SK hynix May Delay HBM4 Equipment Orders to 2027 as DRAM Profits Surge
SK hynix is reportedly considering postponing some of its HBM4 backend manufacturing equipment orders until 2027, raising questions about the company's production strategy as conventional DRAM prices continue climbing. According to a report from South Korean publication Dealsite, equipment originally scheduled for the second half of 2026 could be deferred as the manufacturer evaluates shifting additional investment toward more profitable general purpose DRAM production.
The potential adjustment comes as DDR5 and server memory prices continue rising amid strong demand from AI infrastructure, enterprise computing, and consumer hardware manufacturers. With conventional DRAM generating increasingly attractive returns, SK hynix may have greater financial incentives to prioritize existing memory production rather than immediately expanding its more complex HBM4 packaging operations. This possibility aligns with industry projections showing conventional DRAM could deliver higher margins than HBM, particularly as advanced memory stacking increases manufacturing costs.
The report also suggests that SK hynix could initially allocate more capacity at its upcoming Yongin semiconductor complex toward conventional DRAM. However, the company has not confirmed a change in its manufacturing strategy, and a representative reportedly stated that backend equipment investments are proceeding according to the original schedule.
SK hynix HBM4 Production Outlook
| Development | Current Status |
|---|---|
| HBM4 Backend Equipment | Some orders reportedly could move to 2027 |
| Conventional DRAM Investment | Possible higher priority due to rising profitability |
| Yongin Production Capacity | Initial allocation may favor conventional DRAM |
| HBM4 Packaging Challenges | Industry speculation, not officially confirmed |
| Official Investment Schedule | SK hynix reportedly maintains existing plans |
| HBM4 Production Disruption | No confirmed impact |
Beyond investment priorities, questions have emerged regarding SK hynix's HBM4 packaging technology. Semiconductor commentator @zephyr_z9 highlighted concerns involving Mass Reflow Molded Underfill (MR MUF), the technology used to bond and protect vertically stacked DRAM dies. As HBM4 introduces tighter physical requirements and higher stacking density, maintaining reliable underfill distribution becomes increasingly challenging.
“Some speculate that SK hynix may have encountered technical issues in its HBM4 back-end processing. However, no direct link between the equipment investment schedule and product quality issues has been confirmed so far.” SK hynix is having issues with HBM4.
— Jukan (@jukan05) October 14, 2026
However, these concerns do not establish that SK hynix is experiencing unresolved production failures. Daishin Securities analyst Ryu Hyung Geun reportedly considers the quality concerns exaggerated, noting that advanced packaging challenges are not necessarily insurmountable. SK hynix has already demonstrated its 16 layer, 48 GB HBM4 technology, including a design incorporating an advanced logic base die developed in partnership with TSMC, as detailed in its official technology presentation.
The development comes amid intensifying competition with Samsung and Micron for next generation AI memory contracts. Any significant reduction in SK hynix's HBM4 expansion could create opportunities for competitors, although equipment scheduling changes alone do not establish a delay in product shipments or a loss of market leadership.
The possibility of SK hynix prioritizing conventional DRAM over additional HBM4 capacity highlights an important shift in memory market economics. HBM remains essential for advanced AI accelerators, but increasing DDR5 and server DRAM prices are making conventional memory production exceptionally profitable.
The distinction between equipment delays and manufacturing problems is critical. HBM4 requires increasingly sophisticated packaging, but speculation about technical difficulties should not be treated as evidence of production failures.
For gamers and PC builders, the implications are equally interesting. Greater investment in conventional DRAM could eventually support DDR5 availability, but it does not guarantee lower prices while enterprise customers continue securing substantial production capacity through long term agreements.
Ultimately, SK hynix must balance immediate DRAM profitability against maintaining its technological position in the rapidly expanding AI memory market.
Should SK hynix prioritize expanding conventional DDR5 and server DRAM production while prices remain high, or continue investing aggressively in HBM4 to defend its position in the AI market?
