Now Playing
Memory at 75%, Dell's $95B Backlog & the Convoy Slowing AI Build-Out
Audio is available on Spreaker — see link below.
AI infrastructure is bottlenecked not by GPUs but by optical transceivers, memory costs, and HBM supply — and the data is alarming. From Samsung doubling its HBM share to CXMT crossing 10% DRAM, today's briefing maps every chokepoint in the stack.
Audio is available on Spreaker — see link below.
Optical transceivers are now the slowest thing in AI infrastructure. Not GPUs.
The transceiver crisis doesn't sit in isolation. It's part of a broader pattern where memory and interconnect have quietly displaced compute as the primary constraint.
Inside the memory market, the competitive picture is shifting fast. Samsung doubled its HBM revenue share in a single quarter, jumping from twenty-one percent to thirty-three percent.
A third player is entering this market in a way that changes its structure. China's CXMT has crossed ten percent of global DRAM market share, up from below one percent in twenty twenty-three.
Dell's latest numbers put a commercial frame around all of this. The company is carrying a ninety-five billion dollar AI infrastructure backlog, a record.
The through-line across all of this is a convoy problem. AI infrastructure moves at the speed of its slowest critical component.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.