GPU · 1000 TFLOPS dense BF16
H100 SXM fine-tuning cost estimates
All sixteen scenarios on the H100 SXM: GPU-hours, wall-clock on 8 GPUs and the cost range across 6 providers, at 35% MFU. Prices read 12 November 2026. Estimates, not quotes.
The figure
Where 1000 TFLOPS comes from
“Peak BF16 Tensor Core1 · 1000 TFLOPS · 2000 TFLOPS2 · 800 TFLOPS · 1600 TFLOPS2”
Printed with: “2. Effective TFLOPS / TOPS using the Sparsity feature”
Printed with: “1. Preliminary performance estimates for H100 based on current expectations and subject to change in the shipping products”
The maker labels this figure preliminary. Shipping hardware may differ.
Scenarios
Cost by scenario
| Scenario | GPU-hours | 8 GPUs | Lowest | Median | Highest |
|---|---|---|---|---|---|
| 7B LoRA fine-tune on 100 million tokens | 2.22 | 17 min | $6.56 | $7.04 | $11 |
| 7B LoRA fine-tune on 1 billion tokens | 22 | 2.8 h | $66 | $70 | $113 |
| 7B full fine-tune on 100 million tokens | 3.33 | 25 min | $9.83 | $11 | $17 |
| 7B full fine-tune on 1 billion tokens | 33 | 4.2 h | $98 | $106 | $170 |
| 13B LoRA fine-tune on 100 million tokens | 4.13 | 31 min | $12 | $13 | $21 |
| 13B LoRA fine-tune on 1 billion tokens | 41 | 5.2 h | $122 | $131 | $210 |
| 13B full fine-tune on 100 million tokens | 6.19 | 46 min | $18 | $20 | $31 |
| 13B full fine-tune on 1 billion tokens | 62 | 7.7 h | $183 | $196 | $315 |
| 34B LoRA fine-tune on 100 million tokens | 11 | 1.3 h | $32 | $34 | $55 |
| 34B LoRA fine-tune on 1 billion tokens | 108 | 13 h | $318 | $342 | $549 |
| 34B full fine-tune on 100 million tokens | 16 | 2.0 h | $48 | $51 | $82 |
| 34B full fine-tune on 1 billion tokens | 162 | 20 h | $478 | $513 | $824 |
| 70B LoRA fine-tune on 100 million tokens | 22 | 2.8 h | $66 | $70 | $113 |
| 70B LoRA fine-tune on 1 billion tokens | 222 | 28 h | $656 | $704 | $1,130 |
| 70B full fine-tune on 100 million tokens | 33 | 4.2 h | $98 | $106 | $170 |
| 70B full fine-tune on 1 billion tokens | 333 | 42 h | $983 | $1,057 | $1,696 |
Prices used
H100 SXM price by provider
| Provider | $/GPU-hr | Configuration | Source |
|---|---|---|---|
| Thunder Compute | $2.95 | 1× SXM | price page |
| Civo | $2.99 | 1× SXM | price page |
| Massed Compute | $3.14 | 1× SXM | price page |
| Hyperstack | $3.20 | 1× SXM | price page |
| RunPod | $3.99 | 1× SXM | price page |
| Verda | $5.09 | 1× SXM | price page |
Each provider's lowest listed on-demand configuration for this GPU. Marketplace figures are left out.
Questions
H100 SXM: common questions
What is the lowest cost scenario on the H100 SXM?
7B LoRA fine-tune on 100 million tokens: about $6.56 on Thunder Compute, 2.2 GPU-hours.
What throughput is assumed?
1000 TFLOPS dense BF16 as printed by the maker, times 35% MFU, so about 350 TFLOPS of useful work.
How many providers list it?
6 with a current provider-set on-demand price. Runcrate has no live listing for it.
Related