Research record
Not reviewedNo reviewer, review date, or supporting citations are recorded for this demo thesis. Scores are editorial opinions and have not been verified against cited evidence.
Revision history
No dated revisions are recorded for this thesis.
Investment thesis
The thesis in one line
You can argue about which chip designer wins — but they all manufacture at TSMC. Own the foundry and you own the bottleneck.
The ultimate picks-and-shovels
NVIDIA, AMD, Apple, Broadcom and the hyperscalers' custom silicon share one thing: they are fabless. The actual leading-edge manufacturing happens at TSMC, which holds an effective monopoly on the most advanced (3nm, 2nm) process nodes. Whoever wins the design wars, TSMC collects the toll. There is no more diversified way to own "AI compute" than owning the factory that prints all of it.
A second scarce bottleneck
Beyond wafers, AI accelerators need advanced packaging (CoWoS) to bond logic and high-bandwidth memory together. Packaging capacity has been a genuine constraint on AI-chip supply — and TSMC dominates it. That gives the company two points of leverage over the entire industry's output.
The discount and the reason
TSMC consistently trades at a lower multiple than the fabless designers it enables. The reason is one word: Taiwan. The geopolitical overhang is real and won't disappear, which is why diversification into Arizona and Japan fabs matters. For investors who can stomach that tail risk, you are buying the most indispensable company in technology at a discount to its customers.
Stage of the cycle
Indispensable and compounding — capex-heavy, but every dollar extends a near-monopoly.
Scorecard
Upside 7.5/10 · Safety 7.5/10 · AI exposure 9.5/10. The single most irreplaceable link in the AI chain.
