Bramcott

Compute economics

What accelerator capacity
actually costs.

Bramcott prepares scoped price quotations for GPU fleets: the configuration priced, the term options, the charges that sit beside compute, and the assumptions every figure depends on. Each quotation is a document you can put in front of a procurement review.

Request a fleet quote   See the illustrative quotation documents

  1. 128× L40S inference fleet priced as a committed and on-demand hybrid

    L40S × 128 · air cooled · cost per million tokens

    Hybrid: 96 committed, 32 on demand

    Sample

    sample-quote-l40s-inference-fleet · Valid to 15 Dec 2026

  2. 1,024× B200 rack-scale NVLink domains on a three-year commitment

    B200 × 1,024 · 16 NVLink domains · direct-to-chip liquid cooled

    3-year committed, phased delivery

    Sample

    sample-quote-b200-1024-gpu-nvl-domain · Valid to 31 Jan 2027

  3. 64× H200 fine-tuning and research fleet on a monthly rolling term

    H200 × 64 · InfiniBand NDR · 8 nodes

    Monthly rolling, 30 days’ notice

    Sample

    sample-quote-h200-64-gpu-fine-tuning · Valid to 30 Nov 2026

  4. 256× H100 SXM5 training cluster with one-year and three-year alternatives

    H100 SXM5 × 256 · InfiniBand NDR · 32 nodes

    Monthly, 1-year and 3-year priced side by side

    Sample

    sample-quote-h100-256-gpu-training-cluster · Valid to 31 Dec 2026