The AI Supply Chain: Hidden Investment Opportunities

To get the lastest report, please subscribe to our newsletter or contact us at [email protected]

The AI revolution is constrained not by software but by physical infrastructure. Hyperscalers have committed $650B+ in 2026 capex for AI data centers, yet nearly half of planned U.S. projects are delayed or canceled due to supply chain bottlenecks . The shortages are structural, not cyclical — CoWoS packaging, HBM memory, power transformers, and critical minerals face multi-year constraints. This creates a seller’s market where incumbent suppliers have unprecedented pricing power.

1. The Supply Chain Pyramid

The investment framework visualized — capital flows from abundant top-layer services down to scarce bottom-layer resources. The highest-conviction opportunities sit at the foundation layers where supply constraints cascade throughout the entire chain.

Bottom layers = scarce resources with inelastic demand
Top layers = abundant capital competing for limited supply
Investment edge: own what AI giants cannot operate without
Structural shortages (not cyclical) = multi-year pricing power

2. High Bandwidth Memory (HBM): The #1 Bottleneck

While both are OTC equity derivatives, their risk profiles differ fundamentally. TRS provides linear exposure (like owning on margin), whereas the Accumulator (KODA—Knock-Out Discumulative Accumulator) embeds a knock-out barrier and downward leverage—earning it the nickname “I Kill You Later.” In 2008, Hong Kong investors lost HK$600 billion on Accumulators; CITIC Pacific lost HK$15 billion on AUD KODAs, leading to government bailout.

Key Data Points

HBM market: $38B (2025) → $58B (2026) → ~$95B (2028)
SK Hynix: “Already sold out our entire 2026 HBM supply”
Micron: “HBM capacity for 2025 and 2026 is fully booked”
Samsung: Raising HBM prices by high-teens to low-twenties % for 2026 contracts
SK Hynix overtook Samsung as the world’s largest DRAM manufacturer for the first time in history

Investment Angle

SK Hynix offers the purest HBM play. Micron offers the most dramatic earnings leverage (share growing from ~5% to 21%+). Samsung offers a contrarian recovery as qualification issues resolve in 2026.

3. CoWoS Advanced Packaging: The Hidden Constraint

Without CoWoS packaging, even perfectly fabricated 3nm wafers cannot become functional AI chips. This makes it arguably the most critical single process step.

Key Data Points

TSMC: “CoWoS capacity is very tight and remains sold out through 2025 and into 2026”
NVIDIA: “CoWoS assembly capacity is oversubscribed through at least mid-2026”
TSMC scaling CoWoS from ~35k wafers/month (late 2024) to 130k wafers/month by end of 2026 — nearly 4x
Google cut 2026 TPU production target by ~25% (from 4M to 3M units) due to CoWoS constraints

4. Power Infrastructure: The Gating Factor

The Scale of the Problem

U.S. data center electricity: 176 TWh (2023) → 325-580 TWh by 2028
Goldman Sachs: AI will drive a 165% increase in data center power demand by 2030
Transformer delivery: now takes up to 5 years (was 24-30 months pre-2020)
China controls ~60% of global transformer production capacity

The Nuclear Pivot

Big tech signed contracts for 10+ GW of new nuclear capacity in the past year alone :

5. Cooling: From Air to Liquid (Mandatory Transition)

As AI rack densities exceed 30kW and approach 100kW, liquid cooling becomes mandatory, not optional.

M&A is accelerating: Ecolab acquired CoolIT, Trane acquired LiquidStack, Schneider acquired Motivair, Daikin acquired Chilldyne.

6. Optical Interconnects: The Nervous System

80%+ of hyperscale data center links now use optical solutions . The market is growing at 15.3% CAGR, from $9.9B (2025) to $31B (2033).

Key Drivers

Co-Packaged Optics (CPO): 37% market share in AI data centers
Front-end metro DCI: 34% CAGR through 2029 — highest-growth subsegment
800G transition: Will capture ~70% of DCI market share by 2029
Broadcom targeting 200T optical interconnect solutions

7. Critical Minerals: The Foundation Layer

Copper: The Workhorse Metal

A single hyperscale facility uses ~2,200 tonnes of copper (~27 tonnes per MW)
Data center copper demand: 375,000 tonnes/year by 2030
New mines take 17 years from discovery to production
BHP: Copper demand will rise 72% by 2050

Key target: Freeport-McMoRan (FCX) — largest U.S. copper miner. Rio Tinto partnered with Indium Corporation to mine the first primary gallium in North America .

