Damnang
· Damnang’s Substack
· June 16, 2026 at 08:06
· ⏱ 11 min read
| Read on Substack ↗
Summary
The article argues that TSMC's transition to panel-level packaging (CoPoS) is a clear direction but faces deep uncertainty on where glass substrates fit and when mass production will actually ramp, making pure 'glass plays' risky. Instead, the author proposes evaluating stocks on two axes—CoPoS purity and path survival—and suggests inspection/metrology equipment (which is needed regardless of the winning packaging path) is the most resilient beneficiary, though no specific tickers are named or recommended.
•TSMC's CoPoS shifts from round wafers to square panels (first format 310x310mm, roadmap to 750x620mm) to improve area utilization for large AI chips.
•Pilot line at subsidiary VisEra starts 2026, small trial production targeted 2027, mass production cited 2028-2029 but may slip to 2029-2030 due to warpage and CTE management challenges.
•TSMC validated glass substrates with Ibiden and Innolux but stressed mass production is still far off, and glass's role (interposer vs core substrate) remains undecided.
•NVIDIA is cited as the first CoPoS customer, but the transition coexists with CoWoS and other paths like CoWoP, creating multiple technology forks.
•The article builds a framework with two axes: CoPoS purity (direct exposure to the transition) and path survival (revenue regardless of which packaging route wins), placing inspection equipment at the highest path survival.
•The author explicitly states the piece is for information and industry analysis, not investment recommendations, and may hold or trade mentioned stocks without disclosure.
TSMC is the central architect of CoPoS and the primary beneficiary if the transition succeeds, but timeline and technology risk remain high. The article notes TSMC's pilot line, validation partners, a
TSMC is the central architect of CoPoS and the primary beneficiary if the transition succeeds, but timeline and technology risk remain high. The article notes TSMC's pilot line, validation partners, and mass-production roadmap, making TSM the core name to monitor.
Risk: Mass production delays beyond 2029-2030 could dampen near-term revenue expectations; glass position uncertainty could require additional R&D spending.
NVIDIA is explicitly named as the first CoPoS customer, giving it early access to higher-area packages for larger AI accelerators. However, the article does not quantify NVIDIA's specific benefit or o
NVIDIA is explicitly named as the first CoPoS customer, giving it early access to higher-area packages for larger AI accelerators. However, the article does not quantify NVIDIA's specific benefit or order volume, and the timeline uncertainty means adoption may be gradual.
Risk: If CoPoS mass production slips materially, NVIDIA may rely longer on CoWoS, which is not a negative per se but delays the specific packaging advantage CoPoS offers.