Let's talk about Nvidia earnings …
As we've discussed for much of the past two years, data center revenue has been telling a very clear story. There's a supply issue.
For seven quarters, data center revenue grew at a pace of about $4 billion a quarter – as if it were fixed.
And that put Nvidia’s growth rate on a steep downward slope, as seen here:
However, along the way, the backlog of DEMAND for Nvidia's most advanced chips has been insatiable. Every chip Nvidia receives from its manufacturer, Taiwan Semiconductor, is already sold.
And yet, back in May, they reported the slowest revenue growth since the onset of the AI boom.
Moreover, most of the data center revenue growth in Q1 was fueled by networking equipment — not the coveted GPUs.
Fast forward to Q2.
They reported today, and this time there was no growth contribution from "compute" (GPUs) — actually, it was slightly negative.
And this time, ALL of the data center revenue growth came from networking equipment (graphic below). It amounted to a second consecutive quarter of the slowest total data center growth since the onset of the AI boom.
As we've discussed over the past several quarters, Taiwan Semiconductor seems to have hit capacity/a hard ceiling — at least in what it's capable of producing for Nvidia.
And it seems clear that Nvidia can't chip away at the backlog of demand until new global capacity comes online (which will be in the U.S., next year).
What we don't know, is how Taiwan Semi determines how it allocates its capacity between the U.S. and China? Is there coercion from China on that decision?
