NT$9.21B of financial asset gains inflated EPS to 8.45 while the CFO couldn't decompose a 6-point margin jump into price, volume, or utilization.
Thesis: ABF substrate capacity is the real bottleneck. CoWoS and EMIB are both capacity constrained through 2030, Unimicron is one of three board suppliers in that allocation, and management confirmed price and volume rising together with 80-85% of a 370B TWD capex program committed to ABF. Guangfu Fab 2 Phase 3 opens Q3 2026 with hyperscalers directly planning capacity with them. The headline EPS is inflated by NT$9.21B financial asset gains, but the underlying ABF business (22% QoQ revenue growth, 6-point margin expansion) is structurally advantaged. Own it for the capacity ramp, not the quarter's optics.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Guangfu Fab 2 Phase 3 capacity opening in Q3 2026 with 80-85% of new ABF capex already committed. Management guided Q3 higher on AI GPU, ASIC, and HPC demand with pricing intact.
Key Risk: Hyperscalers moving to COT and building their own supply chains, which Judy confirmed directly. Also EMIB first-phase yield target of only 50% and tool allocation dependent on three Japanese suppliers' ramp speed.
The Tell: When asked to decompose the 6+ point gross margin jump into price, volume, and utilization, the CFO said 'I don't have actual numbers' and could only confirm both moved together. On the exact metric analysts need to trust the margin trajectory, management had no breakdown. This is the Say/Do Gap on the core earnings quality question.
Friction Level: MODERATE_FRICTION — Both sides agree ABF demand is real and capacity-constrained. Disagreement centers on whether NT$8.93B non-operating income and segment reclassification make headline EPS meaningless, and whether COT displacement by hyperscalers is a 2030+ footnote or an active erosion.
Report not found
The report data is no longer available. Please return to the archive.