TSMC CoWoS Supply-Demand Gap Narrowing from 20% to 10% by End-2026

Summary

TrendForce report (2026-06-15) on TSMC's CoWoS advanced-packaging capacity trajectory. This is genuinely new against the existing CoWoS source (tsmc-q2-2026-cowos-sold-out-through-2026, phemex), which carried only the binary "sold out" claim. This source adds the capacity numbers and the demand-gap trajectory the frontier's "who holds the booked queue / how fast does packaging ease" open problem was missing, and directly addresses the flagged gap "CoWoS realized capacity vs the 130K target." TrendForce is a respected semiconductor research house → high-quality analysis-tier (still secondary, not a TSMC filing).

Key figures (as-of report date 2026-06-15)

• TSMC CoWoS monthly capacity, 2026: 120,000–140,000 wafers/month (the ~130K target sits inside this band — on track, possibly a hair short; a separate estimate cited ~127K by end-2026).
• OSAT partners (ASE et al.), 2026: +50,000–60,000 wafers/month of additional CoWoS-class capacity.
• Total industry advanced-packaging capacity target 2026: ~200,000 wafers/month (TSMC + OSAT).
• Supply-demand gap: ~20% now → ~10% by end-2026 — the shortfall is narrowing but not closed; demand still outruns supply through year-end.
• Growth trajectory: TSMC (May Taiwan Technology Symposium) guided CoWoS capacity to >80% CAGR from 2022 to 2027.
• Next-gen: NVIDIA's Feynman platform named as first CoPoS adopter, mass production 2028–2029 (Chiayi + Arizona) — beyond the current bottleneck window.

On-thesis relevance

Refines frontier #1 ("the constraint went plural"). The existing KB stated CoWoS "sold out through end-2026" (weak, one source). TrendForce quantifies the tightness and its trajectory: even with capacity quadrupling toward ~200K industry wafers/month, a ~10% supply-demand gap persists at end-2026 — so packaging stays a binding constraint through the year, not one that resolves. This is the "how fast does packaging ease?" answer: slowly, and not fully within 2026. It also gives the memory/packaging-hits-deployment thesis a supply-side denominator to pair with HBM (still volume-blind in the KB).

Provenance / caveats

• Secondary (TrendForce), not a TSMC disclosure — the KB's flagged gap for "a primary TSMC source" on CoWoS remains open; this is a higher-quality secondary than the prior phemex piece, not a filing.
• Figures are "reportedly" / supply-chain-checks sourced; the 120–140K band and the 20%→10% gap are estimates, dated 2026-06-15.
• "Realized vs target" is not cleanly resolved — the article gives a target/projection band, not audited realized output; the ~127K "by end-2026" figure is a projection.

Source: TSMC CoWoS Supply-Demand Gap Reportedly Seen Narrowing from 20% to 10% by End-2026, TrendForce, 2026-06-15.