1) Jensen once again highlights his view that Moore's Law no longer delivers the rate of innovation that's needed and causes computation inflation in Data Center. As such, the rebuilding of $1T data center infrastructure that's currently installed will drive growth over the next 10 years, bringing material performance improvement and cost savings
2) ROI: Jensen reiterates that while using a GPU to augment a CPU will drive an increase in cost in nominal terms (roughly 2x), in the case of Spark (a distributed processing system and analytics engine for big data), the net cost benefit could be as large as 10x as the application will be sped up 20x.
From a top line generation, he shares that hyperscale customers can generate $5 in rental revenue for every $1 spent on NVDA GPUs, given sustained strength in the demand for accelerated computing.
3) With respect to Nvidia's moat, Jensen highlights that the large install base of NVDA GPUs across multiple platforms, ability to augment their hardware with their proprietary software, and ability to build rack-level systems and innovate across various chips creates an insurmountable moat.
4) Jensen once again compliments his supply chain partners, highlighting TSMC for being the reason for NVDA's exponential revenue growth in 2H23/2024. That being said, he did note that NVDA has adequate in-house IP to shift manufacturing without material disruption if needed. He also stressed that companies are relentlessly demanding new chips so they can be first, to the point where tensions are rising as they demand their chips now.
5) The most important part: Jensen reiterated that the company will begin shipping Blackwell chips in 4Q25 and that shipments will scale in FY26 (Note that NVDA is currently in 3Q25 for their accounting calendar). Demand for Blackwell is incredibly strong and he feels that NVDA has a personal responsibility to ensure all companies are able to obtain the supplies needed to develop AI
Leave a Reply