The sector reports assume you know what a process node is, why HBM sits next to the GPU, and how a datacenter is powered. If you don't — start here. Read modules 01–08 for the technology stack (an afternoon), 09–10 for the industry structure and the metrics that matter. Every number is sourced; every figure states whether it is an original illustration or an institutional image.
Modules
Module 00
How to Read This Series
The ten sectors as one machine: what it looks like, which primer module unlocks which door, and the shortest reading path for your job.
Module 01
From Sand to Silicon
How purified sand becomes the most precisely manufactured object humans make — and why yield math and $20B fabs explain half of the AI supply chain.
Module 02
Transistors and the End of Cheap Scaling
After this module you'll know what a transistor actually does, why fifty years of shrinking made computing exponentially cheaper, and why what ended was not the shrinking — but the cheapness.
Module 03
Anatomy of an AI Accelerator
Read a GPU spec sheet like an architect: what SIMT, Tensor Cores, HBM bandwidth, FP4 and 'NVL72' actually mean — and which of them is the real bottleneck.
Module 04
Training vs Inference
After this module you can look at any AI chip announcement and know which of the two workloads it is really built for — and read a $/Mtok chart the way the buyers do.
Module 05
The Memory Hierarchy and HBM
Understand the register-to-SSD speed pyramid, why DRAM stopped shrinking, and how stacked HBM became 50-60%+ of an AI GPU's manufacturing cost.
Module 06
Advanced Packaging 101
Why the last step of chipmaking became the chokepoint of AI compute — and what chiplets, CoWoS, HBM integration and hybrid bonding actually are.
Module 07
Interconnect: Scale-Up vs Scale-Out
After this module, any interconnect headline — '900 GB/s NVLink', 'CPO kills pluggables' — instantly tells you which of the datacenter's two networks is in play, and whose revenue moves.
Module 08
The Datacenter as a Machine
After this module, 'a 100MW campus', 'PUE 1.1', '800VDC' and 'behind-the-meter gas' stop being buzzwords — you can trace every watt from the grid substation to the sub-1V pin of a GPU, and explain why electrons, not chips, became AI's scarcest input.
Module 09
The Industry Map: Who Does What and Where the Money Pools
After this module you can place any company in the AI supply chain on one map — and predict its gross margin, its capex burden, and its pricing power before you ever open its 10-K.
Module 10
The Investor's Toolkit and Glossary
Every number that moves an AI supply-chain stock — what it actually measures, where it gets published, which traps eat analysts, and the 43 terms that unlock all ten sector briefs.