
NVIDIA L4
৳147*/hrFast inference, video AI, efficient fine-tuning
ComputeBD is validating demand for on-demand AI computing where builders choose an accelerator and pay around the AI Computer hours they actually need.
AI Computer requirements change from project to project. Owning a single accelerator—or leaving expensive hardware idle—may not fit teams with variable training, inference and research demand.
ComputeBD aims to make the accelerator, requested hours and indicative BDT cost visible in a simple Jupyter-first workflow.

Fast inference, video AI, efficient fine-tuning

Rendering, vision, medium training workloads

Generative AI, rendering, larger fine-tunes

LLM fine-tuning, research, high-throughput training

Large model training, transformer workloads, HPC

Memory-heavy LLMs, retrieval, frontier research

Next-gen Blackwell: large-LLM training and the fastest inference

Most powerful: 288 GB for trillion-parameter models and reasoning AI
Yes. From Dhaka or anywhere in Bangladesh you rent ComputeBD GPUs in your browser — there is no hardware to collect or return.
On ComputeBD an NVIDIA H100 is about ৳1,047 per hour and an A100 about ৳638, billed by the minute in taka.
Billing is by the hour, so you pay for the hours you run. Monthly examples for each GPU are on the pricing page and each GPU page.
For most people, rent. Our H100 cost benchmark shows how many thousand rented hours the price of one H100 buys.