Opportunities ahead, traced to the lab

🐂 Tools for 2D-material transistors.

• Very long timeline (judgment)
• Linked research is 6-10 years from market (our formula) (judgment)

Conviction score 26 = 0.4×linked readiness 16 + 0.4×20×exposure 1/5 + 0.2×20×execution 3/5 → Watch. Our judgment.

🐂 Analog in-memory & HBM; Gujarat plant.

• Memory cycle (judgment)
• System-level review: specialised analog/physical accelerators can show 'little or no advantage' once conversion, control and memory access are counted. source

Conviction score 51 = 0.4×linked readiness 27 + 0.4×20×exposure 3/5 + 0.2×20×execution 4/5 → Medium. Our judgment.

🐂 Glass substrate validation with TSMC (reported).

• Glass may arrive late (judgment)
• Glass timelines keep slipping (Tom's Hardware). source

Conviction score 61 = 0.4×linked readiness 52 + 0.4×20×exposure 3/5 + 0.2×20×execution 4/5 → Medium (was Speculative). Our judgment.

🐂 ₹91,526 cr Gujarat fab with PSMC.

• Not directly investable (judgment)
• Glass timelines keep slipping (Tom's Hardware). source

Conviction score 45 = 0.4×linked readiness 52 + 0.4×20×exposure 2/5 + 0.2×20×execution 2/5 → Watch. Not directly investable → Watch. Our judgment.

🐂 Lasers/optics for photonic links & CPO.

• Optics price wars (judgment)
• Nvidia networking SVP: 'If you can use copper, use copper' — optics only where distance forces it. source

Conviction score 64 = 0.4×linked readiness 51 + 0.4×20×exposure 4/5 + 0.2×20×execution 3/5 → Medium. Our judgment.

🐂 Photonics foundry for startups.

• Customer concentration (judgment)
• Nature Photonics review (2026): in the cloud, losses and light-to-electric conversions 'dominate power' and have so far prevented commercial photonic tensor cores. source
• Nvidia networking SVP: 'If you can use copper, use copper' — optics only where distance forces it. source

Conviction score 49 = 0.4×linked readiness 34 + 0.4×20×exposure 3/5 + 0.2×20×execution 3/5 → Medium (was Speculative). Our judgment.

🐂 Glass core substrates.

• Glass adoption slips (judgment)
• Glass timelines keep slipping (Tom's Hardware). source

Conviction score 53 = 0.4×linked readiness 52 + 0.4×20×exposure 2/5 + 0.2×20×execution 4/5 → Medium (was Speculative). Our judgment.

🐂 Edge neuromorphic chips.

• Tiny revenue; dilution (judgment)
• BrainChip, a listed neuromorphic pioneer, reported only US$1.22M revenue and a US$12.0M loss for H1 2026 — commercial demand for brain-like chips is still tiny. source

Conviction score 55 = 0.4×linked readiness 28 + 0.4×20×exposure 5/5 + 0.2×20×execution 1/5 → Speculative (was Watch). Our judgment.

🐂 Hybrid bonding tied to Applied Materials partnership.

• Adoption timing slips (judgment)
• SK hynix: hybrid bonding not ready for HBM4E. source

Conviction score 88 = 0.4×linked readiness 80 + 0.4×20×exposure 5/5 + 0.2×20×execution 4/5 → High (was Medium). Our judgment.

🐂 AI chip testing demand.

• Cyclical; valuation (judgment)
• SK hynix: hybrid bonding not ready for HBM4E. source

Conviction score 64 = 0.4×linked readiness 80 + 0.4×20×exposure 2/5 + 0.2×20×execution 4/5 → Medium. Our judgment.

🐂 OSAT plant in commercial production (ISM).

• Small scale; high expectations priced in (judgment)
• System-level review: specialised analog/physical accelerators can show 'little or no advantage' once conversion, control and memory access are counted. source

Conviction score 34 = 0.4×linked readiness 35 + 0.4×20×exposure 1/5 + 0.2×20×execution 3/5 → Watch (was Speculative). Our judgment.

🐂 Chip packaging JV with Renesas & Stars.

• Execution; ramp delays (judgment)
• SK hynix: hybrid bonding not ready for HBM4E. source

Conviction score 52 = 0.4×linked readiness 80 + 0.4×20×exposure 1/5 + 0.2×20×execution 3/5 → Watch (was Speculative). Our judgment.

🐂 Licenses hybrid-bonding IP.

• Litigation outcomes (judgment)
• SK hynix: hybrid bonding not ready for HBM4E. source

Conviction score 64 = 0.4×linked readiness 80 + 0.4×20×exposure 3/5 + 0.2×20×execution 2/5 → Medium (was Speculative). Our judgment.

🐂 PIC100 photonics in volume.

• Auto/industrial weakness (judgment)
• Nvidia networking SVP: 'If you can use copper, use copper' — optics only where distance forces it. source

Conviction score 66 = 0.4×linked readiness 84 + 0.4×20×exposure 2/5 + 0.2×20×execution 4/5 → High (was Medium). Our judgment.

🐂 HBM4 full-scale shipments started Q3 2026.

• Memory price cycle reversal (judgment)
• SK hynix: hybrid bonding not ready for HBM4E. source

Conviction score 80 = 0.4×linked readiness 80 + 0.4×20×exposure 4/5 + 0.2×20×execution 4/5 → High. Our judgment.

🐂 2nm steep ramp; Q4 guide $52.3-54.3B.

• Taiwan risk; overseas fab margin drag (judgment)
• Glass timelines keep slipping (Tom's Hardware). source

Conviction score 57 = 0.4×linked readiness 52 + 0.4×20×exposure 2/5 + 0.2×20×execution 5/5 → Medium (was High). Our judgment.

🐂 Vera Rubin ramp in full production.

• Customer capex slowdown; China rules (judgment)
• Nvidia networking SVP: 'If you can use copper, use copper' — optics only where distance forces it. source

Conviction score 60 = 0.4×linked readiness 60 + 0.4×20×exposure 2/5 + 0.2×20×execution 5/5 → Medium. Our judgment.

🐂 Q3 results Nov 3 2026; Q3 revenue guide ~$13.0B.

• Execution vs Nvidia (judgment)
• BrainChip, a listed neuromorphic pioneer, reported only US$1.22M revenue and a US$12.0M loss for H1 2026 — commercial demand for brain-like chips is still tiny. source

Conviction score 39 = 0.4×linked readiness 28 + 0.4×20×exposure 1/5 + 0.2×20×execution 5/5 → Watch (was Medium). Our judgment.