SK hynix Dethrones Samsung in DRAM: Shipments, ASPs, Revenue, and the Coming Hybrid Bonding Transition

For more than three decades, Samsung Electronics has stood unchallenged at the top of the DRAM market. No more.

According to Korean media reports, SK hynix has overtaken Samsung for the first time in 33 years, claiming 36.3% of global DRAM market share versus Samsung’s 32.7%. Micron, while still in third place, has benefited from consistent share gains driven by stronger alignment with AI customers.

Shipments: Samsung’s Volume Lead Erodes

Chart 1 shows DRAM shipments measured in millions of 1 Gb equivalents. Although Samsung continues to ship the largest absolute volumes, its trajectory has flattened. SK hynix and Micron have accelerated sharply since 2023, narrowing the gap. The linear trendlines show that Micron has the steepest long-term growth slope.

Chart 1

ASPs: SK Hynix Commands a Premium

Chart 2 shows average selling prices revealing how technology positioning dictates profitability. SK hynix commands the highest ASPs, reflecting its outsized role in HBM3 and HBM3E shipments tied to NVIDIA GPUs. Samsung’s ASPs lag as it remains more exposed to commodity DRAM markets like smartphones and PCs, where oversupply keeps pricing under pressure. Micron sits in between: benefiting from its growing HBM share but still dragged by mainstream exposure.

Chart 2

Revenues: The Market Share Crown Changes Hands

Chart 3, showing DRAM revenues, reveals the culmination of volume and pricing dynamics. For years, Samsung’s revenue line sat comfortably above its rivals. That changed in 2024–2025, as SK hynix’s revenue trajectory surged past Samsung on the strength of higher ASPs and expanding HBM shipments. Micron has also grown revenue alongside its HBM3E ramp, further tightening the competitive field. This revenue crossover marks the first time in three decades that Samsung has lost the DRAM leadership position.

Chart 3

Demand by Client: Nvidia Drives Half of the Market

Chart 4 shows that HBM demand in 2024 was overwhelmingly driven by Nvidia, which accounted for more than half (52%) of total consumption. Broadcom followed at 17%, driven by its custom AI accelerators, while AMD captured 12%. Marvell, Alchip, Tesla, and smaller players round out the remaining share. This client concentration underscores why SK Hynix surged to the top—it has long been Nvidia’s primary HBM partner. By contrast, Samsung’s absence in major HBM contracts explains its loss of market share despite still shipping the highest total DRAM volumes.

Chart 4

The Hybrid Bonding Inflection Point

While SK hynix enjoys today’s crown, the next technological transition could reset leadership again. The industry is preparing to move from thermocompression bonding (TCB)—the current standard using microbumps—to hybrid bonding, which creates direct copper-to-copper connections.

As Table 1 shows, HBM4 is expected by 2026, with bandwidth of 5–6 TB/s per stack, and HBM4e following in 2027 with up to 7–8 TB/s. These figures dwarf HBM3e’s 1.2–1.5 TB/s. Hybrid bonding enables higher interconnect density, lower thermal budgets, and improved efficiency. This will be critical as AI accelerators push past current bandwidth ceilings. And Samsung’s future depends on getting there first.

Samsung’s 12-High HBM4 Gamble with Hybrid Bonding

Samsung’s 12-high HBM4 with wafer-to-wafer hybrid bonding could alter its fate for three reasons.

First are the density and thermal advantages. Hybrid bonding eliminates microbump height, reducing z-stack thickness and improving thermal dissipation. A bonded 12-hi stack has a lower form factor than a TCB-based stack, meaning Samsung could deliver higher density without as much penalty in heat or signal integrity.

Second is yield scaling. Microbump reliability and warpage become major challenges beyond 8-hi. SK hynix has been pushing 12-hi HBM3E with TCB but yields and packaging costs rise sharply. If Samsung’s bonding approach proves manufacturable at scale, it could bypass some of those yield cliffs.

Third is strategic timing. HBM4 adoption coincides with Nvidia’s Blackwell follow-ons and AMD’s next-gen Instinct. If Samsung can validate bonding at 12-hi for HBM4, it has a shot at re-entering the hyperscaler/Nvidia ecosystem at the generation transition, rather than trying to displace entrenched suppliers mid-cycle.

The risk is that bonding at this scale is unproven in HBM mass production. SK hynix has the advantage of entrenched customer trust, while Micron has already demonstrated fast ramps with HBM3E. Samsung has to prove it can deliver both volume and reliability. Its recent HBM track record does not inspire confidence that it can do this.

This is Samsung’s only realistic on-ramp to catch up. If it works, it could be the equalizer. If it slips, Samsung will remain boxed out by SK hynix and Micron.

Investor Takeaway

The DRAM market is undergoing its most significant restructuring since the 1990s. SK hynix has emerged as the immediate winner, translating its early and deep partnership with Nvidia into market leadership. Micron has quietly strengthened its position through HBM3E ramps, and as the sole U.S. DRAM maker, it has a geopolitical edge.

Samsung’s 8.8-percentage point share loss in just six months shows how quickly leadership can shift when execution falters. Yet investors should not dismiss Samsung’s long-term potential. Its pivot to hybrid bonding and roadmap for HBM4e could allow it to rebound just as bandwidth ceilings move higher.

Looking ahead, leadership will depend not only on DRAM shipment volume but also on who dominates hybrid bonding and secures the deepest AI partnerships. SK hynix holds the throne today, but Samsung is preparing its comeback, and Micron remains a credible challenger. In the most plausible scenario, Samsung’s HBM competitiveness will improve gradually over the next two to three years, setting the stage for another reshuffling of leadership.