AI beyond NVIDIA

The AI supply chain has more than one winner.

By Alex Chen

What’s inside

• NVDA NVIDIA: Huang claims $1T of Blackwell/Vera Rubin orders through 2027; still the chain's anchor winner
• AVGO Broadcom: Q3 AI semi revenue $16.7B, up 221%; proof custom accelerators win alongside GPUs
• AMD AMD: The second merchant accelerator source if buyers refuse a single supplier
• TSM TSMC: Sales up 45% and another $100B US pledge; paid no matter which chip designer wins
• ASML ASML: Sole route to leading-edge lithography; the chain cannot expand around it
• MU Micron: Customers getting half to two-thirds of memory requested; pricing power is the payout
• AMAT Applied Materials: New memory and logic fabs mean deposition and etch orders regardless of chip winner
• KLAC KLA: Yield control gets harder at advanced nodes and HBM stacks; inspection demand rises
• ETN Eaton: Power distribution for the gigawatts Broadcom and Nvidia customers keep announcing
• VRT Vertiv: Rack-scale systems need dense cooling and power; a non-chip way to hold the build-out
• CEG Constellation Energy: Supplies the electricity the compute capacity runs on, independent of chip vendor

Why it could work

The AI supply chain has more than one winner. The basket combines direct participants and broader suppliers.

What could break it

Adoption could disappoint, competition could weaken margins, or high valuations could already price in the trend. Some holdings are broad, indirect proxies.

For learning only. Not investment advice.