September 8, 2026, (Inside AI) — Australia's Firmus has signed a multi-year agreement with OpenAI to provide computing capacity from two data centres in Malaysia. The deal positions the ChatGPT maker as an anchor customer for the facilities.
The agreement reflects a broader scramble for AI infrastructure across Southeast Asia. Demand for advanced model training and inference has outpaced supply in major hubs like Singapore and Tokyo. Malaysia offers lower energy costs and available land, making it an attractive alternative.
Firmus is backed by Nvidia, which has invested in the company to expand its data centre footprint. The two Malaysian sites will host Nvidia GPUs, likely including H100 or H200 accelerators. OpenAI will use the capacity to run workloads for its frontier models.
The deal gives OpenAI a foothold in a region where it has been expanding cloud partnerships. It also reduces reliance on Microsoft Azure, OpenAI's primary compute provider. Industry analysts have noted that OpenAI has been diversifying its infrastructure suppliers to avoid bottlenecks.
Malaysia has attracted several data centre investments in the past two years. Johor state, bordering Singapore, has become a hotspot for hyperscale facilities. Firmus has not disclosed the exact locations of the two centres or the total megawatt capacity involved.
Neither company revealed the financial terms or the duration of the contract beyond saying it is multi-year. A Firmus spokesperson said the agreement validates the company's build-out strategy in the region.
OpenAI has been under pressure to secure compute for its next-generation models. Training runs for models like GPT-5 require clusters of tens of thousands of GPUs. Long-term supply deals have become essential to maintain development timelines.
Malaysia's government has courted AI infrastructure investment through tax incentives and streamlined approvals. The country aims to become a regional data hub, competing with Indonesia and Thailand. Local energy providers have been upgrading grid capacity to support new facilities.
Firmus previously operated data centres in Australia and has been expanding into Asia. The Nvidia backing gives it access to scarce GPU allocations, a key advantage in a market where supply remains tight.
The deal could accelerate OpenAI's deployment of services in Southeast Asia. Lower latency from regional data centres improves performance for users in Malaysia, Indonesia, and Singapore. It may also help OpenAI meet data residency requirements in some markets.
Industry observers expect more anchor tenant deals as AI companies lock in capacity years ahead. Hyperscalers like Google and Amazon have also signed long-term agreements with data centre operators in the region.
Firmus said construction on the two Malaysian sites is underway, with initial capacity expected to come online within the next 12 to 18 months. OpenAI will begin using the facilities once they are operational.
