAI’s VP of Compute Strategy, described the partnership as part of expanding global compute capacity to serve “growing demand for OpenAI’s products across the region.”

The Malaysian sites will run on the NVIDIA DSX AI Factory Platform, integrating Vera Rubin NVL72 rack-scale systems with Firmus’s proprietary HyperCube platform. HyperCube is a prefabricated infrastructure system that integrates liquid cooling, mechanical systems, and electrification into a repeatable format, manufactured in regional New South Wales.

Why This Is Firmus’s Most Significant Partnership Yet

Firmus has been building toward a model of vertically integrated AI factories: “energy in and tokens out.” The OpenAI anchor validates that model at scale. Before this deal, Firmus’s public partnerships included NVIDIA’s revenue-share arrangement announced in July 2026 (which covered 210,000 GPUs) and earlier collaborations in Australia and Singapore. Adding OpenAI, the world’s most compute-hungry AI lab, as a dedicated customer changes the company’s fundraising and capacity negotiating position materially.

The 900 MW contracted figure is a portfolio number across all customers, not just OpenAI. The addition of the OpenAI partnership was what pushed Firmus past that threshold.

Vera Rubin at the Edge of the Asia-Pacific Buildout

NVIDIA’s Vera Rubin NVL72 is the next-generation rack-scale system, representing a significant step beyond Blackwell. Its deployment in Malaysian sites positions Firmus to offer frontier compute to regional enterprise, government, and AI-lab customers. Nico Caprez, NVIDIA’s VP of Global AI Infrastructure Growth, said the HyperCube-based factories would “bring advanced capacity online faster and produce intelligence at industrial scale with stronger token economics.”

Firmus also intends to establish an AI Access Program for researchers and organisations applying AI to national and regional priorities, including science, education, agriculture, energy, and resource efficiency.

Key Numbers

MetricValue
Total contracted capacity900+ MW (all customers)
Operational sitesAustralia, Singapore
New development sitesMalaysia (2), Indonesia (and existing dev)
HardwareNVIDIA Vera Rubin NVL72
PlatformNVIDIA DSX + Firmus HyperCube
HyperCube manufacturingRegional New South Wales, Australia
Target readiness (new sites)Within 24 months

The pace of Asia-Pacific AI factory construction is now clearly in gigawatt territory. Firmus is building the argument that vertically integrated infrastructure, manufactured locally and financed by anchor tenants, is a viable alternative to US-centric hyperscaler capacity for the region.