SK hynix is having difficulty with hybrid bonding for high-bandwidth memory (HBM), according to a report relayed by TechPowerUp, and now trails rival Samsung in this specific packaging technology.

The claim concerns one narrow but critical link in the HBM production chain. Hybrid bonding is an advanced packaging technique that connects stacked DRAM dies directly, without solder bumps, enabling higher bandwidth density and better thermals as memory stacks grow taller.

Neither company has commented publicly on the reported gap. SK hynix did not dispute the account, and Samsung has made no official statement claiming leadership in this process step.

What is hybrid bonding and why does it matter?

HBM chips stack multiple DRAM dies on top of a base logic die and place them next to a GPU or accelerator. As the industry moves to taller stacks with more layers, conventional microbump interconnects approach their physical limits.

Hybrid bonding replaces those bumps with direct copper-to-copper connections between dies. The result is a shorter electrical path, lower power per bit transferred, and the headroom needed to keep scaling stack heights in future HBM generations.

Any vendor that cannot master the process risks falling behind in the segments that demand the densest memory configurations — chiefly AI accelerators from the largest chip designers.

Where does this leave SK hynix?

SK hynix currently supplies the majority of HBM used in high-end AI hardware. A stumble in one packaging technique does not erase that position, and the report itself frames the issue as a difficulty rather than a failure.

However, the competitive read is straightforward. If Samsung has advanced further on hybrid bonding, it holds a card it has not held in the HBM market for some time: a genuine process advantage in the technology that defines the next generation of stacked memory.

Samsung has spent the current HBM cycle fighting to win qualification from major customers. A hybrid-bonding lead would strengthen its pitch for future product cycles, where the technique is expected to become a requirement rather than a differentiator.

What happens next?

Watch three signals:

• Supplier disclosures. Both companies discuss packaging roadmaps at earnings calls and industry events; any change in language on bonding methods would confirm or soften the report.

Customer qualification decisions. Accelerator vendors effectively arbitrate this rivalry; their sourcing choices reveal which supplier's technology works.

Next-generation HBM announcements. Roadmaps that rely on hybrid bonding will expose whichever vendor has not solved the process problem.

For now, the story rests on a single industry report. It describes a technical difficulty at the current HBM leader, a claimed advantage for its closest competitor, and a packaging technology that will decide part of the next AI memory cycle.