To The Moon

Nvidia $NVDA Q2 Conference Call Highlights

Blackwell should start shipping in Q4 (billions of dollars). Jensen sounds confident that any issues were largely solved.

Computing requirements could be doubling every year. This supports the argument that hyperscale customers have to keep buying GPUs.

Next generation LLM models could require 20, 30, 40x more compute than current generation (!!!!)

Lots of CPU tasks are just now being migrated to GPU. Things like ad targeting, data processing, search results, etc.

Demand for Hopper is "really strong" ... Demand for Blackwell is "incredible" - Jensen

There's a race to the "next plateau" in AI - which is fueling the demand for Hopper even as Blackwell is in the bullpen.

Jensen: Companies that buy Nvidia get ROI right away. There's already $1T of CPU cloud/compute already - so the ROI isn't great on a new buildout. GPU ROI is way higher because there's far less capacity and it's the best infrastructure available.

Blackwell will have multiple configurations as it relates to cooling. Doesn't have to be liquid cooled - but most data centers will move to liquid cooling because it's (long term) cheaper and it allows you to pack more GPUs into a single data center.

Jensen: next year is going to be a great year, we'll grow our data center business quite significantly. Blackwell is significantly better than Hopper.

Wow what a great call - this really reminds me of the early days of the iPhone where every new model was WAY better than the previous.