Seoul links chip tax breaks to domestic HBM packaging, tightening control of AI memory supply chain

South Korea’s Ministry of Trade, Industry and Energy has issued guidance that makes enhanced K‑Chips tax credits and grants contingent on building or sourcing advanced high bandwidth memory packaging domestically, according to notices circulated in Seoul this week and first reported by ETnews and Maeil Business Newspaper. The move targets through silicon via stacking, 2.5D interposers, and related backend steps for HBM, and it comes as SK Hynix and Samsung Electronics race to add capacity for Nvidia and other accelerator customers.

The change matters because packaging has become the binding constraint in AI hardware, not only wafer output. By linking support to local content in HBM assembly and test, Seoul is steering investment toward Korean OSATs and equipment suppliers and away from overseas providers that dominate 2.5D packaging.

MOTIE’s notice clarifies that HBM die stacking, interposer bonding, and high density redistribution layer processes qualify as national strategic technologies under the existing K‑Chips framework, which offers facility tax credits for large companies and additional deductions tied to supply chain resilience. The guidance also prioritizes cleanroom expansions and process tool purchases for packaging lines located in Korea when allocating cash grants.

Local suppliers stand to benefit. Hanmi Semiconductor, which sells thermal compression bonders used in HBM stacking, and Wonik IPS, which provides deposition and etch tools, have been cited by Korean brokers as near term winners as eligibility shifts toward domestic procurement. OSATs including Amkor Technology Korea, LB Semicon, and Hana Micron are expected to compete for new contracts as memory makers diversify beyond foundry led packaging.

SK Hynix has added HBM capacity at Icheon and Cheongju while qualifying next generation parts with Nvidia. Samsung has guided to higher HBM output from Pyeongtaek as it chases share in HBM3E and future HBM4. Both firms historically leaned on Taiwan for some advanced packaging, including TSMC CoWoS for 2.5D, a dependency Seoul is now trying to reduce.

The policy also aligns with export control realities. With the United States, Japan, and the Netherlands limiting advanced tool shipments to China, keeping HBM packaging onshore gives Korea more control over sensitive steps that straddle memory and logic. For global buyers of AI servers, the guidance signals that the next capacity swing factor is not only wafers but where, and with whom, the most intricate packaging happens.