Custom silicon displacement fight
story cl-0007 · born 2026-05-25 · last seen 2026-06-22 · lifecycle fading
Core thesis
A core contradiction inside AI semis is that custom ASIC and TPU demand can validate infrastructure growth while pressuring Nvidia's share and margins. Bulls in AVGO, AMZN, GOOG and AMD see verticalized silicon optionality, while NVDA bulls argue full-stack lock-in and inference demand still dominate.
Cracks — what would invalidate
The displacement thesis fails if hyperscaler custom chips remain internal, underperform merchant GPUs, or require Nvidia networking/software more than expected. The Nvidia bull case fails if external TPU and ASIC financing becomes a durable margin-transfer mechanism.
Key evidence
Tickers in this story
| ticker | last close | mcap | since last seen (2026-06-22) |
|---|---|---|---|
| AMD | $517.82 | $844.4B | -6.1% |
| AMZN | $242.67 | $2.6T | +4.2% |
| ARM | $315.28 | $335.5B | -22.7% |
| AVGO | $360.45 | $1.7T | -8.1% |
| GOOG | $356.18 | $4.3T | +2.1% |
| GOOGL | $359.91 | $4.4T | +2.9% |
| INTC | $120.35 | $604.9B | -14.6% |
| NVDA | $194.83 | $4.7T | -6.6% |
| ORCL | $140.27 | $404.0B | -19.9% |
| QCOM | $176.25 | $185.8B | -20.6% |
Also in this story, no US price data on file (index / non-US listing): SSNLF.
Who's driving it (author voices)
DriversTrajectory (chronological)
Stories refresh with the weekly run: fresh discovery, SQL Jaccard continuity on ticker sets, lifecycle from measured flow — never model vibes. Dated catalysts get adjudicated (happened / missed) on the next pass.