DeepSeek V4 Pro 1.6T — GB300 NVL72 vs MI355X Performance per Dollar
Cost per million tokens of GB300 NVL72 (NVIDIA Blackwell) versus MI355X (AMD CDNA 4) on DeepSeek V4 Pro 1.6T. 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.
Near the low end of the 13–133 tok/s/user interactivity band — at 43 tok/s/user — GB300 NVL72 runs $0.05 per million tokens on DeepSeek V4 Pro 1.6T while MI355X runs $0.13. GB300 NVL72 is the cheaper choice by 138%.
On DeepSeek V4 Pro 1.6T at 73 tok/s/user, the per-million math comes out to $0.07 for GB300 NVL72 and $0.25 for MI355X; GB300 NVL72 delivers 260% more output per dollar.
At 103 tok/s/user on DeepSeek V4 Pro 1.6T, GB300 NVL72 costs $0.10 per million tokens; MI355X costs $0.52. GB300 NVL72 is 429% more cost-efficient at this operating point. (Numbers reflect the default 8k/1k · fp4 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 · MI355X $1.50/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:$0.055MI355X:$0.130 | GB300 NVL72:$0.069MI355X:$0.249 | GB300 NVL72:$0.099MI355X:$0.523 |
| Concurrency | GB300 NVL72:~4613MI355X:~149 | GB300 NVL72:~1206MI355X:~29 | GB300 NVL72:~587MI355X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.