原文：
So Jensen himself went out to do damage control with $NVDA delays reports apparently:
1. Jensen Huang said no delays:
- 800V and rack-to-rack optical interconnects is progressing fully in line with the original plan with no material delay.
- Debunked the rumor that the next-generation flagship, Rubin Ultra, would be delayed until 2028 and said it would be on track to ship next year
- Rack system optimizations, like replacing Kyber with a new design are architectural optimizations, not impacting delivery timelines.
2. On ASIC competition: no loss of market share.
- Talked about how a leading frontier lab that previously relied heavily on internal ASICs has shifted nearly 50% of its compute load to Nvidia
3. On over-reliance on hyperscalers.
- AI frontier labs currently contributing to ~20% of demand.
- Traditional cloud providers, Nvidia provides Vera CPUs and high-speed networking
- Sovereign AI
翻译：
所以詹森·黄本人亲自出面进行损害控制，针对英伟达延迟的报道：
1. 詹森·黄表示没有延迟：
- 800V 和机架到机架的光互连技术正在按照原计划完全按部就班，没有重大延迟。
- 否认了下一代旗舰产品Rubin Ultra将延迟到2028年发布的传闻，并表示它将按计划明年发货
- 机架系统优化，例如用新设计替换Kyber，这些是架构优化，不会影响交付时间表。
2. 关于ASIC竞争：没有失去市场份额。
- 谈到一个领先的前沿实验室，之前严重依赖内部ASIC，现在已经将其计算负载的近50%转移到了英伟达
3. 关于对超大规模云计算服务商的过度依赖。
- AI前沿实验室目前贡献了需求的约20%。
- 传统云服务商，英伟达提供Vera CPU和高速网络
- 主权AI
So Jensen himself went out to do damage control with $NVDA delays reports apparently:
1. Jensen Huang said no delays:
- 800V and rack-to-rack optical interconnects is progressing fully in line with the original plan with no material delay.
- Debunked the rumor that the next-generation flagship, Rubin Ultra, would be delayed until 2028 and said it would be on track to ship next year
- Rack system optimizations, like replacing Kyber with a new design are architectural optimizations, not impacting delivery timelines.
2. On ASIC competition: no loss of market share.
- Talked about how a leading frontier lab that previously relied heavily on internal ASICs has shifted nearly 50% of its compute load to Nvidia
3. On over-reliance on hyperscalers.
- AI frontier labs currently contributing to ~20% of demand.
- Traditional cloud providers, Nvidia provides Vera CPUs and high-speed networking
- Sovereign AI
翻译：
所以詹森·黄本人亲自出面进行损害控制，针对英伟达延迟的报道：
1. 詹森·黄表示没有延迟：
- 800V 和机架到机架的光互连技术正在按照原计划完全按部就班，没有重大延迟。
- 否认了下一代旗舰产品Rubin Ultra将延迟到2028年发布的传闻，并表示它将按计划明年发货
- 机架系统优化，例如用新设计替换Kyber，这些是架构优化，不会影响交付时间表。
2. 关于ASIC竞争：没有失去市场份额。
- 谈到一个领先的前沿实验室，之前严重依赖内部ASIC，现在已经将其计算负载的近50%转移到了英伟达
3. 关于对超大规模云计算服务商的过度依赖。
- AI前沿实验室目前贡献了需求的约20%。
- 传统云服务商，英伟达提供Vera CPU和高速网络
- 主权AI
Comments
Unlock