8. Supply Chain Bottleneck Severity Matrix

9. Investment Framework: 3 Tiers of Conviction

Tier 1 — Critical Bottlenecks (Highest Conviction)

Direct exposure to the most severe, persistent shortages.

Tier 2 — Infrastructure Enablers (Medium-High Conviction)

Leveraged exposure with more predictable demand.

Tier 3 — Emerging Opportunities (Higher Risk/Reward)

Transformative technologies at early-stage valuations.

Tier 4 — Infrastructure Enablers (Medium-High Conviction)

Leveraged exposure with more predictable demand.

10. Conclusion: The Structural Shortage Thesis

The defining characteristic of the AI supply chain is that shortages are architectural, not cyclical. Past semiconductor cycles resolved in 12-18 months. The AI cycle is different because demand grows exponentially while supply is constrained by physical limits: 3-5 years to build a fab, 2-3 years to qualify new HBM suppliers, 17 years to bring a copper mine online.

As TSMC CEO C.C. Wei stated: “The structural AI-related demand continues to be very strong” . Suppliers agree that if they had more capacity, they could sell 20-50% more — meaning revenue growth is constrained by supply, not demand.

The playbook: Own the bottlenecks. The companies controlling scarce resources in CoWoS, HBM, nuclear power, liquid cooling, and critical minerals will determine AI’s pace, shape its economics, and capture disproportionate value.

11. Reference

1.: SNS Insider. Optical Interconnect Market Report 2035. https://www.snsinsider.com/reports/optical-interconnect-market-7668

2.: DataM Intelligence. Optical Interconnect in AI Data Centers Market 2026-2033. https://www.marketresearch.com/DataM-Intelligence-4Market-Research-LLP-v4207/Optical-Interconnect-AI-Data-Centers-44408277/

3.: McKinsey. Opportunities in networking optics: Boosting supply for data centers. https://www.mckinsey.com/industries/technology-media-and-telecommunications/high-tech/our-insights/opportunities-in-networking-optics-boosting-supply-for-data-centers

4.: Enabling Photonic Technologies. Industry Perspectives: Optical Communications Market Update Q2 2025. https://www.enablingphotonictechnologies.org/industry-perspectives-optical-communications-market-update-q2-2025/

5.: IEA. Global Critical Minerals Outlook 2025. https://www.iea.org/reports/global-critical-minerals-outlook-2025

6.: Oplexa. AI Chip Packaging Bottleneck: TSMC Crisis 2026. https://oplexa.com/ai-chip-packaging-bottleneck-2026/

7.: ETF Trends. Data Centers Embracing Nuclear, SMRs for AI Needs. https://www.etftrends.com/disruptive-technology-channel/data-centers-embracing-nuclear-smrs-ai-needs/

8.: Forbes. Why Microsoft And Amazon Are Turning To Nuclear Power For AI. https://www.forbes.com/sites/rrapier/2026/02/19/why-microsoft-and-amazon-are-turning-to-nuclear-power-for-ai/

9.: Trellis. Amazon, Google, Meta and Microsoft go nuclear. https://trellis.net/article/amazon-google-meta-and-microsoft-go-nuclear/

10.: Carbon Credits. Data Centers’ Copper Hunger. https://carboncredits.com/data-centers-copper-hunger-how-ai-is-driving-a-looming-supply-crunch/

11.: Quest Metals. AI’s Hunger For Copper. https://www.questmetals.com/blog/ai-s-hunger-for-copper

12.: Rankiapro. Copper demand – the impact of Data Centers and AI. https://rankiapro.com/en/insights/copper-demand-impact-data-centers-ai/

13.: BHP. Why AI tools and data centres are driving copper demand. https://www.bhp.com/news/bhp-insights/2025/01/why-ai-tools-and-data-centres-are-driving-copper-demand

14.: IEA. Copper supply shortage threatens AI data centers. https://www.iea.org/reports/global-critical-minerals-outlook-2025

And more.

DISCLAIMER

Past performance does not guarantee future results.

Opinions and estimates offered constitute our judgment and are subject to change without notice, as are statements of financial market trends, which are based on current market conditions. We believe the information provided here is reliable, but do not warrant its accuracy or completeness. This material is not intended as an offer or solicitation for the purchase or sale of any cryptocurrencies. The views and strategies described may not be suitable for all investors. This material has been prepared for informational purposes only, and is not intended to provide, and should not be relied on for, accounting, legal or tax advice. Any forecasts contained herein are for illustrative purposes only and are not to be relied upon as advice or interpreted as a recommendation.

Unless otherwise stated, all data is as of October 7, 2024 or as of most recently available.