OpenAI Locks in Malaysian Compute Capacity Through Multi-Year Deal with Firmus

OpenAI Secures Dedicated AI Compute from Firmus Malaysia Data Centers

OpenAI has locked in a multi-year deal with Firmus, securing dedicated AI compute capacity from two Malaysia data centers operated by the Australian AI infrastructure company1

1

2

2

. The partnership positions OpenAI as an anchor customer and brings Firmus' total contracted capacity across all customers to more than 900MW4

4

. "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," said Sachin Katti, VP of Compute Strategy at OpenAI2

2

.

Nvidia Hardware Powers Malaysian AI Infrastructure

Firmus will deploy Nvidia's next-generation Vera Rubin NVL72 systems on Nvidia's DSX platform at the Malaysian sites1

1

4

4

. The infrastructure integrates with Firmus' proprietary HyperCube system, which combines liquid cooling, mechanical systems and electrification into prefabricated units manufactured in regional New South Wales2

2

. This Australian-made component represents genuine manufacturing export, even as the hardware itself comes from Nvidia and serves an American customer1

1

. The tokens generated will serve OpenAI users worldwide, positioning Malaysia as a crucial node in global AI data center capacity1

1

.

Timing Aligns with Anticipated $15.5 Billion ASX Listing

The Firmus OpenAI deal arrives as the Nvidia-backed Firmus prepares for investor meetings next week ahead of a highly anticipated IPO1

1

3

3

. The company is working on a pathfinder prospectus for what could be a $15.5 billion Australian sharemarket float later this year3

3

. Firmus was valued at above $10.5 billion in its most recent fundraising round, roughly double its valuation four months earlier1

1

4

4

. The company raised $2 billion with backing from Coatue, Nvidia, Blackstone Tactical Opportunities and Jane Street1

1

. Signing the world's most recognizable AI company as an anchor customer immediately before investor meetings creates a compelling narrative for the ASX listing1

1

.

Contracted Capacity Differs from Operational Reality

While Firmus touts 900MW of contracted capacity, only two of its seven AI factories are currently operational1

1

. The company has facilities in Australia and Singapore running now, with five more under development across the Asia-Pacific region targeted for service within the next 24 months2

2

4

4

. The 900MW figure describes commitments rather than delivered infrastructure, which is standard industry practice for financing but represents the gap where data center projects typically fail1

1

. Firmus' Indonesian site in Batam illustrates the scale: a 360MW campus going live in first quarter 2027 with expected offtake running to $25 billion or $30 billion over six years1

1

. Notably, the OpenAI contract does not include any commitment for Firmus' planned Australian facilities3

3

.

Malaysia Emerges as Strategic AI Infrastructure Hub

Malaysia has become Southeast Asia's fastest-growing data center market, though the rapid buildout has drawn scrutiny over electricity and water use4

4

. The Firmus partnership extends the company's footprint in the region beyond Australia, Singapore and Indonesia2

2

. What Malaysia supplies is land, grid connection and power—a meaningful position in the AI infrastructure supply chain, even if not "production of intelligence" as Firmus co-CEO Tim Rosenfield claimed1

1

. The same day as the OpenAI announcement, Malaysia was reportedly weighing Huawei chips for a sovereign AI project despite Washington's warnings about US export controls1

1

. This dual approach—hosting both American and Chinese silicon—positions Malaysia to sell capacity to both sides while committing to neither, a strategy most governments have failed to execute1

1

.

Deal Reflects Intensifying Race for AI Data Center Capacity

The multi-year deal highlights the widening competition among US and Chinese AI groups to secure power and dedicated AI compute for increasingly advanced models4

4

. OpenAI last week unveiled Astra, which it described as its most capable model yet4

4

. The partnership follows a pattern of large AI labs securing Malaysian compute from Nvidia-backed infrastructure providers—Anthropic has signed similar agreements with Volta Infra Holdings and pursued multiple large computing arrangements as demand for its products has grown2

2

. Firmus would join a queue of AI companies pursuing public listings: Anthropic is expected to list within weeks, OpenAI has pointed at 2027, and Moonshot AI is seeking a Hong Kong listing1

1

. Alongside the OpenAI partnership, Firmus announced plans to establish an Australian AI Access Program providing computing access to researchers and organizations working on applications in science, education, agriculture, energy efficiency and climate resilience2

2

.