DeepSeek R1 — GB300 NVL72 vs H100 Performance per Dollar
Cost per million tokens of GB300 NVL72 (NVIDIA Blackwell) versus H100 (NVIDIA Hopper) on DeepSeek R1. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
On DeepSeek R1 at 40 tok/s/user, the per-million math comes out to $0.08 for GB300 NVL72 and $1.20 for H100; GB300 NVL72 delivers 1387% more output per dollar.
At 70 tok/s/user on DeepSeek R1, GB300 NVL72 costs $0.18 per million tokens; H100 costs $4.19. GB300 NVL72 is 2191% more cost-efficient at this operating point.
GB300 NVL72 edges H100 at 99 tok/s/user on DeepSeek R1 — $0.28 per million tokens versus $14.1, a 4867% cost-per-token gap. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Chip pricing (owning hyperscaler): GB300 NVL72 $2.31/chip/hr · H100 $1.17/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | GB300 NVL72:$0.081H100:$1.205 | GB300 NVL72:$0.183H100:$4.189 | GB300 NVL72:$0.283H100:$14.069 |
| Concurrency | GB300 NVL72:~3186H100:~605 | GB300 NVL72:~977H100:~57 | GB300 NVL72:~577H100:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.