Unable to Access High-End GPUs (A100/H100) Despite Research Credits

[Barbhuiya12]

Hi team,

I received a research grant of approximately 1,000 GCP credits, but I am currently unable to access high-end GPUs such as A100 or H100. My requests for these GPUs have been declined repeatedly for over a month, and I have not received a clear explanation or resolution.

Use case

Training a large-scale deep learning model (~100 million parameters)

Dataset size: multiple terabytes

The training workload is research-critical and time-sensitive

Current limitation

Only L4 GPU access (24 GB VRAM) has been granted

This is not sufficient for the workload:

One training epoch takes approximately 12 hours

Full training would take several weeks or longer

Memory constraints significantly limit batch size and training efficiency

Questions and request for guidance

What are the approval criteria for accessing A100 or H100 GPUs under research credits?

Is there a formal justification or review process that I should follow?

Are there any alternative programs, quota increases, or temporary approvals available for research workloads of this scale?

If access to A100/H100 GPUs is not possible, are there recommended configurations or optimizations to make training feasible?

At present, this research cannot realistically progress using only L4 GPUs. Any clarification, guidance, or escalation path would be greatly appreciated.