Samsung’s HBM progress is real, but the larger 2027 question is whether broad memory pricing stays strong enough to turn that mix shift into sustained earnings.
Samsung Electronics’ high-bandwidth memory story is getting stronger. The harder question is whether that strength will be enough to offset a softer conventional-memory cycle in 2027.
That distinction matters because the latest Korean brokerage reports are not mainly arguing about whether Samsung can improve its HBM position. They are arguing about how long the broader memory pricing cycle lasts, and how much of the HBM mix improvement ultimately reaches group earnings.
The earnings debate is already wide
Mirae Asset Securities forecasts Samsung Electronics operating profit of about $385.9 billion in 2027, while BNK Investment & Securities forecasts about $280.6 billion. The gap is roughly 37.5% relative to BNK’s estimate.
Those are broker forecasts, not a consensus range or a probability-weighted outcome. But the difference is useful because it shows where the argument really sits: not in the existence of HBM demand, but in assumptions about DRAM pricing, demand elasticity and product mix.
HBM gives Samsung a stronger mix story
KB Securities estimates Samsung’s HBM market share at 33% in 2Q26 and expects it to approach 40% in 4Q26. It also expects HBM4 to exceed 60% of Samsung’s HBM revenue in the second half of 2026, with 3Q26 HBM4 revenue rising more than threefold quarter on quarter.
The same report argues that Samsung’s foundry execution is improving at the same time. KB says 4nm yield has stabilized above 80% and 2nm GAA yield has improved above 70%. Those figures are brokerage estimates rather than company-reported audited metrics, but they matter to the HBM thesis because Samsung’s advantage could come from more than DRAM alone.
If Samsung can combine memory, advanced logic for HBM base dies, foundry and packaging, the competitive question shifts from “who sells the memory stack?” to “who can optimize the whole system?” That is why foundry recovery can matter to HBM economics even before it becomes a large standalone profit contributor.
The bigger disagreement is DRAM pricing
The contrast between Mirae Asset and BNK is clearest in their DRAM ASP assumptions. Mirae Asset forecasts annual DRAM ASP growth of 25.3% in 2026 and 21.0% in 2027. BNK, by contrast, forecasts a very strong 2026 rebound but then an 8% decline in 2027.
The two brokers may use different product mixes and pricing bases, so the series should not be treated as perfectly harmonized. Still, the directional split is large enough to explain much of the difference in their 2027 earnings outlooks.
HBM can improve mix, margins and customer relevance. But Samsung still sells large volumes of conventional DRAM and NAND. If those products move into a weaker pricing phase, HBM gains have to work harder just to keep aggregate semiconductor earnings flat.
Near-term revisions show how quickly that can matter
BNK’s September report cut its 3Q26 operating-profit estimate by 5%, to about $78.2 billion, and its 4Q26 estimate by 2%, to about $85.6 billion. Those are estimate revisions, not reported results, but they illustrate the downside mechanism.
A small change in DRAM ASP, exchange rates or end-demand assumptions can offset part of the upside from HBM at the group level. BNK therefore does not need to reject the HBM thesis to arrive at a more cautious 2027 earnings path.
The practical takeaway is that Samsung’s 2027 debate has two layers. The first is execution: HBM4 ramp, custom-HBM progress and foundry yield. The second is cycle durability: whether conventional memory pricing remains strong enough for those execution gains to compound rather than merely cushion a downturn.
That makes the next phase less about proving that Samsung can participate in HBM, and more about proving that its improving product mix can outrun a potentially weaker memory cycle.
