Bitdeer AI adds 60MW Malaysia facility, reaches 95% of 350MW target
Bitdeer AI, the neocloud arm of NASDAQ-listed Bitdeer Technologies Group, has entered into an 8-year data centre services agreement for A103, a 60MW AI compute facility on its existing Cyberjaya campus in Malaysia. The deal takes the company's total secured AI Cloud capacity to approximately 333.5MW across sites in Malaysia, Norway and the United States, representing roughly 95% of its stated target of up to 350MW to be delivered by the first quarter of 2028.
A103 is designed for liquid-cooled, rack-scale NVIDIA systems and will be capable of hosting the next-generation NVIDIA Vera Rubin platform when it becomes available. Energisation is expected in the first quarter of 2028. Because A103 sits alongside the already-operational A101 and A102 buildings on the same Cyberjaya site, Bitdeer AI says it can draw on existing power, cooling and networking infrastructure, shortening the time to bring the facility online. Together, the three Cyberjaya buildings will account for approximately 71.5MW of the total campus capacity.
The deal
Michael G. Potter, chief financial officer of Bitdeer Technologies Group, said the co-location with A101 and A102 improves the company's time-to-market. "A103 is also built for the Vera Rubin generation, which is the capacity customers are asking us about for 2028," he added.
Bitdeer AI describes its active pipeline for uncontracted AI Cloud capacity as exceeding $10 billion. The company is careful to note, however, that this figure represents management's estimate of potential contract value under commercial discussion, not executed backlog or binding commitments. No offtake contracts for A103 had been signed as of the release date, and the company has issued standard cautionary language around whether pipeline conversations will convert to deals on the expected timeline.
Market context
The announcement arrives at a moment of intense competition among GPU cloud providers. Hyperscalers such as AWS, Microsoft Azure and Google Cloud are all scaling AI inference and training capacity aggressively, while a cohort of well-capitalised neoclouds, including CoreWeave, Lambda Labs and Crusoe Energy, are racing to secure long-term leases and NVIDIA allocation to service enterprise AI demand. Bitdeer AI's positioning as a preferred NVIDIA Cloud Partner gives it preferential access to hardware allocation, which remains a meaningful constraint across the sector as NVIDIA's Blackwell and forthcoming Vera Rubin platforms are in high demand.
Malaysia has become an increasingly important geography for AI infrastructure investment, drawing data centre operators with its relatively low-cost land and power, a stable political environment, and government incentives for digital infrastructure. The Cyberjaya technology corridor in particular has attracted significant hyperscaler and co-location investment over the past three years, giving Bitdeer AI access to an established ecosystem of power and network providers.
Regulatory and standards read-across
Operators building at scale in Malaysia must navigate the Malaysian Communications and Multimedia Commission's licensing framework as well as data-residency requirements that vary by customer jurisdiction. For Bitdeer AI's anticipated 2028 customer base, EU AI Act obligations around compute traceability and the US Commerce Department's ongoing export-control regime governing advanced AI chips are relevant considerations, particularly for customers with cross-border data flows. The company has not disclosed the nationality profile of its customer pipeline, which would affect which regulatory frameworks apply at the point of contract.
With approximately 16.5MW of capacity still to secure before hitting the 350MW ceiling, Bitdeer AI's next announcement is likely to involve either a fourth facility or an expansion of an existing site. Investors will be watching for the first named customer commitment on A103 and confirmation of NVIDIA Vera Rubin delivery schedules as the key milestones before energisation.