Singapore-headquartered AI cloud and infrastructure provider Bitdeer AI announced last week that it has sold its 9.5MW capacity at its A102 AI cloud data center in Malaysia ahead of energization in Q1 2027. Together with an offtake commitment announced earlier in August, the firm now has in its hand five-year long-term offtake commitments with total expected revenue of over US$800 million at this Johor site.
The A102 facility is a liquid-cooled, multi-customer facility purpose-built for rack-scale NVIDIA GB300 NVL72 deployments, capable of delivering both GPU cloud services and data hosting from a single site. It represents 9.5MW of the 350MW target of its global AI cloud capacity to be delivered by the first quarter of 2028. It’s among the first of these sites to have contracted capacity, Bitdeer Ai said in a press release last week.
These contracts are not expected to have a revenue impact in 2026. Revenue and associated costs are expected to begin in the first quarter of 2027, when services commence. Contract value per megawatt varies with service mix, contract duration and site.
The Malaysia build-out advances Bitdeer AI’s strategy of developing high-density AI infrastructure in markets with strong power availability and proximity to fast-growing enterprise AI demand, it added.
Demand for uncontracted capacity at Bitdeer AI’s other AI Cloud sites remains strong. Bitdeer AI’s active pipeline for AI cloud capacity now exceeds US$ 2 billion, and the firm is now concentrating on its 21.7MW Johor Bahru site in Malaysia as well as capacity at other sites.
The platform funds its development through customer prepayments, operating cash flow, and financing secured against contracted cash flows. Its AI cloud contracts usually provide for customer prepayments to cover more than 50 per cent of the associated capital expenditure. It develops data center capacity in line with contracted demand.
Bitdeer AI, which is part of Nasdaq-listed Bitdeer Technologies Group, expects the pipeline for AI Cloud capacity to increase in the next several quarters. The Group features footprint across the U.S., Norway, Bhutan, Canada, and Malaysia, with a target of up to 350MW of AI-ready data center capacity to be delivered by the first quarter of 2028.