The Altman Buildout

Sam Altman

What’s inside

NVDA NVIDIA: The accelerator supply Altman's model roadmap is priced against; OpenAI is the anchor customer.
AVGO Broadcom: Custom accelerators and switching are how he escapes single-vendor pricing on compute.
TSM TSMC: Every chip in his buildout is fabbed here; his timelines assume leading-edge capacity.
ORCL Oracle: His chosen partner for the vast capacity commitments behind OpenAI's training clusters.
CRWV CoreWeave: Pure-play rented GPU capacity of the kind OpenAI has contracted for at huge scale.
VRT Vertiv: Power and cooling kit is the physical bottleneck on the gigawatt campuses he describes.
ETN Eaton: Electrical distribution inside the buildout he says must keep compounding for years.
CEG Constellation Energy: Firm nuclear output is the nearest-term version of the clean baseload his data centers need.
VST Vistra: Merchant generation that gets repriced when AI demand meets a fixed grid, as he warns.
GEV GE Vernova: Turbines, grid gear and nuclear services are the picks-and-shovels of his energy thesis.
CRWD CrowdStrike: His own warnings about advanced cyber capability argue for far more endpoint defence spend.
PANW Palo Alto Networks: If rogue agents probe institutions, network and cloud defence is the direct offset.
MSFT Microsoft: His largest partner and the channel through which his models reach regulated enterprises.
RDDT Reddit: An early Altman bet worth over a billion, and the human-data layer AI models train on.
ABNB Airbnb: A signature Altman early investment and his template for founder-led compounding.

Why it could work

If Altman is right that capability keeps compounding, the chips, power and buildings he is contracting for stay supply-constrained for years.

What could break it

Safety incidents, training pauses or regulation could stall the spending cycle and leave the buildout overbuilt.

For learning only. Not investment advice.