C↗ COMPUTEATLASA FIELD GUIDE TO COMPUTING
ASSEMBLING THE ATLAS
Loading six scales of computing
SCROLL TO EXPLORE · DRAG TO ROTATE
01 / 06INFRASTRUCTURE

The cloud
has an address.

Behind every answer, stream, and simulation is a very physical place. Let's step inside.

Three systems.
One purpose.

Power, cooling, and networking keep computation moving.

OBSERVATION SCALEBuilding
Meters
FROM INFRASTRUCTURE
TO THE ATOM
00%
COMPUTE ATLAS

A closer look
at computing.

A six-scale, interactive field guide to the physical systems behind AI. Scroll to travel inward, select a chapter to jump, or drag a model to inspect it.

All models were authored in Blender. The facility is conceptual. The rack follows GB200 NVL72 component counts; its arrangement and open inspection tray are illustrative. The Blackwell package illustrates a dual-die GPU and adjacent HBM stacks, not a production CAD model. The accelerator is nested inside an open rack tray; memory, device, and atomic details stay anchored inside their parent components. Camera magnification moves continuously between these details. Relative scale ratios are illustrative, not physically calibrated. The eight-layer HBM stack exposes exaggerated TSVs and interconnects. The one-transistor, one-capacitor DRAM cell is an educational cross-section, not a vendor process reconstruction. Die routing and memory subarrays are schematic. The silicon lattice shows atomic positions and bonds; atoms are not solid balls.

Technical references

Download the Blender source ↓