
Chapter 1
The four billion dollar handshakes
Chapter 2
The hardware heavyweights
Chapter 3
The brain of the operation: Muse
Chapter 4
The circle of silicon
Ch 2 · The hardware heavyweights

steve Named after the astronomer who confirmed dark matter, Nvidia's Vera CPUs handle the heavy math required for training AI.
steve The Hardware Heavyweights
Four billion dollars down to lock our silicon in your grid
steve Meta installed millions of AMD Epyc processors to run its newest high-speed artificial intelligence models.

steve AMD Helios rackscale systems optimize these massive data centers for complex AI models.
Quiz
Which family of AMD processors has Meta deployed by the millions for its AI workload?
Quiz
What is the name of the AMD rackscale solution currently running inside Meta's data centers?
steve Meta uses Nvidia Spectrum-X technology to keep data flowing smoothly between supercomputer chips
Quiz
What type of hardware is Nvidia's BlueField-4, which Meta is deploying in its network?
Quiz
What is the primary purpose of Nvidia's Spectrum-X Ethernet technology in an AI supercomputer?
Fun Fact
Power of a gigawatt: Just one of Meta's four planned superclusters will use enough electricity to power about 750,000 homes simultaneously.
steve Meta is building new superclusters that each require enough electricity to power 750,000 homes.
Quiz
How many gigawatt-scale 'superclusters' is Meta currently building to house its AI hardware?
Glossary
semiconductor (noun)
A solid substance that has a conductivity between that of an insulator and most metals. In October 2026, Nvidia and AMD pledged four billion dollars to fund Meta's massive artificial intelligence expansion.
A solid substance that has a conductivity between that of an insulator and most metals. In October 2026, Nvidia and AMD pledged four billion dollars to fund Meta's massive artificial intelligence expansion.

