# ClearCompute blog

Practical writing on running AI inference: GPU prices across clouds, memory sizing for large models, and when self-hosting beats an API. Every number is dated and sourced.

- [What an H100 costs per hour in October 2026, and what the price leaves out](https://clearcompute.cc/blog/h100-cloud-prices.md) (2026-10-05): On 5 October 2026 one H100 listed at $2.69 to $12.29 per GPU-hour across nine clouds, with a $1.80 marketplace median. Where the spread comes from, and the costs the hourly rate leaves out.
- [How much GPU memory does a 70B model need? Fitting it is the easy part](https://clearcompute.cc/blog/gpu-memory-70b-model.md) (2026-10-05): Llama 3.3 70B needs 141 GB of weights at BF16 and 73 GB at FP8, and every token of a user's context adds 320 KiB of KV cache. The arithmetic for how many users fit on an H100, two H100s, an H200 or a B200.
- [Moving off a closed API: when do your own GPUs pay off?](https://clearcompute.cc/blog/self-hosting-vs-api-cost.md) (2026-10-05): We priced the same Llama 3.3 70B traffic three ways on 5 October 2026: a closed API, a per-token open-model API and dedicated H100s. Dedicated GPUs beat Sonnet-class prices above about 4% utilisation; for small or spiky traffic on a widely hosted model, the cheapest per-token hosts still cost less.
