Semiconductors

HBM, CoWoS, advanced packaging, and the manufacturing logic of competitive leverage.

Semiconductor advantage is not created by demand alone. It is created where a process can be manufactured at yield, qualified by the right customer, expanded without eroding returns, and replaced only at real cost. HBM, CoWoS, ABF substrates, and co-packaged optics expose this logic clearly. A component may be technically available yet commercially non-substitutable because switching requires redesign, requalification, new process windows, or years of capacity build-out. That is why market share is often a lagging signal. I look instead for customer prepayments, platform-specific qualification, yield learning, capital intensity, and the time required to establish an equivalent supply path. This pillar follows competitive power from device architecture into packaging and materials—and then stress-tests it against algorithmic compression, demand reversal, geopolitical concentration, glass substrates, and process-integration risk. The question is not only where the bottleneck sits, but how long its owner can keep the leverage.

中文｜核心觀點

半導體的競爭優勢，不只來自需求成長，而在於能否達到量產良率、通過關鍵客戶驗證、以合理成本擴產，以及替代供應要多久才能建立。HBM、CoWoS、ABF 與 CPO 顯示，護城河藏在製程整合、認證鎖定與學習曲線。本支柱追蹤優勢如何形成，也檢驗演算法壓縮、需求反轉、地緣風險、玻璃基板與新製程如何削弱它。

Start Here — Reading Path

Read these in sequence: from technical foundations, to control, primary evidence, and moat expiration.

1. What Are CoWoS, HBM, and ABF — And Why Do They Matter So Much in the AI Era? — Start with a plain-language map from GPU and memory to packaging and substrate, and why chip design can decentralise faster than the manufacturing ecosystem can be replicated.

2. Article 2 | The Real AI Supply Chain: A Power Map Beyond the GPU — Move from what each layer does to who controls it, using qualification barriers, scarcity, and replacement time to separate pricing power from narrative exposure.

3. Article 3 | How Deep Is the Moat? Reading TSMC, SK Hynix, and Micron Through Their SEC Filings — Test the framework against original filings, where customer prepayments, margin structure, capital intensity, and risk disclosures reveal more than market-share headlines.

4. Article 4 | Stress-Testing the Moat: Four Threats That Could Rewrite the AI Supply Chain — Challenge today’s leaders with algorithmic compression, HBM demand reversal, geopolitical concentration, and glass substrates, because every moat has an expiration condition.

Papers

• Featured research — The Architecture of Leverage: Structural Concentration and Competitive Moats in the AI Compute Supply Chain — Introduces the author’s IRI—technology scarcity, supply concentration, qualification barriers, and replacement-cycle length—to compare leverage across CoWoS, HBM, and ABF, then stress-tests the conditions under which those positions could erode.

• Related research — Infrastructure-Led Leading Indicators in Technology Investment Cycles: Evidence from the Semiconductor Industry — Uses a 31-year semiconductor dataset to test the equipment-first hypothesis and introduces ECDR, adding a cycle-level check to company-level moat analysis.

All Posts in This Cluster

• Copper Is Running Out of Room. But Light Has a Manufacturing Problem.
• Article 4 | Stress-Testing the Moat: Four Threats That Could Rewrite the AI Supply Chain
• Article 3 | How Deep Is the Moat? Reading TSMC, SK Hynix, and Micron Through Their SEC Filings
• Article 2 | The Real AI Supply Chain: A Power Map Beyond the GPU
• What Are CoWoS, HBM, and ABF - And Why Do They Matter So Much in the AI Era?
• Infrastructure-Led Leading Indicators in Technology Investment Cycles: Evidence from the Semiconductor Industry
• 2025 AI晶片產業趨勢

Working on These Questions?

Facing a semiconductor or advanced-materials decision? I undertake defined-scope project engagements on manufacturing constraints, supplier leverage, and technology transitions.