High Bandwidth Memory

3 videos across 3 channels

A look at how memory bandwidth shapes AI compute, from the global bottlenecks driving up costs to the way memory, packaging, and substrates constrain throughput. The discussion ties High Bandwidth Memory to supply-chain realities, power and cooling limits, and the governance needed across vendors and hyperscalers to avoid bottlenecks as chip production becomes a vast, multi-country factory. It highlights why memory speed and reliability are as critical as raw compute in sustaining AI progress.

Elon won after all thumbnail

Elon won after all

The video explains a global compute crunch driving up costs and bottlenecks across AI makers, hardware suppliers, and po

00:23:21
I Burned 500 Million Tokens Last Week. Do You Know Yours? thumbnail

I Burned 500 Million Tokens Last Week. Do You Know Yours?

The speaker argues that AI capacity isn’t limited by GPUs alone but by an entire industrial supply chain—from chips, mem

00:23:37