Bangladesh has launched its first government-run, shareable cloud computing facility powered by high-performance graphics processing units (GPUs), marking a major step in advancing higher education, research, and machine learning-based skills across the country.
The initiative was announced by Faiz Ahmad Taiyeb, special assistant to the chief adviser at the Ministry of Posts, Telecommunication and Information Technology, in a Facebook post on January 15. He said the Bangladesh Hi-Tech Park Authority has integrated more than 20 NVIDIA Volta Architecture Tensor Core GPUs into a cloud resource management platform, making the facility operational.
It is the first shared GPU cloud of its kind in Bangladesh’s public sector. The infrastructure will be maintained by the National Data Centre team of the Bangladesh Computer Council.
GPUs are specialised processors designed for massive parallel computations and are significantly more efficient than traditional CPUs for artificial intelligence and deep learning tasks. According to the announcement, a single GPU delivers computing power equivalent to around 45 CPUs, giving the platform a combined capacity of more than 900 CPU-equivalent processing power.
The system offers up to 2,240 teraFLOPS of deep learning capability. A teraFLOPS measures computing speed and represents one trillion floating-point calculations per second, a standard benchmark for AI and supercomputing performance.
The facility will support a wide range of high-performance computing activities, including machine learning dataset training, threat intelligence analysis, and geoscience modelling. It will enable users to run small-scale AI simulations, train models, and conduct inference tests to evaluate how trained models perform on new data.
Universities, research institutions, organisations, and individuals interested in using the platform have been invited to express their interest by emailing da********@*****ov.bd.
