Event ·Multi-source

黄仁勋：模型推理速度翻倍等于收入翻倍

NVIDIA CEO 黄仁勋提出，通过软件或 kernel 优化把模型推理速度提升 2 倍，在不新增 GPU、电力和机房投入的情况下，AI 实验室的收入可以翻倍。其逻辑是推理收入等于服务 token 数乘以 token 单价，更快的 kernel 在同样硬件上每秒产出更多 token。前提条件是需求持续超过供给，每个新增 token 都能按现价卖出去。

Current read

黄仁勋算了一笔账：不改硬件、只靠 kernel 优化把推理提速 2 倍，token 卖得掉的话收入就翻倍，适合做 AI 业务的人看看这笔账怎么算。

11 sources66° AI heatLast updated 2026/10/28 22:28:51

Evidence

11 sources

Primary sourcerohanpaul_ai

2026/10/28 22:28:51View source

Related source 1techcrunch

2026/10/27 11:00:00View source

Related source 2Tech in Asia

2026/10/27 01:25:42View source

Related source 3Nikkei Asia

2026/10/27 06:15:06View source

Related source 4arXiv cs.AI

2026/10/27 08:46:21View source

Related source 5Decoder

2026/10/27 09:58:16View source

Related source 6Genk

2026/10/27 11:45:00View source

Related source 7CNBC Indonesia Tech

2026/10/27 14:00:00View source

Related source 8IT之家

2026/10/28 00:09:31View source

Related source 9Dylan Patel (SemiAnalysis)

2026/10/28 00:57:13View source

Related source 10The Information

2026/10/28 13:23:09View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story