Now Playing
Broadcom's $73B Backlog, HBM Shortage & the AI Infrastructure Spread
Audio is available on Spreaker — see link below.
AI hardware investment is diversifying fast — Broadcom's custom chip backlog hit $73B, memory stocks are lapping Mag Seven, and Nvidia's China revenue is effectively zero. Today's briefing maps who's capturing margin at every bottleneck in the AI infrastructure stack.
Audio is available on Spreaker — see link below.
Broadcom's custom chip backlog just hit seventy-three billion dollars. That single number tells you more about where AI infrastructure money is actually landing than almost anything else right now.
Custom accelerators from Broadcom now account for nearly twenty-eight percent of AI server shipments. Volume grew forty-four point six percent year over year.
The memory story is equally sharp. Micron is up two hundred and twenty percent year to date.
Nvidia understands this. The Vera Rubin architecture, now entering volume shipments, generates forty billion dollars of revenue opportunity per gigawatt of compute.
One development that deserves more attention than it's getting is the ACIE customer segment, AI clouds, industrial, enterprise. This cohort grew twenty-five percent quarter over quarter versus thirteen percent for hyperscalers.
On China, the situation is straightforward. Q3 guidance assumes essentially zero GPU revenue from China.
That pressure is being absorbed by a global data center pipeline that shifted sharply in August. Australia, two point two gigawatts, thirty-one billion Australian dollars.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.