AI 导读
DeepSeek 开源推理框架 DeepInfer，用于 DeepSeek V4.1 Flash 的大规模推理服务，单集群吞吐达 155 万亿 token/天，按非高峰价计算每卡每小时收入 $38。作者指出 DeepInfer 同时支持 NVIDIA Blackwell 和华为昇腾，并估算即使按 $7/GPU 小时租用 B300，毛利率也约 82%。
推荐理由
作者在转述数字之外补充了毛利测算，读者可以借此理解 DeepSeek 推理服务的成本结构。
正文
“A single cluster currently serves 155 trillion tokens/day, generating $38 in revenue per GPU-hour (at off-peak pricing).”
Huge flex from DeepSeek. Not only did they open-source their inference engine, but they also open-sourced their revenue numbers.
They didn't disclose which GPUs they're using here, but DeepInfer supports both NVIDIA Blackwell and Huawei Ascend.
Even assuming they're renting B300s at $7 per GPU-hour, that's ~82% gross margin.
来源：Yuchen Jin · x.com