英伟达财报解读:需求远超供应,云积压增至2.5万亿美元
$NVDA earnings transcript + industry read through. To sum it up: "Demand far exc…
英伟达最新财报核心数据汇总,直接反映AI算力产业链供需格局与定价权变化,对判断市场趋势极具参考价值。
$NVDA earnings transcript + industry read through. To sum it up: "Demand far exceeds supply".
英伟达财报电话会议纪要及行业解读。总结来说:"需求远超供给"。
- Good news for the $MU / Samsung memory longs:
- 对$MU / 三星内存多头而言是利好消息:
"Our supply commitments increased from $279B last quarter to $390B, primarily related to the multi-year procurement of memory supply". (from SEC filings, not call)
"我们的供应承诺从上季度的2790亿美元增至3900亿美元,主要与多年期内存采购有关"。(来自SEC文件,非电话会议)
This is a a major leap Q/Q in supply procurement.
这是季度环比供应采购的大幅跃升。
- "We expect revenue to grow approximately 75% year-over-year, up from 70%." (upward revision, likely due to capacity constraints easing)
- "我们预计收入将同比增长约75%,此前预期为70%"。(上调预测,可能由于产能约束缓解)
(new agreements with Kinsus and Samsung Electro-Mechanics support additional ABF substrate capacity)
(与金像电子和三星电机的新协议支持增加ABF基板产能)
- NVIDIA said industry cloud backlog increased from $2 trillion to $2.5 trillion in a single quarter. While GPU rental pricing is continuing to rise dramatically:
- 英伟达表示,行业云积压订单在单个季度内从2万亿美元增至2.5万亿美元。尽管GPU租赁价格持续大幅上涨:
Hopper GPUs: 2x their original pricing B300 GPUs: +60% since introduction Utilization: NVIDIA says its own compute is fully utilized across the hyperscalers and neoclouds it serves.
Hopper GPU:原价的2倍 B300 GPU:自推出以来上涨60% 利用率:英伟达称其服务于超大规模企业和新云厂商的计算资源已完全满载。
- "Our co-packaged optics roadmap remains on track"
- "我们的共封装光模块路线图进展顺利"
Optical networking keeps growing rapidly ($21B+ in quarterly networking), CPO scale out switches has entered volume production. With Lambda earlier announcing production deployment of Nvidia photonics CPO switches.
光网络持续增长迅速(季度网络收入超过210亿美元),CPO扩展交换机已进入量产阶段。Lambda此前宣布部署英伟达光子CPO交换机。
- "Under these agreements, we will earn revenue on the upfront sale of our infrastructure" So there's additional recurring economics coming from Neoclouds.
- "根据这些协议,我们将通过基础设施的前期销售获得收入"。因此,新云厂商带来了额外的经常性经济收益。
But just points to industry acceleration from memory to CPO to neoclouds, relative to markets pricing in collapse.
但这仅指向行业从内存到CPO再到新云厂商的加速发展,而市场却在定价崩溃。
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力