Samsung vs SK Hynix

HBM Market Dominance and the Profit Flip

The most dramatic development in the Samsung–SK Hynix rivalry came in January 2026 when SK Hynix reported 47.2 trillion won in operating profit for 2025, surpassing Samsung's 43.6 trillion won — the first time SK Hynix has ever beaten Samsung on this metric. This profit flip was driven almost entirely by SK Hynix's dominance in HBM, where it supplies the majority of chips used in NVIDIA's AI GPUs.

SK Hynix held between 57% and 62% of the HBM market by revenue through 2025, while Samsung trailed at roughly 22%. Notably, Micron actually overtook Samsung for the number-two spot earlier in 2025, adding urgency to Samsung's turnaround efforts. However, Samsung has publicly stated its goal to push HBM market share above 30% in 2026, and early signs suggest this target is achievable given its new customer wins.

The HBM4 Race: Neck and Neck on Timelines

Both companies began mass production of HBM4 in early 2026, targeting the NVIDIA Vera Rubin GPU platform. SK Hynix demonstrated the industry's first 16-layer 48GB HBM4 at CES 2026, building on its 12-layer 36GB product that achieved record 11.7 Gbps speeds. Samsung countered at NVIDIA GTC 2026 by unveiling HBM4E — a next-generation product delivering 16 Gbps per pin and 4.0 TB/s bandwidth.

Samsung's HBM4 achieves consistent 11.7 Gbps processing speeds with an enhanced mode reaching 13 Gbps. Critically, Samsung is leveraging its foundry division to fabricate HBM4 base dies in-house, allocating over 50% of its Pyeongtaek foundry capacity to this effort. SK Hynix, lacking its own foundry, relies on external partners for logic dies — a potential vulnerability as HBM4 integrates more complex logic layers.

The competitive dynamics for 16-layer HBM4, which NVIDIA has requested for the second half of 2026, remain fluid. SK Hynix leads with its refined MR-MUF packaging process, while Samsung and Micron are fighting aggressively for supply contracts.

Strategic Partnerships and Customer Diversification

SK Hynix's relationship with NVIDIA remains its greatest asset — it has been the primary HBM supplier since the H100 era and is positioned as the first-mover supplier for the Rubin platform. This deep partnership has given SK Hynix unmatched volume visibility and co-engineering advantages.

Samsung, however, has made significant strides in customer diversification. In March 2026, Samsung secured an exclusive deal to supply HBM4 to OpenAI for its in-house Titan AI chip — a major validation. Samsung also signed a memorandum of understanding with AMD to serve as a primary HBM4 supplier for the Instinct MI455X GPU. These partnerships reduce Samsung's dependence on NVIDIA qualification and open new high-volume channels.

Manufacturing Technology and Packaging

The two companies take fundamentally different approaches to HBM packaging — a critical differentiator as layer counts increase. SK Hynix uses its proprietary MR-MUF (Mass Reflow Molded Underfill) process, which it has refined over multiple HBM generations. For HBM4, SK Hynix introduced an advanced EMC material that improves heat dissipation by approximately 1.6× over previous versions, a crucial advantage as 16-layer stacks generate significant thermal challenges.

Samsung employs its own advanced packaging technology and benefits from vertical integration with Samsung Foundry. By fabricating HBM4 logic base dies internally on its advanced process nodes, Samsung can co-optimize the memory and logic layers in ways that SK Hynix cannot easily replicate. Samsung has reported logic die yields exceeding 90%, though its 1c DRAM yields remain around 50% — suggesting room for improvement in cost efficiency.

Vertical Integration vs. Pure-Play Focus

Samsung's breadth — spanning memory, foundry, mobile, and consumer electronics — is both a strength and a liability. The foundry division gives Samsung a unique ability to manufacture HBM4 base dies in-house, but it also means the company must allocate capital and engineering talent across competing priorities. Samsung Foundry has struggled to match TSMC in advanced logic manufacturing, and critics argue that spreading resources too thin has contributed to Samsung's HBM market share losses.

SK Hynix, as a pure-play memory specialist, can concentrate all R&D and capital expenditure on memory and HBM. This focus has enabled SK Hynix to consistently be first to market with new HBM generations and maintain tighter relationships with key customers. However, SK Hynix's lack of foundry capability means it depends on third parties for HBM4 logic dies — a dependency that could become a bottleneck as base die complexity increases with HBM4's shift toward integrating more processing logic.