Amkor triples Gwangju investment as AI packaging demand grows

Amkor tripling its Gwangju commitment is not a capacity story—it is a geopolitical re-architecture of the OSAT landscape. The AI chip bottleneck is migrating from "who fabricates the die" to "who stacks the dies," and advanced packaging has replaced wafer manufacturing as the supply-chain chokepoint. Technically, 2.5D interposer and chiplet integration are the battleground, directly mirroring TSMC's CoWoS scaling cadence. Upstream, ABF substrate makers (Ibiden, Unimicron) and underfill suppliers (Ajinomoto) will see order books stretch beyond 18 months, shifting pricing power toward materials. The compliance angle is sharper than it appears. Seoul's southwest strategy is explicitly a friendly-shoring play: anchoring a US-listed firm in Gwangju sidesteps packaging capacity in China and reduces single-point dependency on TSMC's internal packaging. The hidden cost? Amkor's capex is now partially tethered to Korean industrial policy, with subsidy clawback clauses as a latent liability. Competitively, ASE Group will likely accelerate Southeast Asian expansion to defend share, while TSMC continues internalizing CoWoS, compressing the third-party OSAT TAM. Amkor's thesis: the chiplet ecosystem fragments enough to keep outsourced packaging relevant. 12–24 month outlook: once the Gwangju line is live, capacity will almost certainly be pre-allocated to two or three hyperscaler customers. The real inflection point is not the capacity number—it is whether Amkor's 2.5D interposer yield can match TSMC's in-house process. If it cannot, triple the investment is just expensive inventory.